Drill string mechanical parameter calculation method based on hybrid model and self-learning optimization

Abstract

The invention relates to the technical field of oil and gas drilling, and discloses a drill string mechanical parameter calculation method based on a hybrid model and self-learning optimization, and the method comprises the steps: 1, obtaining the engineering data of drilling, including a well depth parameter, a drill string parameter, a drilling parameter and real-time drilling data; 2, well track segmentation is conducted, and a mechanical model is selected based on the well deviation change rate Ka to calculate the normal constraint force of the current well depth; 3, correcting the friction factor by adopting a physical information neural network to obtain a friction factor correction amount; 4, a standard well database is established according to engineering data of well drilling, and the friction torque of a target well is calculated through a difference quantization and linear superposition method; and 5, performing multi-end early warning and regulation according to the friction torque of the target well. By means of the scheme, the friction torque prediction precision of the complex structure well is improved, and multi-end early warning is achieved.

Description

Technical Field

[0001] This invention relates to the field of oil and gas drilling technology, specifically to a method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization. Background Technology

[0002] In oil and gas drilling operations, friction and torque are key factors affecting drilling efficiency, cost, and safety as the drill string operates within the wellbore. Accurate prediction of friction and torque is crucial for optimizing drilling design, guiding field operations, and ensuring equipment safety. With the development of drilling technology, higher demands are placed on the accuracy and reliability of friction and torque calculations. Existing methods for calculating drill string friction and torque mainly suffer from the following technical problems: The trade-off between computational accuracy and efficiency: Existing computational methods, such as those based on the finite element method, while highly accurate, suffer from complex modeling and time-consuming calculations, making them unsuitable for real-time on-site analysis. Simplified models (flexible / rigid rod models), on the other hand, offer fast computation but lack sufficient accuracy.

[0003] Incomplete consideration of influencing factors: Existing methods often ignore key factors such as the coupling effect between cuttings bed and wellbore trajectory, resulting in large deviations in the calculation of friction coefficient.

[0004] Poor model adaptability: The single model in the existing technology is difficult to adapt to the changes in mechanical properties of different well sections (vertical well section, deviated well section, horizontal section) and lacks an intelligent switching mechanism.

[0005] Insufficient real-time parameter correction capability: Traditional methods lack a self-learning optimization mechanism based on real-time drilling data, and cannot dynamically correct key parameters such as friction factor. Summary of the Invention

[0006] This invention provides a method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization, which can achieve accurate prediction and risk warning of friction torque in wells with complex structures.

[0007] Therefore, the present invention provides the following technical solution: A method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization, the method comprising: Step 1: Obtain drilling engineering data, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data; Step 2: Segment the wellbore trajectory based on the well inclination rate. K a Select a mechanical model to calculate the normal constraint force at the current well depth; Step 3: Use a physical information neural network to correct the friction coefficient and obtain the friction coefficient correction amount; Step 4: Establish a standard well database based on the drilling engineering data, and calculate the friction torque of the target well using the difference quantization and linear superposition method; Step 5: Based on the friction torque of the target well, perform multi-terminal early warning and control.

[0008] Optionally, in step 1, the well depth parameters include well depth, well inclination angle, azimuth angle, and dogleg degree; the drill string parameters include drill string assembly, drill string specifications, and material properties; the drilling parameters include drilling pressure, rotational speed, displacement, hook load, and torque; and the real-time drilling data includes drilling pressure vibration spectrum, torque phase difference, and measured friction torque data.

[0009] Optionally, in step 2, the well inclination change rate is... K a A well section with a rate of change greater than 1° / m is defined as a high rate of change section. A soft rod model is used to calculate the normal constraint force at the current well depth, taking into account the bending deformation of the drill string. The rate of change of well inclination K a The well section with a rate of change of less than 0.3° / m is defined as the low rate of change section, and a rigid bar model is used to calculate the normal constraint force at the current well depth. The well inclination change rate is 0.3° / m≤ K a The well section with a radius of ≤1° / m is defined as the transition section. A weighted fusion model is used to calculate the normal constraint force at the current well depth.

[0010] Optionally, in step 2, during the transition section, the formula for calculating the normal constraint force at the current well depth using the weighted fusion model is as follows: ; In the formula, Weighting factor, dimensionless, derived from The contribution ratios of the flexible bar model and the rigid bar model were calculated. The mixed force is the normal constraint force at the current well depth, expressed in N, representing the weighted and fused comprehensive mechanical calculation result. Normal constraint force calculated for a soft-bar model; Normal constraint force calculated for a rigid bar model.

[0011] Optionally, in step 3, based on the drilling pressure vibration spectrum and torque phase difference in the real-time drilling data, a self-learning optimization module is introduced. The loss function of the physical information neural network includes torque prediction error and physical constraints to obtain the friction coefficient correction. The loss function is: 2 +0.01 2 ; In the formula, The loss function is dimensionless and is used to measure the overall error of the neural network's prediction of the physical information; the training process involves minimizing this error through an optimization algorithm. ; T actual This is the actual measured torque value; T pred The torque value predicted by the physical information neural network is a torque prediction value calculated by the physical information neural network based on the drill pressure vibration spectrum and torque phase difference in real-time drilling data. μ is the gradient of the friction coefficient μ, i.e., the friction coefficient correction amount; 0.01 is the regularization weight coefficient.

[0012] Optionally, step 4 includes: Step 41: Perform data preprocessing on the drilling engineering data to construct the structured standard well database; Step 42: Based on the standard well database, perform difference quantification and similarity analysis. First, find the reference well in the standard well database that is most similar to the target well in terms of characteristics. Then, determine the applicable weight of the reference well's calculation results by quantifying the differences between the two. Step 43: Calculate the friction torque of the target well using the linear superposition method to obtain the predicted curve of the friction torque of the entire well section of the target well, including axial friction and rotational torque. Step 44: The target well friction torque calculated using the linear superposition method is compared with the target well friction torque calculated using high-fidelity finite element analysis based on the accurate three-dimensional model of the target well to verify its accuracy. Step 45: Identify key risk points to obtain risk coefficients and calculate the accuracy report.

[0013] Optionally, in step 43, the frictional torque of the target well is calculated by using the frictional torque of reference wells at corresponding well depths in the standard well database, combined with the similarity weights between the target well and each reference well; the calculation formula is as follows: target( )=∑(j=1~m) ref, ( )+Δ correction; in: target( ) is the target well at the well depth Predicted friction torque at the location; ref, ( ) is the first The frictional torque at the corresponding well depth of the reference well; Indicates the first The weight of the calculation results of the reference well; Δ The correction term is based on the specific operating conditions of the target well, and its calculation formula is as follows: Δ correction= corr F ; corr is the change in the friction coefficient correction amount obtained through the physical information neural network relative to the average friction coefficient of the reference well; F n To determine the inclination rate of the target well K a Select the normal constraint force at the current well depth calculated by the mechanical model; The equivalent radius of the contact between the drill string and the wellbore is determined by the drill string parameters.

[0014] Optionally, in step 5, a friction profile and a buckling risk cloud map are generated based on the friction torque of the target well. A three-dimensional wellbore buckling cloud map is rendered on the web using WebGL. A compressed alarm message containing the risk location is pushed to the mobile terminal. Real-time data is transmitted to the well control system for closed-loop control of adjusting drilling pressure.

[0015] A drill string mechanical parameter calculation system based on a hybrid model and self-learning optimization, the system comprising: The engineering data acquisition unit acquires drilling engineering data, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data. The normal constraint force calculation unit segments the wellbore trajectory based on the well inclination rate. K a Select a mechanical model to calculate the normal constraint force at the current well depth; The friction coefficient correction unit uses a physical information neural network to correct the friction coefficient and obtain the friction coefficient correction amount; The friction torque calculation unit establishes a standard well database based on drilling engineering data and calculates the friction torque of the target well through differential quantification and linear superposition methods. The multi-terminal early warning and control unit performs multi-terminal early warning and control based on the friction torque of the target well.

[0016] A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to perform the steps of the drill string mechanical parameter calculation method based on a hybrid model and self-learning optimization.

[0017] The present invention provides a method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization. It selects the mechanical model based on the well inclination change rate and introduces a Physical Information Neural Network (PINN) to correct the friction factor, thereby improving the prediction accuracy of friction torque in wells with complex structures and enabling multi-terminal early warning. Compared with existing technologies, the present invention has the following significant advantages: Improved computational efficiency: Through a hybrid model switching mechanism, computational speed is increased by 3-5 times while maintaining accuracy; Significantly improved accuracy: The self-learning optimization module reduces the error in friction coefficient calculation to within 5%; Enhanced adaptability: It can automatically adapt to the characteristics of different well sections and solve the problem of calculation jumps when switching models; Real-time optimization: Supports real-time data access and rapid analysis on-site to meet the real-time decision-making needs of drilling projects; Risk warning capability: Integrates a multi-terminal early warning system to identify risks such as buckling and stuck drill in advance. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a drill string mechanical parameter calculation method based on a hybrid model and self-learning optimization in a specific embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a drill string mechanical parameter calculation system based on a hybrid model and self-learning optimization in a specific embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] like Figure 1 As shown, Figure 1 This is a flowchart of a drill string mechanical parameter calculation method based on a hybrid model and self-learning optimization in a specific embodiment of the present invention. The method includes: Step 1: Obtain well depth parameters, drill string parameters, drilling parameters, and real-time drilling data.

[0023] Well depth parameters include well depth, inclination angle, azimuth angle, and dogleg degree; drill string parameters include drill string assembly, drill string specifications, and material properties; drilling parameters include drilling pressure, rotational speed, displacement, hook load, and torque; real-time drilling data includes dynamic signals such as drilling vibration spectrum, torque phase difference, and measured friction torque data.

[0024] Step 2: Segment the wellbore trajectory based on the well inclination change rate. K a Choose a mechanical model to calculate the normal constraint force at the current well depth.

[0025] The rate of change in well inclination is the change in well inclination angle per unit length of well section, in order to... K a It is represented as the first derivative of the well inclination angle α with respect to the well depth L.

[0026] For high rate of change ( K a >1° / m): Using a soft rod model, the bending deformation of the drill string is considered as follows: ; in M b Let be the bending moment, and s be the well depth. K Where T is the wellbore curvature and T is the torque. F n For the normal constraint force, q m K represents the mass per unit length of the drill string. f It represents the equivalent stiffness when the drill string is in contact with the wellbore, or a parameter related to the friction model. α is the well inclination angle, which is the angle between the wellbore axis and the vertical direction.

[0027] For the low rate of change segment ( K a<0.3° / m): A rigid bar model is used, with normal constraint force. F n It consists of two parts: the gravitational component and the geometric effect caused by the axial force. The simplified calculation is as follows: ; In the formula, F n Wcosα is the normal constraint force; Wcosα is the axial component of gravity; ±μWsinα·sin(v) is the axial component of friction. The sign ± depends on the direction of motion (lifting or lowering), and sin(v) represents the proportion of the circumferential component of friction projected onto the axial direction.

[0028] 0.35 EI (d²θ / ds²) represents the additional axial force (or moment conversion effect) caused by wellbore curvature (dogleg). This value is used to supplement the calculation of additional forces generated by drill string bending deformation in curved sections (high d²θ / ds²). 0.35 is an empirical coefficient, derived from historical data or simulation calibration, used to balance the differences between theoretical models and actual operating conditions.

[0029] For the transition section (0.3° / m ≤ K a ≤ 1° / m): The weighted fusion model is as follows: ; Weighting factor, dimensionless, derived from The contribution ratios of the flexible bar model and the rigid bar model were calculated. The mixed force is the normal constraint force at the current well depth, expressed in N, representing the weighted and fused comprehensive mechanical calculation result. Normal constraint force calculated for a soft-bar model; Normal constraint force calculated for a rigid bar model.

[0030] Step 3: Use a physical information neural network (PINN) to correct the friction coefficient in real time.

[0031] Inputs: Drilling pressure vibration spectrum S(f) and torque phase difference Δ Drilling pressure vibration spectrum S(f) and torque phase difference Δ The two parameters are read from the real-time data acquisition and monitoring system at the drilling site.

[0032] Output: Friction coefficient correction amount Δμ, which is the friction coefficient μPINN output by the PINN network in real time.

[0033] Introducing a self-learning optimization module, the loss function of the physical information neural network includes torque prediction error and physical constraints: 2 +0.01 2 ; The loss function is dimensionless and measures the overall prediction error of a neural network. The training process involves minimizing this loss function using optimization algorithms (such as gradient descent). T actual The actual torque measurement value comes from real torque time series data from field sensors (such as top drive torque sensors). pred The torque value predicted by the neural network is determined by the PINN model based on inputs such as the drilling pressure vibration spectrum S(f) and the torque phase difference Δ. The torque prediction value calculated by (etc.). The gradient of the friction coefficient μ represents the rate of change of μ in space (e.g., well depth s) or time. This term serves as a physical regularization term, constraining μ to avoid drastic, non-physical abrupt changes and ensuring its spatial / temporal smoothness. 0.01 is the regularization weight coefficient, used to balance the relative importance of the data fitting term (first term) and the physical constraint term (second term). This value is typically determined through hyperparameter tuning or engineering experience.

[0034] ||T actual - T pred ||², the data fitting term, ensures that the torque predicted by the network is as close as possible to the measured value, and is the foundation of model accuracy. 0.01|| μ‖², a physical regularization term, introduces prior knowledge that the friction coefficient should change smoothly, avoiding the network from outputting an unphysical, violently oscillating μ curve in order to forcibly fit the data, thereby improving the model's generalization ability and stability.

[0035] Step 4: Calculate and verify the friction torque. Establish a standard well database, and quickly calculate the friction torque of the target well using differential quantization and linear superposition methods. Specifically, this includes: Step 41, establish a standard well database. First, collect and organize complete engineering data from multiple completed wells, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data, including: Well depth parameters: well depth, inclination angle, azimuth, dogleg degree; Drill string parameters: drill string assembly, specifications, and material properties; Drilling parameters: drilling pressure, rotation speed, displacement, hook load, torque; Measured friction torque data: used as a verification benchmark.

[0036] Then, the data is preprocessed: the data is normalized, denoised, and feature extracted to form a structured standard well database.

[0037] Step 42: Perform difference quantification and similarity analysis. To quickly calculate the friction torque of the target well, first, a "reference well" with characteristics most similar to the target well is found in the standard well database. By quantifying the differences between the two, the applicable weight of the reference well's calculation results is determined.

[0038] The formula for quantifying differences is: ;

[0039] in: For the target well and the first The overall difference of the reference wells (dimensionless, the smaller the value, the more similar). Representing the Several characteristic parameters (such as: build-up point depth, maximum well inclination angle, horizontal section length, drill string assembly similarity, etc.). ,targets It is the value of the target well on the i-th feature parameter. P i,ref Let be the value of the j-th reference well on the i-th characteristic parameter. The weight of this parameter is determined through historical data regression analysis or expert experience. These are the maximum and minimum values ​​of the parameter in the database, used for normalization.

[0040] The formula for calculating similarity weight is: ;

[0041] In the formula, Indicates the first The weights of the calculation results from the reference wells are used for subsequent overlay calculations. Δ j The overall difference between the target well and the j-th reference well. Δ k The overall difference between the target well and the k-th reference well.

[0042] Step 43: The linear superposition method is used to quickly calculate the friction torque of the target well and obtain the predicted friction torque curve of the entire well section of the target well (including axial friction and rotational torque).

[0043] Using known calculation results from reference wells in the standard well database (these results have been obtained using this method or the finite element method), and combining the similarity weights between the target well and each reference well, the frictional torque of the target well is quickly calculated. The calculation formula is as follows: target( )=∑(j=1~m) ref, ( )+Δ correction; in: target( The target well is at a depth of [missing information]. Predicted frictional torque at the location (N·m); ref, ( ) is the first The frictional torque of the reference well at the corresponding well depth can be quickly calculated using its well depth parameters and the hybrid model from step 2: T ref,j (s)= μ ref,j ( s ) F n,ref,j (s) R j ; μ ref,j ( s ): Reference well depth s The coefficient of friction at the point (dimensionless); F n,ref,j (s): Normal contact force (N / m) calculated from step 2; R j The equivalent friction radius (m) of the reference well drill string in contact with the well wall is calculated from the drill string specifications (such as the outer diameter of the drill pipe joint, the outer diameter of the drill collar, etc.).

[0044] Δ The correction term is based on the specific operating conditions of the target well, and its calculation formula is as follows: Δ correction= corr F n ; corr is the change in the friction factor μPINN output by the PINN network in step 3 relative to the average friction factor μref,avg of the reference well (i.e.: corr=μPINN-μref,avg); F nIn step 2, based on the actual wellbore structure of the target well (well inclination rate) K a The normal constraint force at the current well depth calculated by the selected mechanical model (soft rod, rigid rod, or hybrid model); The equivalent radius (m) of the contact between the drill string and the well wall is determined by the drill string parameters (step 1).

[0045] Step 44: Compare the results with the finite element calculation to verify the accuracy.

[0046] To ensure the reliability of this rapid calculation method, its prediction results are compared with those of high-fidelity finite element analysis (FEA) based on an accurate three-dimensional model of the target well.

[0047] The verification metrics are: ; T FEA (s) The torque of the target well at a depth s in finite element analysis.

[0048] The accuracy requirement is: through the self-learning optimization in step 3 and the superposition correction in step 4, ensure that the relative error of the predicted torque for the entire well section is stable within 5%.

[0049] The verification process involves selecting key feature points of the target well (such as the build-up point, target point, and dogleg abrupt change point) for comparison. This includes: 1) Output torque-depth comparison curves for the entire well section.

[0050] 2) If the error exceeds the threshold, it is fed back to step 3 to further optimize the training data of the PINN network or adjust the model parameters.

[0051] Step 45: Identify key risk points and calculate the accuracy report.

[0052] Key risk points were identified by combining the model output from step 2 with the torque curve, identifying risk locations including high friction and high torque well sections.

[0053] The calculation accuracy report includes a comparative analysis with the finite element results and error statistics, demonstrating the effectiveness and reliability of this method.

[0054] Step 5: Perform multi-terminal early warning and control based on friction and torque. Calculate parameters such as friction and torque to generate a friction profile and a buckling risk cloud map. The buckling risk cloud map is generated based on load parameters (including torque and drilling pressure), wellbore parameters (including wellbore curvature), and friction parameters (including real-time friction).

[0055] Web-based: Renders 3D wellbore buckling cloud maps using WebGL, supporting interactive viewing; Mobile devices: Push compressed alarm messages (<5KB) via MQTT protocol, including the risk location and buckling risk factor. R b Key information such as values; buckling risk coefficient R b This is the ratio of the current axial pressure to the critical buckling load.

[0056] Well control system: Transmits real-time data via the OPC UA protocol and supports closed-loop control for automatic adjustment of drilling pressure. Real-time data includes pressure data (standby pressure, casing pressure, etc.), mud / circulation system data (inlet and outlet flow rates, etc.), drilling condition data (drilling speed, torque, etc.), and downhole data (annular pressure, etc.).

[0057] The present invention provides a method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization. First, all necessary input parameters are obtained; then, according to… K a Intelligent selection of mechanical model to calculate accurate normal constraint force F n This forms the basis for torque correction. Then, a high-precision friction coefficient is used for real-time correction. PINN This significantly improves the accuracy of torque calculation. Finally, it provides precise mechanical parameters and risk indicators, driving multi-terminal early warning and optimized control.

[0058] In one embodiment, this method is applied in a three-section horizontal well (vertical depth 2500m, horizontal section length 800m) in an oilfield: in the build-up section ( K a =1.5° / m) adopts a soft rod model; stable slope section ( K a =0.2° / m) A rigid bar model was adopted; the friction coefficient was corrected in real time through PINN network, and the error between the final calculated torque and the measured value was only 3.2%.

[0059] In another specific embodiment of the present invention, risk warning for extended reach wells is performed. During the drilling of a certain extended reach well (water-to-vertical ratio 3.2), the system monitors the buckling risk coefficient in real time. R b When the value reaches 0.85, an early warning is triggered in time and the drilling pressure parameters are automatically adjusted, successfully avoiding a stuck drill accident.

[0060] Accordingly, embodiments of the present invention also provide a drill string mechanical parameter calculation system based on a hybrid model and self-learning optimization, such as... Figure 2 The diagram shown is a structural schematic of the system. This drill string mechanical parameter calculation system based on a hybrid model and self-learning optimization includes the following modules: The engineering data acquisition unit 201 acquires drilling engineering data, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data. Normal constraint force calculation unit 202 performs wellbore trajectory segmentation based on well inclination rate. K a Select a mechanical model to calculate the normal constraint force at the current well depth; Friction coefficient correction unit 203 uses a physical information neural network to correct the friction coefficient and obtain the friction coefficient correction amount; The friction torque calculation unit 204 establishes a standard well database based on drilling engineering data and calculates the friction torque of the target well through differential quantification and linear superposition method. The multi-terminal early warning and control unit 205 performs multi-terminal early warning and control based on the friction torque of the target well.

[0061] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0062] The present invention also provides a storage medium, which is a computer-readable storage medium storing a computer program thereon, the computer program being executable when it runs. Figure 1 The method shown may include some or all of the steps. The storage medium may include read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc. The storage medium may also include non-volatile memory or non-transitory memory, etc.

[0063] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data provider to another website, computer, server, or data provider via wired or wireless means.

[0064] The embodiments of the present invention have been described in detail above. Specific implementation methods have been used to illustrate the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the methods and systems of the present invention, and are merely some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention, and the content of this specification should not be construed as a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization, characterized in that, The method includes: Step 1: Obtain drilling engineering data, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data; Step 2: Segment the wellbore trajectory based on the well inclination rate. K a Select a mechanical model to calculate the normal constraint force at the current well depth; Step 3: Use a physical information neural network to correct the friction coefficient and obtain the friction coefficient correction amount; Step 4: Establish a standard well database based on the drilling engineering data, and calculate the friction torque of the target well using the difference quantization and linear superposition method; Step 5: Based on the friction torque of the target well, perform multi-terminal early warning and control.

2. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 1, characterized in that, In step 1, the well depth parameters include well depth, inclination angle, azimuth angle, and dogleg degree; the drill string parameters include drill string assembly, drill string specifications, and material properties; the drilling parameters include drilling pressure, rotational speed, displacement, hook load, and torque; and the real-time drilling data includes drilling pressure vibration spectrum, torque phase difference, and measured friction torque data.

3. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 1, characterized in that, In step 2, the well inclination change rate K a A well section with a rate of change greater than 1° / m is defined as a high rate of change section. A soft rod model is used to calculate the normal constraint force at the current well depth, taking into account the bending deformation of the drill string. The rate of change of well inclination K a The well section with a rate of change of less than 0.3° / m is defined as the low rate of change section, and a rigid bar model is used to calculate the normal constraint force at the current well depth. The well inclination change rate is 0.3° / m ≤ K a Well sections ≤ 1° / m are defined as transition sections, and a weighted fusion model is used to calculate the normal constraint force at the current well depth.

4. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 3, characterized in that, In step 2, during the transition section, the formula for calculating the normal constraint force at the current well depth using the weighted fusion model is as follows: ； In the formula, Weighting factor, dimensionless, derived from The contribution ratios of the flexible bar model and the rigid bar model were calculated. The mixed force is the normal constraint force at the current well depth, expressed in N, representing the weighted and fused comprehensive mechanical calculation result. Normal constraint force calculated for a soft-bar model; Normal constraint force calculated for a rigid bar model.

5. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 1, characterized in that, In step 3, based on the drilling pressure vibration spectrum and torque phase difference in the real-time drilling data, a self-learning optimization module is introduced. The loss function of the physical information neural network includes torque prediction error and physical constraints to obtain the friction coefficient correction. The loss function is: 2 +0.01 2 ； In the formula, The loss function is dimensionless and is used to measure the overall error of the neural network's prediction of the physical information; the training process involves minimizing this error through an optimization algorithm. ; T actual This is the actual measured torque value; T pred The torque value predicted by the physical information neural network is a torque prediction value calculated by the physical information neural network based on the drill pressure vibration spectrum and torque phase difference in real-time drilling data. μ is the gradient of the friction coefficient μ, i.e., the friction coefficient correction amount; 0.01 is the regularization weight coefficient.

6. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 1, characterized in that, Step 4 includes: Step 41: Perform data preprocessing on the drilling engineering data to construct the structured standard well database; Step 42: Based on the standard well database, perform difference quantification and similarity analysis. First, find the reference well in the standard well database that is most similar to the target well in terms of characteristics. Then, determine the applicable weight of the reference well's calculation results by quantifying the differences between the two. Step 43: Calculate the friction torque of the target well using the linear superposition method to obtain the predicted curve of the friction torque of the entire well section of the target well, including axial friction and rotational torque. Step 44: The target well friction torque calculated using the linear superposition method is compared with the target well friction torque calculated using high-fidelity finite element analysis based on the accurate three-dimensional model of the target well to verify its accuracy. Step 45: Identify key risk points to obtain risk coefficients and calculate the accuracy report.

7. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 6, characterized in that, In step 43, the friction torque of the target well is calculated by using the friction torque of the reference wells at the corresponding well depth in the standard well database and combining the similarity weights between the target well and each reference well. The calculation formula is: target( )=∑(j=1~m) ref, ( )+Δ correction; in: target( The target well is at a depth of [missing information]. Predicted friction torque at the location; ref, ( ) is the first The frictional torque at the corresponding well depth of the reference well; Indicates the first The weight of the calculation results of the reference well; Δ The correction term is a modification based on the specific operating conditions of the target well, and its calculation formula is as follows: D correction= correct F ; corr is the change in the friction coefficient correction amount obtained through the physical information neural network relative to the average friction coefficient of the reference well; F n To determine the inclination rate of the target well K a Select the normal constraint force at the current well depth calculated by the mechanical model; The equivalent radius of the contact between the drill string and the wellbore is determined by the drill string parameters.

8. The method for calculating drill string mechanical parameters based on a hybrid model and self-learning optimization according to claim 6, characterized in that, In step 5, based on the frictional torque of the target well, a frictional profile map and a buckling risk cloud map are generated. A three-dimensional wellbore buckling cloud map is rendered on the web using WebGL. A compressed alarm message containing the risk location is pushed to the mobile terminal. Real-time data is transmitted to the well control system for closed-loop control of adjusting drilling pressure.

9. A drill string mechanical parameter calculation system based on a hybrid model and self-learning optimization, characterized in that, The system includes: The engineering data acquisition unit acquires drilling engineering data, including well depth parameters, drill string parameters, drilling parameters, and real-time drilling data. The normal constraint force calculation unit segments the wellbore trajectory based on the well inclination rate. K a Select a mechanical model to calculate the normal constraint force at the current well depth; The friction coefficient correction unit uses a physical information neural network to correct the friction coefficient and obtain the friction coefficient correction amount; The friction torque calculation unit establishes a standard well database based on drilling engineering data and calculates the friction torque of the target well through differential quantification and linear superposition methods. The multi-terminal early warning and control unit performs multi-terminal early warning and control based on the friction torque of the target well.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it performs the steps of the drill string mechanical parameter calculation method based on hybrid model and self-learning optimization as described in any one of claims 1 to 8.

Images