Well drilling trajectory real-time dynamic tracking control method and system based on deep learning

Abstract

The invention discloses a drilling trajectory real-time dynamic tracking control method and system based on deep learning, and relates to the technical field of petroleum drilling, and the method comprises the steps: generating well trajectory control parameters according to the basic information of a well and preset well trajectory data; setting an optimal drilling fluid performance parameter corresponding to the current working condition according to the stratum environment where the drilling tool is located; deep learning is conducted on the well track control parameters and the optimal drilling fluid performance parameters through the correlation model, a drilling parameter setting scheme is generated, and an underground drilling tool is controlled to conduct drilling; comparing the drilling data with data in the drilling parameter setting scheme, and determining deviation in the drilling data; and determining a corresponding fault mode, and determining a corresponding solution from the solution library to correct the drilling parameter setting scheme. According to the method, deep learning is carried out on geological conditions and historical data, the well trajectory control model and other models are constructed, the model can be continuously updated and optimized along with data accumulation, and the drilling trajectory tracking and control precision is continuously improved.

Description

Technical Field

[0001] The present application relates to the field of oil drilling technology, and in particular to a method and system for real-time dynamic tracking control of drilling trajectories based on deep learning. Background Art

[0002] Drilling trajectory tracking refers to the technology that monitors and controls the drill bit's path in real time during the drilling process, ensuring it closely matches the pre-designed wellbore trajectory. In petroleum engineering, drilling trajectory tracking and control technology is crucial for drilling quality, recovery efficiency, and capital investment.

[0003] Currently, global oil and gas exploration is trending toward ultra-deepwater, ultra-deep formations, low permeability, and unconventional drilling. This presents significant challenges for drilling trajectory tracking and control technology. As the oil and gas industry continues to demand higher recovery rates and higher drilling costs, traditional drilling trajectory tracking and control technology is no longer able to meet the trajectory accuracy and drilling efficiency requirements of complex reservoirs. Breakthroughs in drilling trajectory tracking and control technology are urgently needed. Research on drilling trajectory tracking and control technology is currently focused on horizontal wells, directional wells, multilateral wells, and multi-bottom wells. Theoretically, research on drill string mechanics and drill tool assemblies is relatively mature, and various directional drilling technologies have also achieved considerable progress. However, with the increasing complexity of oil and gas reservoirs, drilling trajectory tracking and control technology faces significant challenges due to the high temperature, high pressure, and steepness of these reservoirs. Furthermore, issues such as nonlinearity, strong interference, high coupling, hysteresis, and time variability present during the drilling process, presenting numerous challenges for drilling trajectory tracking and control technology.

[0004] Patents CN111810113A and CN110500081A describe establishing a rotary steerable drilling database, recording actual or simulated drilling data, forming a drilling model based on the drilling database, inputting operational well data into the drilling model, providing drilling tool assembly recommendations for the operational well, and outputting a drilling plan for the operational well. The model then measures wellbore trajectory data in real time, evaluates tool status, identifies the degree of wellbore trajectory compatibility with the design, plans wellbore control plans, allocates the latest control instruction drilling sections, and adjusts drilling parameters. However, neither patent describes updating data to the model in real time during drilling.

[0005] As drilling progresses, new formation information and drilling conditions are constantly changing. If these new data are not incorporated into model training in a timely manner, the accuracy of the model will decrease over time. At the same time, the trained drilling model mixes all drilling data together for training, and interference may occur between different types of data, making it difficult for the model to accurately learn the key features of each type of data. Summary of the Invention

[0006] An embodiment of the present application provides a real-time dynamic tracking and control method and system for drilling trajectories based on deep learning, which is used to solve the problems in the above-mentioned prior art that the drilling model established by the rotary steerable drilling system based on deep learning continues to change with the formation information and drilling conditions, resulting in a decrease in the accuracy of the model, the drilling model is difficult to accurately learn the key features of each data, and there are no processing measures for abnormal fluctuations in the measurement data of downhole sensors.

[0007] On the one hand, an embodiment of the present application provides a real-time dynamic tracking control method for drilling trajectories based on deep learning, the method comprising: S1. Generate wellbore trajectory control parameters based on real-time acquired well basic information and preset wellbore trajectory data using a pre-built wellbore trajectory control model; and set optimal drilling fluid performance parameters corresponding to the current operating conditions based on the formation environment in which the drilling tool is located using a pre-built drilling fluid formation parameter coupling model; S2. Deeply learn the wellbore trajectory control parameters and optimal drilling fluid performance parameters through a pre-built correlation model, generate a drilling parameter setting plan, and control the downhole drilling tool drilling according to the drilling parameter setting plan; S3. Using a pre-built multi-index fusion evaluation model, the real-time acquired drilling data is compared with the data in the drilling parameter setting plan to determine whether there is any deviation in the drilling data; S4. Determine the failure mode corresponding to the deviation from a pre-built failure mode library, and determine the corresponding solution from a solution library based on the failure mode to modify the drilling parameter setting solution.

[0008] Preferably, before step S1, the following steps are included: Constructing a deep learning database based on historical data of the area to be drilled, where the types of historical data include one or more of the following: geographic location of the well, well depth, drilling fluid performance parameters, wellbore trajectory data, drilling tool type, drilling tool mechanical parameters, basic formation parameters of the well to be drilled, fault type, fault characteristics, and solutions; Perform data refinement processing on the data in the deep learning database, including data cleaning, data deduplication, data standardization, data anomaly processing and noise reduction.

[0009] Preferably, the method further comprises: Conduct deep learning on wellbore trajectory data and drilling tool mechanical parameters to build a wellbore trajectory control model; Conduct deep learning on basic formation parameters and drilling fluid performance parameters to build a drilling fluid-formation parameter coupling model; Conduct deep learning on wellbore trajectory data and drilling fluid performance parameters to build correlation models; Conduct deep learning on historical data to build a multi-indicator fusion evaluation model.

[0010] Preferably, determining whether the drilling data has a deviation in step S3 includes: The wellbore trajectory data acquired in real time exceeds the wellbore trajectory data in the drilling parameter setting plan; and / or the drill tool mechanical data acquired in real time exceeds the drill tool mechanical data in the drilling parameter setting plan; and / or the drilling fluid performance data acquired in real time exceeds the drilling fluid performance data in the drilling parameter setting plan.

[0011] Preferably, before step S4, the method further includes: Constructing a fault mode library based on historical data and empirical data, wherein the fault mode library includes: historical fault types and deviation characteristics corresponding to the historical fault types; Build a solution library based on pre-set rule base and optimization algorithm; Determining the failure mode corresponding to the deviation in step S4 includes: Match the characteristics of the deviation with the failure modes in the failure mode library and identify the most similar failure mode; Obtain corresponding solutions from the solution library based on the failure mode.

[0012] Preferably, the method further comprises: using the drilling data acquired in real time as new samples to update the parameters of the wellbore trajectory control model and the drilling fluid formation parameter coupling model.

[0013] On the other hand, the embodiment of the present application also provides a real-time dynamic tracking and control system for drilling trajectories based on deep learning, the system comprising: A database for storing historical data of the area to be drilled; The model training module is used to conduct deep learning on the data in the database to build a wellbore trajectory control model, a drilling fluid formation parameter coupling model, and a multi-index fusion evaluation model. It also conducts deep learning on the data parameters output by the wellbore trajectory control model and the drilling fluid formation parameter coupling model to build an association model. The drilling parameter generation module is used to generate a drilling parameter setting plan based on the basic information of the well, the preset well trajectory data and the formation environment where the drilling tool is located through the well trajectory control model, the drilling fluid formation parameter coupling model and the association model; The drilling data verification module is used to compare the real-time acquired drilling data with the data in the drilling parameter setting plan to determine whether there is any deviation; The drilling parameter optimization module is used to obtain corresponding solutions and optimize the drilling parameter setting plan when drilling data deviates.

[0014] Preferably, the drilling parameter verification module includes: Real-time monitoring unit, used to obtain drilling data in real time; The comparison unit is used to determine that the drilling data has a deviation when the drilling data exceeds the parameters in the drilling parameter setting plan.

[0015] Preferably, the drilling parameter optimization module includes: A fault pattern matching unit is used to match the deviation characteristics with the fault patterns in the preset fault pattern library, identify the most similar fault pattern, and obtain the corresponding solution from the preset solution library based on the fault pattern; The parameter optimization unit is used to integrate the solution into the drilling parameter setting plan for correction.

[0016] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed, it implements the above-mentioned deep learning-based real-time dynamic tracking and control method for drilling trajectories.

[0017] The real-time dynamic tracking method and system for drilling trajectories based on deep learning in this application have the following advantages: Through deep learning training of complex and changeable geological conditions and a large amount of historical multi-source data from different sensors and geology, processing models such as wellbore trajectory control model, drilling fluid formation parameter coupling model, multi-index fusion evaluation model and correlation model are constructed. Real-time feedback of logging data gives the processing model self-learning ability, which can continuously update and optimize the model as data accumulates, continuously improving the drilling trajectory tracking and control accuracy without frequent manual reprogramming or algorithm adjustment. At the same time, the multi-index fusion evaluation model provides reliable guidance for drilling operations, and can simultaneously consider multiple data states such as wellbore trajectory, drilling equipment status, drilling fluid performance, etc., and make comprehensive analysis and decisions to formulate more scientific and reasonable trajectory tracking plans, thereby improving overall efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 A flowchart of a method for real-time dynamic tracking of drilling trajectories based on deep learning provided in an embodiment of the present application.

[0020] Figure 2 A schematic diagram of the structure of a real-time dynamic tracking system for drilling trajectories based on deep learning provided in an embodiment of the present application.

[0021] Figure 3Schematic diagram of a real-time dynamic tracking system for drilling trajectories based on deep learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0023] refer to Figure 1 The figure shows a flow chart of a method for real-time dynamic tracking of drilling trajectories based on deep learning provided by an embodiment of the present application. The method for real-time dynamic tracking and control of drilling trajectories based on deep learning provided by an embodiment of the present application includes: S1. Generate wellbore trajectory control parameters based on the real-time acquired well basic information and preset wellbore trajectory data through a pre-built wellbore trajectory control model, and set the optimal drilling fluid performance parameters corresponding to the current working conditions based on the formation environment where the drill tool is located through a pre-built drilling fluid formation parameter coupling model.

[0024] In this embodiment, before drilling begins, a pre-built wellbore trajectory control model is first used to generate wellbore trajectory control parameters, such as the weight-on-bit adjustment value, steering tool deflection angle, and rotational speed optimization recommendations, based on real-time acquired basic well information, such as current well depth, formation lithology, and wellbore temperature, as well as preset wellbore trajectory data, such as target well inclination, azimuth, and target point coordinates. When the drill enters a high-pressure formation, the drilling fluid density needs to be increased to balance the formation pressure and prevent blowouts or wellbore collapse. Therefore, a pre-built drilling fluid-formation parameter coupling model is used to set optimal drilling fluid performance parameters for the current operating conditions, based on the drill's current bottom environment, such as current formation lithology, porosity, and fracture pressure, as well as density and viscosity ranges set based on formation stability requirements. For example, "increase the drilling fluid density to 1.25 g / cm³ and adjust the viscosity to 45 mPa·s."

[0025] S2. Deeply learn the wellbore trajectory control parameters and the optimal drilling fluid performance parameters through a pre-built association model to generate a drilling parameter setting plan. The drilling parameter setting plan includes but is not limited to: wellbore trajectory control instructions, drill bit mechanical parameter setting instructions, drilling pump operating parameters, and drilling fluid performance parameters.

[0026] In this embodiment, a pre-built correlation model is used to perform deep learning data analysis on wellbore trajectory control parameters such as the bit pressure adjustment value, steering tool deflection angle, and speed optimization suggestions, as well as optimal drilling fluid performance parameters such as drilling fluid density and viscosity. This generates a drilling parameter setting plan, which includes: wellbore trajectory control instructions, such as maintaining the bit pressure at 85 kN and adjusting the tool face angle to 12°; drill bit mechanical parameter setting instructions, such as increasing the speed to 120 RPM (revolutions per minute) and limiting the torque to 50 kN·m; drilling pump operating parameters, such as maintaining the pump pressure at 20 MPa and adjusting the displacement to 1.2 m³ / min; and drilling fluid performance parameters, such as density of 1.2 g / cm³, viscosity of 40 mPa·s, and shear force of 4 Pa.

[0027] S3. Compare the real-time acquired drilling data with the data in the drilling parameter setting plan through a pre-built multi-index fusion evaluation model to determine whether there is any deviation in the drilling data.

[0028] In this embodiment, during the drilling process, real-time measurement of drilling data, including current wellbore trajectory data, drill tool mechanical data, and drilling fluid performance data, is also required to comprehensively determine whether the current drilling data is biased. For example, measurement while drilling (MWD) tools are used to obtain precise well inclination, azimuth, and tool face angle data, as well as real-time well depth information recorded by well depth measurement sensors. Geosteering instruments and logging equipment are used to collect formation parameters such as lithology, porosity, permeability, and formation pressure. Drilling fluid performance monitoring instruments are used to obtain drilling fluid density measured by a densitometer, drilling fluid viscosity measured by a viscometer, drilling fluid shear force measured by a shear force meter, and drilling fluid fluid loss measured by a fluid loss meter. Drill tool mechanical data sensors are used to obtain drilling pressure data measured by a pressure sensor installed on the drill tool, drill tool rotational speed recorded by a rotational speed sensor, and torque data monitored by a torque sensor. In this embodiment, before comparison, the collected data must be cleaned to remove noise, outliers, and missing values. Data of different types and ranges must be normalized to the same scale.

[0029] By comparing the real-time measured current wellbore trajectory data, drill tool mechanical data, and drilling fluid performance data with the data in the drilling parameter setting plan output by the above-mentioned correlation model, when the real-time measured data is inconsistent with the data in the drilling parameter setting plan or the difference reaches a certain threshold, the drilling data is judged to be deviated.

[0030] In this embodiment, the wellbore trajectory data such as the well inclination angle, azimuth angle, and well depth obtained by real-time measurement are compared with the parameters in the wellbore trajectory control instructions output by the associated model. For example, when the difference between the actual well inclination angle and the designed well inclination angle exceeds the specified error range (for example, ±0.5°), or the actual azimuth angle deviates from the designed azimuth angle to a certain extent (for example, ±1°), it can be determined that the wellbore trajectory data has a deviation.

[0031] In this embodiment, real-time measured drill tool mechanical data, such as weight on bit (WOB), rotational speed, and torque, are compared with parameters in the drill tool mechanical parameter setting instructions output by the correlation model. For example, if WOB suddenly fluctuates significantly, deviating from the normal operating range (e.g., WOB is set at 80-100 kN, but the actual WOB momentarily drops below 60 kN or exceeds 120 kN); or if torque increases abnormally, exceeding the drill tool's normal tolerance (e.g., normal torque is 30-50 kN·m, but the actual torque reaches 70 kN·m or above), these indicate abnormal drill tool operating conditions and drilling data deviations. In this embodiment, real-time measured drilling fluid properties, such as density, viscosity, and shear force, are compared with standard values, determined based on formation conditions and drilling process requirements, from the drilling fluid performance parameters output by the correlation model. For example, if the drilling fluid density deviates from the standard density by more than ±0.05 g / cm³, the viscosity fluctuates by more than ±5 mPa・s, or the shear force changes significantly (e.g., the standard shear force is 3-5 Pa, but the actual shear force is lower than 2 Pa or higher than 6 Pa), then this indicates that the drilling fluid performance data has deviated.

[0032] S4. If yes, determine the failure mode corresponding to the deviation, and modify the drilling parameter setting plan using the solution corresponding to the failure mode.

[0033] In this embodiment, after determining that the drilling data has deviated, the most similar fault mode is found by comparing the type, magnitude, change trend and other characteristics of the deviation. Furthermore, a suitable solution is selected from a pre-established solution library and integrated into the drilling parameter setting plan. The drilling parameters are adjusted and the problems that arise are corrected in real time.

[0034] In an embodiment of the present application, before drilling operations begin, a database for deep learning is constructed based on historical well logging data, simulated drilling data, historical geological exploration data, and historical fault information in the area to be drilled. The constructed database contains at least the following information: the geographic location of the well, well depth, drilling fluid performance parameters, wellbore trajectory data, drill tool type, drilling tool mechanical data, and basic formation parameters to be drilled. Drilling fluid performance parameters include: drilling fluid density, viscosity, pH value, temperature, and conductivity, etc., and data can be obtained through a drilling fluid performance monitoring system established during the drilling process. Drill tool mechanical parameters include: the bit pressure applied to the drill bit, the rotation speed of the drill string and drill bit, and the torque during drill string rotation. Wellbore trajectory data includes: well inclination angle, azimuth angle, and tool face angle. Drill tool types include: push-type rotary steerable drilling tools, pointing rotary steerable drilling tools, etc. The basic formation parameters of the well to be drilled include: gamma ray data, resistivity data, formation pore pressure data, formation rock fracture pressure data; fault information includes: fault type, fault characteristics and solutions, etc.

[0035] In this embodiment, to facilitate deep learning model training, the data in the deep learning database also undergoes data refinement, including data cleaning, data deduplication, data standardization, data anomaly processing, and noise reduction. For example, the database includes data integrity checks for well geolocation, drilling tool mechanical parameters, and wellbore trajectories, and missing values are filled. For a small number of missing values, an appropriate interpolation method can be selected based on the data's characteristics. Data with different characteristics can be standardized to have similar scales and distributions.

[0036] In an embodiment of the present application, deep learning is used to train wellbore trajectory parameters, drill tool types, and drill tool mechanical parameters to obtain a wellbore trajectory control model; deep learning is used to train drilling fluid performance parameters and basic formation parameters to be drilled to obtain a drilling fluid-formation parameter coupling model; and all data are learned and analyzed to obtain a multi-index fusion evaluation model. In this embodiment, an incremental learning method is used to update the wellbore trajectory control model and the drilling fluid-formation parameter coupling model. When new real-time data arrives, the model adapts to the new data characteristics by adjusting the model parameters based on the original knowledge. For example, the real-time drilling data obtained is used as a new sample to update the parameters of the wellbore trajectory control model and the drilling fluid-formation parameter coupling model.

[0037] In an embodiment of the present application, data analysis and deep learning are performed on the optimal control strategy and control instructions calculated based on the desired wellbore trajectory from the wellbore trajectory control model, including weight on bit, rotational speed, and steering tool parameters; as well as drilling fluid performance parameters such as density, viscosity, and shear force, which are dynamically adjusted based on real-time formation parameters, obtained from the drilling fluid-formation parameter coupling model. A correlation model between the wellbore trajectory control parameters and the drilling fluid performance parameters is established to quantify the degree of their mutual influence. Based on the goal of real-time dynamic tracking and control of the drilling trajectory during drilling operations, a comprehensive objective function is determined. The objective function takes a weighted sum of multiple indicators such as mechanical penetration rate, drilling fluid parameters, and wellbore trajectory deviation. An optimization algorithm is used to comprehensively optimize the wellbore trajectory control parameters and drilling fluid performance parameters, while satisfying various constraints (such as the drilling tool load capacity and formation pressure limits), to determine an optimal set of drilling parameter combinations.

[0038] Based on the optimized drilling parameter combination, a detailed drilling parameter setting plan is formulated, including specific parameter values such as bit pressure, rotation speed, steering tool settings, drilling fluid density, viscosity, shear force, as well as the parameter adjustment range and adjustment timing.

[0039] In an embodiment of the present application, a fault pattern library is constructed based on historical and empirical data. The fault pattern library includes historical fault types and corresponding deviation characteristics. A solution library is constructed based on a pre-defined rule library and optimization algorithm. To facilitate finding appropriate solutions based on fault patterns, the data in the fault pattern library and the solution library are correlated. This allows for the identification of deviations in real-time drilling data, enabling the identification of corresponding solutions based on the characteristics of the deviation.

[0040] refer to Figure 2 and Figure 3 The real-time dynamic tracking and control system of drilling trajectory based on deep learning provided in an embodiment of the present application includes: a database, a model training module, a drilling parameter generation module, a drilling data verification module and a drilling parameter optimization module.

[0041] In an embodiment of the present application, the system stores existing historical data of the area to be drilled, such as historical well logging data, simulated drilling data, historical geological exploration data, and historical fault information, in a database for subsequent deep learning. In this embodiment, the stored data should include the geographical location of the well, well depth, drilling fluid performance parameters, wellbore trajectory data, drill tool type, drilling tool mechanical parameters, and basic formation parameters to be drilled. Among them, the drilling fluid performance parameters include: drilling fluid density, viscosity, pH value, temperature and conductivity, etc., and the data can be obtained through the drilling fluid performance monitoring system established during the drilling process. The drilling tool mechanical parameters include: the drilling pressure applied to the drill bit, the rotation speed of the drill string and the drill bit, and the torque when the drill string rotates. The wellbore trajectory data include: well inclination angle, azimuth angle and tool face angle. The drilling tool types include: push-type rotary steerable drilling tools, pointing rotary steerable drilling tools, etc. Basic formation parameters for the well to be drilled include gamma ray data, resistivity data, formation pore pressure data, and formation rock fracture pressure data. Fault information includes fault type, fault characteristics, and solutions. In this embodiment, after historical data is stored in the database, the database is also checked for completeness, including well location, drilling tool mechanical parameters, and wellbore trajectory data, to fill in missing values. For a small number of missing values, an appropriate interpolation method can be selected based on the data characteristics. Data with different characteristics are normalized to have similar scale and distribution.

[0042] The system uses a model training module to conduct deep learning on the data stored in the database. According to the different types of parameters that need to be set in the rotary steerable drilling system, the system classifies and trains the model for data such as the mechanical data of the drill tool, wellbore trajectory control data, drilling fluid performance parameters, and formation parameters.

[0043] In one embodiment, wellbore trajectory parameters, drill tool type, and drill tool mechanical parameters are used as input features, and a wellbore trajectory control model is constructed through deep learning training. The model can quickly calculate the wellbore trajectory control parameters based on the desired wellbore trajectory.

[0044] In one embodiment, the drilling fluid performance parameters and the basic formation parameters to be drilled are trained through deep learning to obtain a drilling fluid formation parameter coupling model. The model can dynamically adjust the drilling fluid performance parameters according to the formation parameters obtained in real time, so that the drilling fluid performance is adapted to the formation conditions in real time.

[0045] In one embodiment, a multi-index fusion evaluation model is constructed by learning from all historical data. Downhole sensor data, such as wellbore trajectory parameters, drill tool mechanical parameters, and drilling fluid performance parameters, is acquired in real time and input into the evaluation model. This model can promptly assess drilling conditions and provide decision-making support for optimizing drilling parameters. In this embodiment, the constructed multi-index fusion evaluation model is also used to establish a fault mode library based on historical data and experience.

[0046] In one embodiment, optimal control strategies and control instructions, including drilling pressure, rotational speed, and steering tool parameters, calculated based on the desired wellbore trajectory, are obtained from a wellbore trajectory control model. Drilling fluid performance parameters, such as density, viscosity, and shear force, dynamically adjusted based on real-time formation parameters, are obtained from a drilling fluid-formation parameter coupling model. Data analysis and deep learning are then performed to establish a correlation model between the wellbore trajectory control parameters and the drilling fluid performance parameters. This model is used to quantify the degree of mutual influence between the wellbore trajectory control parameters and the drilling fluid performance parameters. The model determines a comprehensive objective function based on the goal of real-time dynamic tracking and control of the drilling trajectory during drilling operations. This objective function takes a weighted sum of multiple indicators, such as mechanical penetration rate, drilling fluid parameters, and wellbore trajectory deviation. An optimization algorithm is used to comprehensively optimize the wellbore trajectory control parameters and drilling fluid performance parameters, while satisfying various constraints (such as the drilling tool load capacity and formation pressure limits), to determine an optimal set of drilling parameter combinations. Based on the optimized drilling parameter combination, a detailed drilling parameter setting plan is formulated, including specific parameter values such as bit pressure, rotation speed, steering tool settings, drilling fluid density, viscosity, shear force, as well as the parameter adjustment range and adjustment timing.

[0047] The system generates a drilling parameter setting plan for drilling through a drilling parameter generation module. First, by running the wellbore trajectory control model obtained above, the wellbore trajectory control parameters are quickly calculated according to the expected wellbore trajectory; by running the drilling fluid formation parameter coupling model obtained above, the drilling fluid performance parameters are dynamically adjusted according to the real-time acquired formation parameters; further, by running the association model, a drilling parameter setting plan is obtained according to the wellbore trajectory control parameters and the drilling fluid performance parameters.

[0048] The system runs a multi-index fusion evaluation model through the drilling data verification module, comparing real-time drilling data with the data in the drilling parameter setting plan to determine whether any deviations occur. In this embodiment, real-time wellbore trajectory data, such as inclination, azimuth, and tool face angle, are compared with the set wellbore trajectory to calculate deviations. The smaller the deviation, the closer the current wellbore trajectory is to the planned trajectory. The three-dimensional coordinates of the drill bit at the bottom of the well are measured (typically using a measurement while drilling instrument) and compared with the coordinates of the target point. The distance between the two is calculated. The closer the distance, the higher the target accuracy. In this embodiment, the drilling tool's operating efficiency is assessed by analyzing its mechanical parameters. For example, excessive torque or abnormal power consumption during drilling indicates drilling problems, which can be used to diagnose drilling problems and optimize wellbore trajectory control. In this embodiment, fluctuations in drilling fluid properties, such as density, viscosity, and shear force, are evaluated during drilling to assess the degree of compatibility between drilling fluid properties and real-time formation parameters. Drilling fluid with good adaptability can effectively reduce the adverse impact of the formation on the wellbore trajectory. For example, in formations prone to collapse, appropriate drilling fluid density and viscosity can prevent well wall collapse and ensure the smooth extension of the wellbore trajectory.

[0049] When drilling data deviates, the system uses the drilling parameter optimization module to match the current deviation characteristics with various fault modes in the fault mode library. By comparing characteristics such as the type, magnitude, and trend of the deviation, the system identifies the most similar fault mode. Furthermore, based on the fault mode, the system selects an appropriate solution from a pre-defined solution library. This solution is then integrated into the drilling parameter setting plan, allowing drilling parameters to be adjusted and issues corrected in real time.

[0050] In this embodiment, after identifying the most similar failure modes, preliminary decision recommendations are generated based on pre-defined rules and expert knowledge. These recommendations are sent to the drilling equipment control system and ground operators to automatically adjust the drilling parameter settings. After the parameters are adjusted, changes in relevant indicators are continuously evaluated based on real-time feedback data to assess the effectiveness of the decision. If deviations persist or the expected improvement is not achieved, further problem diagnosis and analysis are performed to readjust the decision.

[0051] In an embodiment of the present application, the drilling data verification module includes: a real-time monitoring unit and a comparison unit.

[0052] In this embodiment, the drilling data verification module acquires drilling data in real time through a real-time monitoring unit. For example, it acquires wellbore trajectory data such as inclination, azimuth, and well depth through a measurement while drilling (MWD) tool; acquires formation parameters such as lithology, porosity, permeability, and formation pressure through geosteering instruments and logging equipment; and measures drilling fluid density, viscosity, shear force, and filtration loss in real time using drilling fluid performance monitoring instruments. The acquired drilling data is compared with the drilling parameter setting plan through a comparison unit. A deviation in the drilling data is determined when the wellbore trajectory data exceeds the parameters in the drilling parameter setting plan, for example, when the wellbore trajectory data in the drilling data deviates from the preset wellbore trajectory data; when the drill tool mechanical data in the drilling data exceeds the normal operating range; or when the drilling fluid performance data exceeds the standard drilling hydraulic performance data.

[0053] In this embodiment, the drilling data verification module also includes an updating unit for cleaning the drilling data acquired by the real-time monitoring unit, such as removing noise, outliers, and missing values, normalizing data of different types and ranges to make them have the same scale, and using an incremental learning method to adapt the wellbore trajectory control model and the drilling fluid formation parameter coupling model to new data features by adjusting the model parameters based on the original knowledge.

[0054] In an embodiment of the present application, the drilling parameter optimization module includes: a fault mode matching unit and a parameter optimization unit.

[0055] In this embodiment, the drilling parameter optimization module uses a fault pattern matching unit to match deviation characteristics with fault patterns in a preset fault pattern library, identify the most similar fault pattern, and retrieve a corresponding solution from a preset solution library based on the fault pattern. The parameter optimization unit integrates the solution into the drilling parameter setting plan for correction. In this embodiment, after the fault matching unit determines the most similar fault pattern, the expert rule library generates preliminary decision recommendations based on pre-set rules and expert knowledge. These recommendations are sent to the drilling equipment control system and ground operators to automatically adjust the drilling parameter setting plan.

[0056] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed, it implements the above-mentioned deep learning-based real-time dynamic tracking and control method for drilling trajectories.

[0057] In summary, this application utilizes a deep learning model to train complex and changing geological conditions, a large amount of multi-source data from various sensors and geology, and real-time feedback of logging data enables the model to have autonomous learning capabilities. This allows the model to be continuously updated and optimized as data accumulates, continuously improving drilling trajectory tracking and control accuracy without the need for frequent manual reprogramming or algorithm adjustment. Simultaneously, a multi-index fusion evaluation model is trained to provide reliable guidance for drilling operations. This model can simultaneously consider multiple data states, such as wellbore trajectory, drilling equipment status, and drilling fluid properties, and conduct comprehensive analysis and decision-making to develop a more scientific and reasonable trajectory tracking plan, thereby improving overall efficiency.

[0058] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0059] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A real-time dynamic tracking control method for drilling trajectory based on deep learning, characterized in that: The method comprises: S1. Generate wellbore trajectory control parameters based on real-time acquired well basic information and preset wellbore trajectory data using a pre-built wellbore trajectory control model; and set optimal drilling fluid performance parameters corresponding to the current operating conditions based on the formation environment in which the drilling tool is located using a pre-built drilling fluid formation parameter coupling model; S2. Deeply learning the wellbore trajectory control parameters and the optimal drilling fluid performance parameters using a pre-built correlation model to generate a drilling parameter setting plan, and controlling the downhole drilling tool to drill according to the drilling parameter setting plan; S3. Comparing the real-time acquired drilling data with the data in the drilling parameter setting plan using a pre-built multi-index fusion evaluation model to determine whether the drilling data has any deviation; S4. Determine a failure mode corresponding to the deviation from a pre-built failure mode library, and determine a corresponding solution from a solution library based on the failure mode to modify the drilling parameter setting solution.

2. The real-time dynamic tracking control method for drilling trajectory based on deep learning according to claim 1, characterized in that: Before step S1, the process includes: Constructing a deep learning database based on historical data of the area to be drilled, wherein the types of historical data include one or more of the following: geographic location of the well, well depth, drilling fluid performance parameters, wellbore trajectory data, drilling tool type, drilling tool mechanical parameters, basic formation parameters of the well to be drilled, fault type, fault characteristics, and solutions; The data in the deep learning database is subjected to data refinement processing, including data cleaning, data deduplication, data standardization, data anomaly processing and noise reduction.

3. The real-time dynamic tracking control method for drilling trajectory based on deep learning according to claim 2, characterized in that: The method further comprises: performing deep learning on the wellbore trajectory data and the drilling tool mechanical parameters to construct the wellbore trajectory control model; Performing deep learning on the basic formation parameters of the well to be drilled and the drilling fluid performance parameters to construct a drilling fluid formation parameter coupling model; performing deep learning on the wellbore trajectory data and the drilling fluid performance parameters to construct the correlation model; Deep learning is performed on the historical data to construct the multi-index fusion evaluation model.

4. The real-time dynamic tracking control method for drilling trajectory based on deep learning according to claim 1, characterized in that: Determining that the drilling data has a deviation in step S3 includes: The wellbore trajectory data acquired in real time exceeds the wellbore trajectory data in the drilling parameter setting plan; and / or the drill tool mechanical data acquired in real time exceeds the drill tool mechanical data in the drilling parameter setting plan; and / or the drilling fluid performance data acquired in real time exceeds the drilling fluid performance data in the drilling parameter setting plan.

5. The real-time dynamic tracking control method for drilling trajectory based on deep learning according to claim 1, characterized in that: Before step S4, the method further includes: Constructing a fault mode library based on historical data and empirical data, wherein the fault mode library includes: historical fault types and deviation features corresponding to the historical fault types; Build a solution library based on pre-set rule base and optimization algorithm; Determining the failure mode corresponding to the deviation in step S4 includes: Matching the characteristics of the deviation with the failure modes in the failure mode library to identify the most similar failure mode; A corresponding solution is obtained from the solution library based on the failure mode.

6. The real-time dynamic tracking control method for drilling trajectory based on deep learning according to claim 1, characterized in that: The method further comprises: The real-time acquired drilling data is used as a new sample to update the parameters of the wellbore trajectory control model and the drilling fluid formation parameter coupling model.

7. A real-time dynamic tracking and control system for drilling trajectories based on deep learning, characterized in that: The system comprises: A database for storing historical data of the area to be drilled; a model training module for performing deep learning on the data in the database to construct a wellbore trajectory control model, a drilling fluid formation parameter coupling model, and a multi-index fusion evaluation model, and for performing deep learning on the data parameters output by the wellbore trajectory control model and the drilling fluid formation parameter coupling model to construct an association model; A drilling parameter generation module is used to generate a drilling parameter setting plan based on the basic information of the well, the preset well trajectory data and the formation environment where the drilling tool is located through the well trajectory control model, the drilling fluid formation parameter coupling model and the association model; A drilling data verification module is used to compare the drilling data acquired in real time with the data in the drilling parameter setting plan to determine whether there is a deviation; The drilling parameter optimization module is used to obtain a corresponding solution to optimize the drilling parameter setting solution when deviation occurs in the drilling data.

8. The real-time dynamic tracking and control system for drilling trajectory based on deep learning according to claim 7, characterized in that: The drilling data verification module includes: Real-time monitoring unit, used to obtain drilling data in real time; The comparison unit is configured to determine that a deviation occurs in the drilling data when the drilling data exceeds a parameter in the drilling parameter setting plan.

9. The real-time dynamic tracking and control system for drilling trajectory based on deep learning according to claim 7, characterized in that: The drilling parameter optimization module includes: A fault pattern matching unit is used to match the deviation characteristics with the fault patterns in a preset fault pattern library, identify the most similar fault pattern, and obtain a corresponding solution from a preset solution library based on the fault pattern; A parameter optimization unit is used to integrate the solution into the drilling parameter setting plan for correction.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed, implements the real-time dynamic tracking and control method for drilling trajectories based on deep learning as described in any one of claims 1 to 6.

Images