Intelligent drilling guiding method and system for petroleum drilling tool face

Abstract

The invention provides an intelligent drilling guide method and system for a petroleum drilling tool face, and relates to the technical field of intelligent control, and the method comprises the steps: determining a first target point, a second target point and a third target point in the process of tracking a target azimuth angle according to a dynamic adjustment instruction; constructing a dynamic triangle through the first target point, the second target point and the third target point, and taking the barycentric coordinate of the dynamic triangle as an azimuth angle correction value; correcting forward and reverse rotation angles and periodic parameters of the drill string according to the azimuth angle correction value so as to obtain a corrected dynamic adjustment instruction; and controlling a drill string to act based on the corrected dynamic adjustment instruction, so that the underground tool surface continuously approaches the target azimuth angle through closed-loop iteration adjustment. The control efficiency and stability of the tool face under the complex working condition can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an intelligent drilling guide method and system for a petroleum drilling tool face. Background Art

[0002] In the field of oil drilling, especially in directional and horizontal well drilling operations, precise control of the tool face azimuth is a key technology to ensure that the wellbore trajectory meets the design requirements. The tool face azimuth determines the directional direction of the downhole drilling tool, and its control accuracy directly affects the drilling efficiency, trajectory deviation and operation safety.

[0003] In traditional drilling operations, tool face adjustment mainly relies on manual operation. The driller manually controls the angle and cycle of the top drive oscillation based on experience. Therefore, there are the following problems: On the one hand, manual adjustment has operational delays and precision errors, and it is difficult to respond to complex downhole conditions (such as changes in formation lithology and fluctuations in friction torque) in real time, resulting in frequent deviations of the wellbore trajectory from the target, increasing the cost of subsequent corrections and the risk of stuck drill. On the other hand, in conventional sliding drilling technology, the support pressure problem caused by the large friction torque of the drill string is particularly prominent, and the drilling pressure is difficult to be effectively transmitted to the drill bit, resulting in a significant decrease in drilling speed. Although the existing top drive torsion swing technology can reduce friction by periodically rotating the drill string, the setting of torsion swing parameters (such as forward and reverse angles and periods) depends on manual experience. When the parameters are unreasonable, "over-adjustment" or "insufficient drag reduction" is prone to occur, which not only affects the directional accuracy but also aggravates the fluctuation of mechanical load. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent drilling guide method and system for a petroleum drilling tool face, which can improve the control efficiency and stability of the tool face under complex working conditions.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows: In a first aspect, an intelligent drilling method for a petroleum drilling tool face is provided, the method comprising: Step S1: acquiring the azimuth data of the downhole tool face in real time, and synchronously recording the historical azimuth offset; Step S2: Processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; Step S3: According to the dynamic adjustment instruction generated in step S2, in the process of tracking the target azimuth, determine the first target point, the second target point and the third target point; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; Step S4: correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; Step S5: Based on the corrected dynamic adjustment instruction, the drill string movement is controlled so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

[0006] In a second aspect, an intelligent drilling guide system for a petroleum drilling tool face includes: An acquisition module is used to acquire the azimuth data of the downhole tool face in real time and simultaneously record the historical azimuth offset; A generating module, used for processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; A correction module is used to determine the first target point, the second target point and the third target point in the process of tracking the target azimuth according to the generated dynamic adjustment instruction; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; A processing module, used for correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; The control module is used to control the drill string action based on the corrected dynamic adjustment instruction, so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

[0007] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.

[0008] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.

[0009] The above solution of the present invention includes at least the following beneficial effects: The downhole sensor array collects and pre-processes azimuth data in real time to ensure that the data source of the input algorithm is real-time, continuous and reliable. The historical offset is recorded simultaneously to provide a data basis for subsequent dynamic compensation, avoiding adjustment lags or deviation accumulation caused by relying solely on real-time data, and improving the system's ability to analyze tool face change trends.

[0010] A hybrid neural network (LSTM + CNN) is used to extract the spatiotemporal characteristics of azimuth data, and the dynamic response of the drill string is predicted in combination with drilling parameters, breaking through the limitations of traditional manual experience in setting parameters. The dynamic adjustment instructions generated by multi-objective optimization (accuracy, load, energy consumption) can adapt to complex downhole conditions (such as formation hardness and mud resistance) and achieve a balance between control accuracy and equipment safety.

[0011] The target orientation is dynamically corrected through the "offset compensation point". The proportional compensation mechanism of the historical offset is used to offset in advance the lag or overshoot caused by friction, inertia and other factors during the tool face adjustment process. The correction value based on the center of gravity of the triangle integrates the real-time position, target position and historical deviations, making the adjustment direction closer to the actual dynamic needs, significantly improving tracking accuracy and reducing oscillation.

[0012] The forward and reverse angles and cycles of the drill string are dynamically adjusted according to the corrected azimuth, and the response speed is dynamically optimized in combination with the historical offset frequency and the current working conditions (such as shortening the cycle to cope with rapid changes and extending the cycle to reduce the load). Through multi-objective constraint verification (accuracy threshold, load safety range, energy consumption limit), equipment overload or invalid adjustment caused by unreasonable parameters is avoided to ensure long-term stable operation of the system.

[0013] Through the closed-loop feedback mechanism of "data acquisition-algorithm processing-parameter correction-execution control", the tool face azimuth is fully automatically tracked, and formation changes can be responded to in real time without human intervention; the iterative adjustment process continuously reduces the deviation to ensure that the wellbore trajectory meets the design requirements with high precision, while reducing safety risks such as drill sticking and well wall collapse caused by manual operation delays or lack of experience, thereby improving drilling efficiency and operation safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of an intelligent drilling guide method for a petroleum drilling tool face provided by an embodiment of the present invention.

[0015] Figure 2 The figure is a schematic diagram of an intelligent drilling guide system for a petroleum drilling tool face provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in a form and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0017] like Figure 1As shown, an embodiment of the present invention provides an intelligent drilling method for a petroleum drilling tool face, the method comprising the following steps: Step S1: acquiring the azimuth data of the downhole tool face in real time, and synchronously recording the historical azimuth offset; Step S2: Processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; Step S3: According to the dynamic adjustment instruction generated in step S2, in the process of tracking the target azimuth, determine the first target point, the second target point and the third target point; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; Step S4: correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; Step S5: Based on the corrected dynamic adjustment instruction, the drill string movement is controlled so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

[0018] In the embodiment of the present invention, the tool face azimuth data is collected and preprocessed in real time through the downhole sensor array to ensure that the data source of the input algorithm has high precision and low latency characteristics; the historical azimuth offset is recorded synchronously, and a time series database of tool face changes is established to provide a historical deviation basis for the calculation of the offset compensation point in step S3, so as to avoid blind adjustment caused by relying solely on real-time data and improve the system's ability to learn the fluctuation law of the tool face. Based on the hybrid architecture of long short-term memory network (LSTM) and convolutional neural network (CNN), the time series characteristics (such as periodic fluctuations) and spatial correlation characteristics (such as the impact of formation changes on the tool face) of azimuth data are deeply mined, breaking through the limitations of traditional manual experience setting parameters and realizing the adaptive generation of adjustment instructions. Combined with multi-objective optimization such as azimuth tracking accuracy, drill string mechanical load balance, energy consumption threshold, etc., the generated dynamic adjustment instructions (forward and reverse angles, periodic parameters) can balance control accuracy and equipment reliability in complex downhole working conditions (such as high friction formations, mud resistance changes), avoiding load overload or energy waste caused by single parameter optimization. By introducing an "offset compensation point" based on historical offsets, proportional compensation is performed on the reverse extension line between the real-time position point and the theoretical target point, and the adjustment lag or overshoot caused by factors such as drill string inertia and downhole friction is corrected in advance, so that the tool face adjustment direction is closer to the actual dynamic needs. The real-time position (first target point), theoretical target (second target point), and historical deviation compensation (third target point) are used to construct a dynamic triangle, with the center of gravity coordinates as the correction value, integrating the current state, target orientation, and historical experience, effectively suppressing the oscillation phenomenon in the tool face tracking process, and improving the stability and accuracy of trajectory control.

[0019] The forward and reverse angles and cycle parameters of the drill string are dynamically adjusted according to the azimuth correction value. For example, the response cycle is adaptively adjusted through the historical offset fluctuation frequency (the cycle is shortened when it changes rapidly, and the cycle is extended when it is stable), so as to avoid efficiency loss or control delay caused by "one-size-fits-all" adjustment; the corrected parameters are verified through a multi-objective optimization model to ensure that the azimuth tracking accuracy meets the standard, the drill string load is within the safe range, and the energy consumption does not exceed the rated power of the equipment. Mechanical failures (such as torque overload) or invalid adjustments caused by unreasonable parameters are avoided at the algorithm level, thereby improving the long-term stability of the system; through the closed-loop feedback mechanism of "data collection-intelligent processing-execution correction", unmanned automatic tracking of the tool face azimuth is realized, and formation changes (such as friction changes caused by sudden changes in lithology) are responded to in real time, solving the delay problem of traditional manual adjustment and greatly shortening the tool face correction time. The iterative adjustment process continuously reduces the deviation between the actual tool face and the target angle, ensuring that the wellbore trajectory strictly meets the design requirements; at the same time, it avoids operational errors caused by insufficient manual experience (such as excessive twisting and swinging causing drill bit sticking), reduces safety risks such as wellbore collapse and equipment damage, and significantly improves drilling efficiency and operational safety under complex working conditions.

[0020] In a preferred embodiment of the present invention, step S1: real-time acquisition of azimuth data of the downhole tool surface and synchronous recording of historical azimuth offsets, including: The tool face azimuth data is collected in real time through a downhole sensor array, and the azimuth data is preprocessed to generate a preprocessed azimuth data stream; The historical azimuth data is stored in a preset time window length, and the historical azimuth offset is calculated.

[0021] In an embodiment of the present invention, a three-axis gyroscope, an accelerometer and a magnetic compass are integrated in a downhole measurement while drilling (MWD) tool to construct a multi-dimensional sensor array, and the original data of the tool face azimuth (unit: degree, accuracy ±0.1°) is collected in real time at a frequency of not less than 10 Hz; a sliding average filtering algorithm (window length 50 ms) is used to remove high-frequency electromagnetic interference and mechanical vibration noise, and retain the true azimuth change trend; through threshold detection (such as 3 consecutive sampling points with a deviation of more than 5° are judged as abnormal), the abnormal value is repaired by linear interpolation of the data before and after; the original attitude data (Euler angle) output by the sensor is converted into the tool face azimuth in the geographic coordinate system, and the coordinate system deviation caused by the rotation of the drill string is eliminated; the pre-processed azimuth data is sorted by timestamp to generate a continuous azimuth data stream (format: timestamp-azimuth value) as the input data source of the subsequent algorithm.

[0022] In the embodiment of the present invention, a circular buffer is established in the control system memory to store historical azimuth data with a preset time window length (such as 300 seconds). When new data is written, the earliest outdated data is automatically overwritten to ensure that the historical records of the last 5 minutes are always retained. The difference between the current azimuth and the preset target angle (△θ=current value-target value) is calculated, and the absolute value |△θ| is taken as the real-time offset amplitude; For all data in the time window, the azimuth change (△θ t =θ t -θ t-1 ), forming a historical offset sequence {△θ 1 , △θ 2 ,…,△θ n}, used to analyze the frequency and amplitude characteristics of tool face fluctuations; calculate the mean (reflecting the long-term trend), standard deviation (reflecting the fluctuation amplitude) and maximum offset value (reflecting the extreme deviation) of the historical offset in real time as the basis for calculating the offset compensation point in step S3.

[0023] In the embodiment of the present invention, through multi-sensor fusion and multi-level preprocessing (filtering, wild value repair, coordinate system conversion), the influence of complex downhole environment (high temperature, high pressure, strong electromagnetic interference) on data acquisition is effectively suppressed, ensuring that the azimuth data error of the input algorithm is ≤0.5°, providing a reliable data source for precise control. Real-time generation of continuous data streams and use of circular buffers to manage historical data avoids information faults caused by traditional discrete sampling, enabling the system to capture high-frequency fluctuations of the tool face (such as rapid angle changes caused by drill string vortex).

[0024] In a preferred embodiment of the present invention, step S2: real-time processing of the azimuth data is performed, and a dynamic adjustment instruction is generated in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes the forward and reverse angles and cycle parameters of the drill string, including: The real-time azimuth data stream, the historical azimuth offset and the preset target angle outputted in step S1 are inputted into a pre-trained artificial intelligence model, wherein the artificial intelligence model is a hybrid architecture based on a long short-term memory network and a convolutional neural network, and is used to learn the mapping relationship between the dynamic change of the azimuth and the control parameters of the drill string; The artificial intelligence model is used to extract the spatiotemporal characteristics of the azimuth data, and the real-time dynamic response of the drill string is predicted in combination with the current drilling parameters to obtain a prediction result; Generate an initial dynamic adjustment instruction according to the prediction result, wherein the initial dynamic adjustment instruction includes a start angle threshold for forward and reverse rotation of the drill string, a rotation angle increment, and a rotation cycle parameter, wherein the start angle threshold for forward and reverse rotation of the drill string is dynamically corrected by the amplitude and direction of the historical offset; The initial dynamic adjustment instruction and the target angle constraint conditions are subjected to multi-objective optimization, wherein the optimization conditions include the azimuth tracking accuracy, the mechanical load balance of the drill string and the energy consumption threshold of the adjustment action, and finally a feasible solution that meets the constraints is generated as the dynamic adjustment instruction.

[0025] In an embodiment of the present invention, the real-time azimuth data stream, the historical azimuth offset and the preset target angle outputted in step S1 are inputted into a pre-trained artificial intelligence model, which is a hybrid architecture based on a long short-term memory network (LSTM) and a convolutional neural network (CNN). The construction and training process of this model is as follows: A large amount of historical drilling data covering azimuth data, drill string control parameters and drilling condition information is collected, cleaned and normalized, and divided into training set, validation set and test set.

[0026] Model construction: CNN layer: Convolution operations are performed on the real-time azimuth data stream using convolution kernels of different sizes (e.g., 3, 5, and 7). Each convolution layer contains 32 convolution kernels. After the convolution, a maximum pooling layer (window size 2, stride 2) and ReLU activation function are added to extract the local spatial features of the azimuth data.

[0027] LSTM layer: The feature sequence output by the CNN layer is input into the LSTM layer containing 64 LSTM units to learn the dynamic change pattern of the azimuth data over time.

[0028] Fully connected layer: The output of the LSTM layer is concatenated with the current drilling parameters and then input into the fully connected layer (containing 128 neurons). A Dropout layer is added (Dropout rate 0.2). Finally, the real-time dynamic response prediction results of the drill string are output through the output layer.

[0029] Using the mean square error loss function and Adam optimizer (initial learning rate 0.001), the training set data is input into the model in batches (64 samples per batch) for training. The learning rate is adjusted according to the performance of the validation set, and the early stopping strategy is used to prevent overfitting. The model performance is evaluated using the test set, and indicators such as the root mean square error and mean absolute error are calculated. If the performance is not ideal, the model is further optimized by adjusting the hyperparameters. The model obtained through the above construction and training process can learn the mapping relationship between the dynamic change of the azimuth angle and the control parameters of the drill string.

[0030] Combined with the current drilling parameters, such as drilling pressure, rotation speed, mud displacement, etc., as additional input information, the extracted spatiotemporal features are fused with the drilling parameters through the fully connected layer to predict the real-time dynamic response of the drill string under different control parameters, such as torque change, displacement, etc., and obtain the prediction results.

[0031] According to the prediction results, the initial values ​​of the start angle threshold, rotation angle increment and rotation cycle parameters for the forward and reverse rotation of the drill string are determined. For example, when the azimuth deviation is predicted to be large, the rotation angle increment and the start angle threshold are appropriately increased, and the start angle threshold is dynamically corrected according to the amplitude and direction of the historical offset. If the amplitude of the historical offset is large and the direction is stable, the start angle threshold is appropriately increased to reduce unnecessary adjustments; if the historical offset fluctuates greatly, the start angle threshold is reduced to improve the response speed of the system.

[0032] By extracting the spatiotemporal characteristics of azimuth data through deep learning models, the dynamic change law of azimuth can be captured more accurately, thereby generating more accurate dynamic adjustment instructions, improving the control accuracy of tool face azimuth, and predicting the real-time dynamic response of the drill string in combination with the current drilling parameters, so that the system can adaptively adjust the control parameters of the drill string according to different drilling conditions, enhancing the adaptability and flexibility of the system. The mechanical load balance of the drill string is considered in the multi-objective optimization process, avoiding excessive torque and stress on the drill string during the adjustment process, reducing the wear and damage of the equipment, and extending the service life of the equipment. By dynamically correcting the start angle threshold, unnecessary adjustment actions are reduced, the frequent start and stop times of the equipment are reduced, and the equipment is further protected. The energy consumption threshold of the adjustment action is included in the multi-objective optimization, and the adjustment instructions are optimized, so that the drill string can reduce energy consumption as much as possible while reaching the target azimuth, reducing the drilling cost; the dynamic adjustment instructions are automatically generated based on the artificial intelligence algorithm, reducing the reliance on manual experience, reducing the impact of human factors on the drilling process, and improving the automation level and reliability of the drilling operation.

[0033] In a preferred embodiment of the present invention, the initial dynamic adjustment instruction and the target angle constraint condition are subjected to multi-objective optimization, wherein the optimization condition includes the azimuth tracking accuracy, the mechanical load balance of the drill string and the energy consumption threshold of the adjustment action, and finally a feasible solution satisfying the constraint is generated as the dynamic adjustment instruction, including: Based on the forward and reverse angle thresholds, rotation angle increments and cycle parameters in the initial dynamic adjustment instructions, a multi-objective evaluation function including azimuth tracking accuracy error, drill string axial torque fluctuation value and single adjustment energy consumption is constructed, where the weight coefficient of each objective is dynamically allocated according to the real-time working conditions downhole. The multi-objective evaluation function is iteratively solved by using a constrained particle swarm optimization algorithm, wherein the constraints include: the azimuth tracking accuracy error does not exceed a preset threshold, the drill string mechanical load fluctuation value is within a safe range, and the energy consumption of a single adjustment is lower than the rated power of the equipment; In the iterative process, the candidate solution set with the smallest azimuth tracking accuracy error is retained first, and the final instruction parameter combination of the mechanical load fluctuation value and the energy consumption comprehensive score is screened in the candidate solution set to generate the preliminary optimization instruction; Performing downhole dynamic response simulation verification on the preliminary optimization instruction, simulating the trajectory of tool face azimuth change after executing the preliminary optimization instruction through a preset drill string dynamics model, and determining it as a feasible solution if the azimuth in the simulation result converges to the target angle and meets the load and energy consumption constraints; If the simulation verification fails, the weight coefficient of the multi-objective evaluation function is adjusted according to the error direction, and new optimization instructions are re-iterated until a dynamic adjustment instruction that meets all constraints is output to obtain the final optimized dynamic adjustment instruction.

[0034] In an embodiment of the present invention, a group of particles are randomly generated in a feasible solution space according to the variable range of the problem, and the position vector of each particle is composed of the forward and reverse angle threshold, the rotation angle increment and the period parameter in the initial dynamic adjustment instruction. For example, assuming that the forward and reverse angle threshold range is [-180°, 180°], the rotation angle increment range is [-10°, 10°], and the period parameter range is [0.5, 5] seconds, then the position vector of each particle is composed of values ​​randomly generated within these ranges. By randomly generating particles, it is possible to conduct an extensive search in the entire feasible solution space, provide a variety of initial solutions for the subsequent optimization process, and increase the possibility of finding the global optimal solution.

[0035] In an embodiment of the present invention, according to a multi-objective evaluation function, the position vector of each particle is substituted into the function for calculation. The multi-objective evaluation function includes objectives such as azimuth tracking accuracy error, drill string axial torque fluctuation value, and single adjustment energy consumption. For example, for a particle, the corresponding azimuth tracking accuracy error (the difference between the current azimuth and the target azimuth) is first calculated according to its position vector, and then the drill string axial torque fluctuation value is calculated according to the drill string dynamics model, and the single adjustment energy consumption is calculated according to the adjustment action parameters. Then, according to the weight coefficient of each objective (dynamically allocated according to the real-time working conditions downhole), these objective values ​​are weighted and summed to obtain the fitness value of the particle; the fitness value is used as an indicator to measure the quality of the particle, and can comprehensively reflect the performance of each particle in the multi-objective optimization. By calculating the fitness value, the particles can be compared and screened.

[0036] In an embodiment of the present invention, the construction process of the drill string dynamics model is as follows: Collect basic drill string parameters: drill string length, diameter, material elastic modulus, density, joint friction coefficient, etc.; Real-time working condition data: current drilling pressure, rotation speed, mud density, formation friction coefficient (obtained through historical data or measurement while drilling), current azimuth angle of tool face and target angle; Environmental parameters: downhole temperature, pressure (affecting material mechanical properties), mud viscosity (affecting fluid resistance).

[0037] Based on the concentrated mass method, the drill string is discretized into multiple units, and each unit considers the balance of axial force, torque, and bending moment; when the drill string contacts the well wall, the tangential friction force and normal positive pressure are calculated according to the friction coefficient and contact angle to simulate the constraint of the formation on the drill string; inertial force and damping force are added to consider the centrifugal force and Coriolis force during the rotation of the drill string, as well as the viscous damping effect caused by the mud flow.

[0038] After inputting the adjustment command parameters (forward and reverse rotation angles, cycles), the model calculates the torque change curve when the drill string rotates: based on the rotation angle increment and cycle, combined with the formation friction, the torque required for each step of rotation and the torque fluctuation range are predicted. By integrating the drill string rotation angle and combining the inertial navigation principle (gyroscope, accelerometer data), the change trend of the tool face angle over time under different adjustment parameters is calculated; Key output indicators: torque fluctuation peak (reflecting mechanical load), azimuth convergence speed (reflecting tracking accuracy), energy consumption (calculated based on motor power and action time).

[0039] The model is trained using historical drilling data (such as actual torque and azimuth changes under known adjustment parameters) to correct unknown parameters such as friction coefficient and damping coefficient. Error thresholds are set (such as torque simulation error ≤5%, azimuth trajectory error ≤1°) to ensure that the model prediction accuracy meets engineering requirements.

[0040] When it comes to specific applications, it can be achieved through the following cases: Drill string basic parameters and discretization, basic parameter collection: Drill string length , outer diameter , inner diameter , Material density , elastic modulus , joint friction coefficient μ; Real-time working conditions: drilling pressure P (N), speed n ( , which needs to be converted into angular velocity , ), mud density (kg / m 3 ), formation friction coefficient ; Environmental parameters: downhole temperature T (°C), pressure p (Pa), mud viscosity .

[0041] Lumped mass method discretization: Divide the drill string into Unit, Units: Unit Length: ; Unit Quality: ,in is the cross-sectional area (m 2 ); Moment of Inertia: ,in, Indicates the unit outer diameter (m). If the drill string has the same diameter, then ; Indicates the inner diameter of the unit. If the drill string has the same diameter, then .

[0042] Axial force at top of each unit: , represents the axial force at the top of unit i, initial value (Drilling pressure), It represents the axial distributed load of unit i (N / m), including the weight of the drill string and the buoyancy of the mud.

[0043] Torque Balance: Considering formation friction and internal torque transmission: ,in, Indicates the top torque of unit i (N・m); Indicates the internal distributed torque of the unit (N・m / m), caused by the rotational resistance of the drill string; Represents the friction torque between the unit and the wellbore wall.

[0044] Bending moment balance: Considering lateral loads (such as wellbore contact forces): ,in, represents the bending moment at the top of element i (N·m); represents the lateral contact force of element i.

[0045] Normal contact force: When the drill string contacts the wellbore wall: ,in, represents normal positive pressure (N); k represents the formation stiffness coefficient (N / m), which is determined by the formation lithology; It represents the radial displacement of unit i (m), that is, the compression of the gap between the drill string and the well wall.

[0046] Tangential friction ,in, represents the formation friction coefficient (dimensionless); represents the unit tangential velocity (m / s), is a sign function, which determines that the direction of friction is opposite to the direction of motion.

[0047] Centrifugal force (rotational inertia force) ,in, Indicates the distance from the unit center of mass to the drill string rotation axis (m), in the bending state ; Represents the angular velocity of rotation.

[0048] Coriolis force (radial motion inertia force) ,in, Indicates the radial movement speed of the unit (m / s).

[0049] Mud viscous damping force ,in, represents the damping coefficient (N・s / m), which is positively correlated with the mud viscosity and unit surface area; Indicates the movement speed of the unit relative to the mud (m / s).

[0050] Adjust command response calculation: According to the rotation angle increment and cycle : ,in, Indicates the basic torque (N・m), which is determined by the drilling pressure and friction; Indicates the torque fluctuation amplitude (N・m), which is related to the formation heterogeneity; Represents the phase angle, initialized to 0.

[0051] By integrating the angular velocity and combining it with the inertial navigation data: ;in, represents the initial azimuth; It represents the well inclination angle, that is, the angle between the wellbore axis and the plumb line, which is the effective component affecting the change of azimuth. is the tool face azimuth at time t, which indicates the angle of the tool face relative to the initial reference direction when the drill string rotates; ω(t) is the drill string rotation angular velocity, which indicates the rotation speed of the drill string around its own axis (changing with time t).

[0052] Energy consumption per adjustment ,in, represents instantaneous power (W), which is calculated by multiplying torque and angular velocity; where, Indicates the energy consumption (joules) of a single adjustment process, reflecting the total energy consumption of the motor or power system when the drill string rotates to adjust the tool face. It is used to evaluate the energy efficiency of drilling operations and optimize adjustment parameters (such as cycle, angle) to reduce energy consumption; Indicates the rotation torque (N m); t0 and t1 represent the start and end time of the adjustment process, corresponding to the cycle in the adjustment instruction .

[0053] Model training and calibration: Parameter correction (taking friction coefficient as an example) Using historical measured torque With simulated torque , that is, the theoretical torque: ; in, Indicates the correction factor (0< <1), avoid over-adjustment; Error threshold setting: Torque simulation error: ; Azimuth track error: .

[0054] in, Represents the corrected formation friction coefficient (dimensionless), which is used to update the friction calculation when the drill string contacts the wellbore / formation in the model; Indicates the formation friction coefficient before correction (initial value or last iteration value). The initial value can be set by the empirical value of the formation lithology (such as 0.3 for sandstone and 0.5 for shale), or initialized based on historical similar working condition data; if =0.1, which means that each correction is 10% of the deviation ratio, gradually approaching the true value. A value that is too small will lead to slow convergence, and a value that is too large may cause parameter oscillation (for example, if it exceeds 1, the friction coefficient may change in the opposite direction); Indicates the measured tool face azimuth; It represents the simulated azimuth, i.e. the theoretical azimuth trajectory; the threshold value of 1° represents the maximum angle deviation allowed in directional drilling.

[0055] According to the fitness value and historical optimal position of the particle, the update formula of the particle swarm optimization algorithm is used to update the position and speed of the particle. Specifically, the speed update formula of each particle will take into account the current speed of the particle, the difference between the particle's own historical optimal position and the current position, and the difference between the global optimal position and the current position. Then, the position of the particle is updated according to the updated speed. For example, if the historical optimal position of a particle has a better fitness value in a certain direction, it will tend to move in this direction when updating the speed, thereby guiding the particle to search for a better solution space. By continuously updating the position and speed of the particle, the particle can gradually move in the solution space and tend to a better solution. This search method based on swarm intelligence can make full use of information sharing and collaboration between particles, accelerate convergence speed, and improve optimization efficiency.

[0056] Check whether the position vector of each particle meets the constraints such as the azimuth tracking accuracy error does not exceed the preset threshold, the drill string mechanical load fluctuation value is in the safe range, and the single adjustment energy consumption is lower than the rated power of the equipment. For example, if the azimuth tracking accuracy error of a particle exceeds the preset threshold, or the drill string mechanical load fluctuation value exceeds the safe range, or the single adjustment energy consumption is higher than the rated power of the equipment, then the particle does not meet the constraints, ensuring that the solutions generated during the optimization process are within the practical feasible range, avoiding infeasible solutions that lead to system failure or performance degradation. By checking the constraints, the reliability and safety of the system can be guaranteed.

[0057] For particles that do not meet the constraints, a certain correction strategy is adopted to make them meet the constraints. For example, if the azimuth tracking accuracy error of a particle is too large, the error can be reduced by adjusting its forward and reverse angle threshold or rotation angle increment; if the mechanical load fluctuation value of the drill string exceeds the safe range, the cycle parameter or other related parameters can be adjusted to reduce the load fluctuation; if the energy consumption of a single adjustment is too high, the parameters of the adjustment action can be optimized to reduce energy consumption. The specific correction method can be designed according to the characteristics of the problem and the actual situation to ensure that all particles in the particle swarm meet the constraints, so that the optimization process can be carried out in the feasible solution space, and the feasibility and practicality of the optimization results are improved.

[0058] Repeat the steps until the preset termination condition is met. The termination condition can be reaching the maximum number of iterations, the fitness value converges to a certain accuracy, or finding the optimal solution that meets specific requirements. For example, when the number of iterations reaches the set maximum value, or when the fitness value of the optimal solution in multiple consecutive iterations changes less than a certain threshold, the optimization process is considered to have converged and the iteration is stopped. Through continuous iteration, the particle swarm can be continuously optimized and gradually approach the global optimal solution or the solution that meets specific requirements. After the termination condition is met, the optimization result obtained can achieve a good balance between multiple objectives and improve the overall performance of the system; after each iteration, a group of candidate solutions with the smallest azimuth tracking accuracy error are selected from all particles. This can be achieved by comparing the azimuth tracking accuracy error value of each particle, and the particles with the smallest error value and their corresponding solutions are retained to form a candidate solution set. Ensure that the important goal of azimuth tracking accuracy is always focused on during the optimization process, and give priority to retaining the solution that performs best on this goal.

[0059] For the retained candidate solution set, each candidate solution is comprehensively scored according to the mechanical load fluctuation value and energy consumption of the drill string. The comprehensive score can be obtained by weighted summation. The corresponding weights are assigned according to the importance of the mechanical load fluctuation value and the energy consumption. The scores of the two are added to obtain the comprehensive score. Then, the candidate solution with the highest comprehensive score is selected as the preliminary optimization instruction. Under the premise of ensuring the azimuth tracking accuracy, the mechanical load fluctuation value and the energy consumption are comprehensively considered, so that the generated preliminary optimization instructions can achieve better performance in multiple objectives, thereby improving the overall efficiency and economy of the system. The embodiment of the present invention can realize multi-objective optimization based on the particle swarm optimization algorithm, improve the system performance in terms of azimuth tracking accuracy, drill string mechanical load balance and adjustment action energy consumption threshold, while ensuring that the solution results meet the constraints and improve the reliability and safety of the system.

[0060] The specific determination process of the above comprehensive score is as follows: Collect the torque fluctuation peaks corresponding to all candidate solutions in the current iteration, determine the maximum and minimum values, and normalize the mechanical load fluctuation value by minimum-maximum normalization to obtain the normalized load value; Count the energy consumption values ​​of all candidate solutions, determine the upper limit (energy consumption corresponding to the rated power of the equipment) and lower limit (theoretical minimum energy consumption), and normalize the energy consumption values ​​through minimum-maximum normalization to obtain the normalized energy consumption values; If the formation friction suddenly increases (judged by mud pressure mutation or torque sensor data), increase the mechanical load weight (such as mechanical load fluctuation value = 0.6, single adjustment energy consumption = 0.4) to avoid overload; if the equipment is low on power or needs to operate for a long time, increase the energy consumption weight (such as mechanical load fluctuation value = 0.4, single adjustment energy consumption = 0.6) to reduce energy consumption; under normal working conditions, use a balanced weight (mechanical load fluctuation value = single adjustment energy consumption = 0.5); for each candidate solution, calculate the normalized load value and the normalized energy consumption value, and obtain a comprehensive score through weighted summation based on the normalized load value and the normalized energy consumption value, and select the candidate solution with the highest score as the preliminary optimization instruction to ensure a balance between load and energy consumption.

[0061] In the embodiment of the present invention, multi-physics field coupling modeling is used to accurately simulate the dynamic response of the drill string under different adjustment parameters, predict the torque fluctuation and azimuth change trend in advance, and avoid overshoot or lag problems in actual execution; calibrate model parameters based on historical data to adapt to complex downhole environments (such as high friction resistance and rock property mutations), improve the engineering reliability of simulation results, and normalize to eliminate the influence of indicators of different dimensions, so that load and energy consumption are comparable; dynamic weight allocation strategy is combined with real-time working conditions to flexibly balance equipment protection and energy efficiency optimization. Through downhole dynamic response simulation verification, the feasibility and effectiveness of preliminary optimization instructions can be evaluated in advance, unforeseen problems can be avoided in actual applications, and the reliability and stability of the system can be improved.

[0062] In a preferred embodiment of the present invention, based on the forward and reverse angle thresholds, rotation angle increments and period parameters in the initial dynamic adjustment instructions, a multi-objective evaluation function including azimuth tracking accuracy error, drill string axial torque fluctuation value and single adjustment energy consumption is constructed, wherein the weight coefficient of each objective is dynamically allocated according to the real-time working conditions downhole, including: Real-time collection of downhole working parameters, including drilling pressure, drill string speed, formation lithology hardness coefficient and mud flow resistance, as the basis for weight allocation; Calculate the initial weight ratios of azimuth tracking accuracy error, torque fluctuation and energy consumption in the multi-objective evaluation function according to the operating condition parameters; Dynamically correcting the initial weight ratio through a pre-trained fuzzy logic rule base to obtain a corrected weight coefficient, wherein the rule base is constructed based on a mapping relationship between operating parameters and weight adjustments in historical drilling data; The corrected weight coefficients are respectively integrated with the azimuth tracking accuracy error, the drill string axial torque fluctuation value and the single adjustment energy consumption to form weighted parameter items corresponding to each objective. A multi-objective evaluation function matching the current downhole working conditions is constructed through the combination of various weighted parameter items.

[0063] In the embodiment of the present invention, sensors are used to collect parameters such as drilling pressure, drill string speed, formation lithology hardness coefficient, and mud flow resistance in real time. For example, drilling pressure is measured by a pressure sensor, drill string speed is obtained by a speed sensor, formation lithology hardness coefficient can be determined based on geological exploration data or relevant information of measurement while drilling, and mud flow resistance can be calculated by mud performance parameters and wellbore conditions.

[0064] Providing an accurate real-time data basis for weight allocation enables the weights to be dynamically adjusted according to actual downhole conditions, thereby more accurately reflecting the importance of each target under current conditions.

[0065] According to the collected working condition parameters, the initial weight ratio of azimuth tracking accuracy error, torque fluctuation and energy consumption is calculated using a pre-set algorithm. For example, the weight of torque fluctuation can be determined based on the formation lithology hardness coefficient and mud flow resistance. When the formation lithology hardness is high or the mud flow resistance is large, the weight of torque fluctuation is appropriately increased; the weight of azimuth tracking accuracy error is determined based on drilling pressure and drill string speed. When the drilling pressure is large or the drill string speed is high, the weight of azimuth tracking accuracy error is appropriately increased; the weight of energy consumption is determined based on the relationship between energy consumption and other parameters, as well as the energy consumption requirements of the operation.

[0066] The weight ratio of each target is preliminarily determined based on the operating condition parameters, and the influence of different operating conditions on each target is taken into consideration, providing a reasonable starting point for subsequent dynamic correction.

[0067] The calculated initial weight ratio is input into the pre-trained fuzzy logic rule base. The rule base dynamically corrects the initial weight ratio based on the mapping relationship between the operating parameters and the weight adjustment in the historical drilling data. For example, if the historical data shows that the weight of the torque fluctuation should be increased by a certain ratio under a certain combination of formation lithology hardness coefficient and mud flow resistance, the rule base will adjust the initial weight of the torque fluctuation accordingly based on the current operating parameters. In this way, the corrected weight coefficient is obtained.

[0068] The dynamic correction of initial weights using historical data and fuzzy logic rule base can more flexibly adapt to different downhole working conditions, further improve the accuracy and rationality of weight allocation, and make the multi-objective evaluation function more in line with actual conditions.

[0069] The corrected weight coefficients are respectively integrated with the azimuth tracking accuracy error, the drill string axial torque fluctuation value and the single adjustment energy consumption. For example, the azimuth tracking accuracy error is multiplied by the corresponding weight coefficient to obtain the weighted parameter item of the azimuth tracking accuracy error; the drill string axial torque fluctuation value is multiplied by its weight coefficient to obtain the weighted parameter item of the torque fluctuation; the single adjustment energy consumption is multiplied by the corresponding weight coefficient to obtain the weighted parameter item of the energy consumption. Then, these weighted parameter items are combined to form a multi-objective evaluation function that matches the current downhole working conditions. For example, these weighted parameter items can be combined in a weighted summation manner to obtain the final multi-objective evaluation function.

[0070] By integrating the corrected weight coefficient with each objective, the constructed multi-objective evaluation function can comprehensively consider the real-time downhole working conditions and accurately evaluate the comprehensive performance of different adjustment instructions in terms of multiple objectives such as azimuth tracking accuracy, drill string axial torque fluctuation value and single adjustment energy consumption.

[0071] In another preferred embodiment of the present invention, when applied specifically, the above content can be implemented in the following manner: Sensor Data Acquisition Model Obtain the working condition parameter vector in real time through downhole sensors: ; in, Indicates drilling pressure (N), which is measured by the pressure sensor and reflects the force exerted by the drill bit on the formation; Indicates the drill string speed (rad / s), obtained by the speed sensor, which affects the torque fluctuation and azimuth change rate; It represents the formation lithology hardness coefficient (dimensionless), which is calibrated by logging while drilling (LWD) data or historical geological exploration data to characterize the formation's anti-drilling strength; Indicates the resistance to mud flow (N), based on mud density , Viscosity And calculation of wellbore geometry parameters: ; Where, L is the well depth, is the mud flow rate, is the ratio of the wellbore radius to the drill string radius.

[0072] The initial weight ratio is calculated based on the weight distribution model of the working condition, and the initial weight vector is calculated using a nonlinear mapping function: in: ;in: ; in: is the Sigmoid function, which maps the input to the (0, 1) interval; , , and is the calibration coefficient (determined by historical data regression); is drilling pressure; It is the reference value of drilling pressure; is the drill string rotation speed; is the reference value of the drill string rotation speed; is the formation hardness coefficient, dimensionless; It is the reference value of the formation hardness coefficient; is the mud flow resistance; are the reference values ​​of mud flow resistance, which are used to normalize the working condition parameters; , and They are the initial weights of azimuth accuracy, torque fluctuation, and energy consumption, respectively, and their sum is 1.

[0073] Fuzzy logic dynamic correction, fuzzy rule base and weight correction model: The following is a textual description of the specific calculation process of the fuzzy logic rule base and weight correction model: Fuzzification process (input variable processing): For each input variable (such as drilling pressure, drill string speed, formation lithology hardness coefficient, mud flow resistance), it is converted into a membership value in the fuzzy set, which means determining the degree of membership of each input variable in different fuzzy subsets (such as low, medium, and high).

[0074] Taking the formation lithology hardness coefficient as an example, three fuzzy subsets are defined: low, medium and high. The membership of the formation lithology hardness coefficient in each subset is calculated by a specific function (such as a Gaussian membership function). This function determines the membership based on the input value and pre-set parameters (such as mean and standard deviation). For example, when the input formation lithology hardness coefficient is close to the mean of a fuzzy subset (such as "high"), its membership in this subset will be higher.

[0075] Fuzzy rule base construction: The fuzzy rule base is composed of a series of rules, each of which has a premise and a corresponding conclusion. The premise is a combined description of the fuzzy subsets of the input variables, while the conclusion is a fuzzy subset description of each output variable (i.e., the weight correction of the azimuth tracking accuracy error, the weight correction of the axial torque fluctuation value of the drill string, and the weight correction of the energy consumption of a single adjustment).

[0076] For example, a rule may be: if the formation lithology hardness coefficient is "high" and the mud flow resistance is "high", then the weight correction of the drill string axial torque fluctuation value is "increase 0.15", the weight correction of the azimuth tracking accuracy error is "decrease 0.05", and the weight correction of the single adjustment energy consumption is "decrease 0.10". The "high", "increase 0.15", etc. here are fuzzy concepts, and the specific numerical range is determined by the membership function defined above.

[0077] Fuzzy reasoning process: For each rule, its activation degree must be calculated first. The activation degree indicates the matching degree between the current input working condition parameters and the premise conditions of the rule. The calculation method is to take the minimum value of the membership degree of each input variable in the premise condition in the corresponding fuzzy subset.

[0078] For example, for the above-mentioned rules, it is necessary to find the membership of the formation lithology hardness coefficient in the "high" subset and the membership of the mud flow resistance in the "high" subset, and then take the minimum value of these two memberships as the activation degree of the rule. This activation degree reflects the degree of compliance of the current working conditions with the rule. The higher the activation degree, the stronger the applicability of the rule to the current working conditions.

[0079] For the output of each rule, the calculated activation degree is used to truncate the membership function of the corresponding output variable, which means that the membership of the output variable in its fuzzy subset is adjusted according to the value of the activation degree, so that the output fuzzy subset is more in line with the current working conditions.

[0080] Defuzzification process: The outputs of all rules are synthesized, and then the final output value is calculated through the centroid method, that is, the specific weight correction of each output variable. The calculation process of the centroid method is to first calculate the central value of the fuzzy subset of each rule output, then multiply the activation of each rule by the corresponding central value and sum them, and finally divide it by the sum of the activations of all rules. The calculation result is the weight correction of each output variable, which is a specific value used for subsequent adjustment of the weight.

[0081] Weight modification constraints: In order to ensure that the sum of the corrected weights is always 1, a normalization factor is introduced. This factor is calculated by adding the inverse of the sum of the current initial weights and the respective weight corrections. Then, this normalization factor is multiplied by the initial weight of each output variable plus the weight correction to obtain the final corrected weight. This ensures that after considering the weight corrections of all objectives, their sum is still 1, thereby maintaining the rationality and effectiveness of the multi-objective evaluation function.

[0082] Through the above series of calculation processes, the fuzzy logic system can dynamically adjust the weight of each objective in the multi-objective evaluation function according to the downhole operating parameters collected in real time, so as to better adapt to different downhole operating conditions and achieve a more optimized drilling control strategy.

[0083] In a preferred embodiment of the present invention, step S3: according to the dynamic adjustment instruction generated in step S2, in the process of tracking the target azimuth, determining the first target point, the second target point and the third target point; constructing a dynamic triangle through the first target point, the second target point and the third target point, and using the centroid coordinates of the dynamic triangle as the azimuth correction value, including: The current tool face real-time azimuth is converted into the real-time position point in the polar coordinate system as the first target point. The wellhead center is taken as the coordinate origin, and the unit radius is taken to construct the azimuth coordinate plane. The coordinates of the first target point corresponding to the real-time azimuth are (cosθ 1 , sinθ 1 ), where θ 1 The tool face azimuth collected in real time; The preset target tool face angle is converted into a theoretical position point in the polar coordinate system as the second target point. The coordinates of the theoretical position point are (cosθ t, sinθ t ), where θ t is the preset target tool face angle; The third target point, i.e., the offset compensation point, is calculated based on the historical azimuth offset, and the direction of the line connecting the real-time position point and the theoretical position point is determined. The offset distance is determined in a proportional relationship according to the magnitude of the historical azimuth offset along the reverse extension line of the line. The real-time position point is used as the starting point, and the distance proportional to the magnitude of the historical offset is extended in the reverse direction of the line to obtain the coordinates of the third target point. The proportional relationship is pre-set according to the mapping relationship between the offset and the adjustment amplitude in the historical drilling data. The coordinates of the three target points in the polar coordinate system are converted into rectangular coordinates, a dynamic triangle is constructed with the three rectangular coordinate points, the arithmetic mean of the coordinates of the three vertices of the dynamic triangle is calculated as the centroid coordinates, the centroid coordinates are converted into angle values ​​in the polar coordinate system, and the angle values ​​are used as azimuth correction values.

[0084] In the embodiment of the present invention, the specific calculation process of step S3 can be implemented as follows: Step 1: Determine the first target point (real-time location point): Coordinate plane setting: With the wellhead center as the coordinate origin, an azimuth coordinate plane (polar coordinate system) with a unit radius is constructed, and all angles are measured with this origin as the reference; the tool face azimuth angle (denoted as θ 1 , such as the actual angle of the current tool face) is converted into a coordinate point in the polar coordinate system; in polar coordinates, the rectangular coordinates corresponding to any angle on the unit circle are (cosθ, sinθ), so the real-time azimuth angle θ 1 The corresponding coordinates of the first target point are (cosθ 1 , sinθ 1 ), which represents the actual position of the tool face on the azimuth plane.

[0085] Step 2: Determine the second target point (theoretical position point) Target angle conversion: Convert the preset target tool face angle (denoted as θ t , such as the tool face angle required by the design) is also converted into coordinate points in the polar coordinate system. According to the rule of converting polar coordinates to rectangular coordinates, the target angle θ t The corresponding coordinates of the second target point are (cosθ t , sinθ t ), which represents the theoretical target position that the tool face needs to track and is the ideal state for directional drilling.

[0086] Step 3: Determine the third target point (offset compensation point) Analyze historical offsets: review the historical azimuth offsets recorded in step S1 (i.e., the difference between the actual azimuth and the target angle), and calculate the offset amplitude (such as the average offset amplitude or the maximum offset amplitude) in a period of time before the current moment.

[0087] Determine the compensation direction and distance: Connection direction: Connect the first target point (real-time position) and the second target point (theoretical position) to form a straight line. The offset compensation point needs to be along the reverse extension line direction of the straight line (that is, starting from the real-time position point and extending in the direction away from the theoretical position point, which is opposite to the connection direction). According to the mapping relationship between the "offset size" and the "effective adjustment range" in the historical drilling data (for example, when the historical average offset is 10°, a compensation adjustment of 5° is required to offset the lag), a proportional coefficient (such as 30% of the offset amplitude) is pre-set.

[0088] Calculate the coordinates of the compensation point: start from the real-time position point and extend a distance in the reverse extension direction. The distance is proportional to the amplitude of the historical offset. For example, if the historical average offset amplitude is △θ, the compensation distance is k×△θ (k is the preset proportional coefficient), and finally the coordinates of the third target point are obtained.

[0089] Step 4: Construct dynamic triangle and calculate barycentric coordinates Coordinate unified conversion: The polar coordinates (angle-radius) of the three target points are uniformly converted into coordinates (x, y) in the rectangular coordinate system, and the three rectangular coordinate points (the first target point, the second target point, and the third target point) are used as vertices to form a dynamic triangle. The shape of the triangle changes with the real-time azimuth and historical offset.

[0090] Calculate the center of gravity: The center of gravity coordinates are the arithmetic mean of the coordinates of the three vertices, that is: Center of gravity x coordinate = (first target point x + second target point x + third target point x) ÷ 3; Center of gravity y coordinate = (first target point y + second target point y + third target point y) ÷ 3; The center of gravity represents the combined position of the three target points, integrating the current state, goal orientation, and historical compensation.

[0091] Step 5: Convert the center of gravity coordinates to azimuth correction values Polar coordinate conversion: Convert the rectangular coordinates (x, y) of the center of gravity into the angle value (i.e. azimuth) in the polar coordinate system, and calculate the angle between the center of gravity and the origin through the inverse tangent function. This angle is the azimuth correction value.

[0092] Correction logic: This correction value comprehensively considers the current tool face position (first point), the target position (second point) and the compensation position caused by the historical offset (third point), and is used to adjust the rotation direction and amplitude of the drill string so that the tool face angle gradually approaches the target through closed-loop adjustment.

[0093] In an embodiment of the present invention, the first point reflects the current state, the second point clarifies the target orientation, and the third point compensates for the adjustment lag caused by the drill string inertia or friction in advance through the historical offset to avoid overshoot or oscillation. The center of gravity, as the average position of the three, can smooth the deviations of different target points, make the correction direction closer to the actual dynamic needs, and improve the stability and accuracy of trajectory control. The compensation distance of the third target point is set based on the historical offset law, so that the system has self-learning ability and can adapt to different downhole working conditions (such as adjustment delays in high friction formations).

[0094] In a preferred embodiment of the present invention, step S4: correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction, including: Calculate the angle deviation between the azimuth correction value and the current tool face real-time azimuth; According to the circular characteristics of the azimuth angle, the minimum angle difference in the clockwise direction is taken as the effective adjustment deviation; According to the preset forward and reverse angle increment range in the dynamic adjustment instruction, the forward and reverse direction of the drill string and the corrected rotation angle are determined in combination with the effective adjustment deviation: if the correction value is greater than the real-time azimuth and the deviation is within the range of 0-180 degrees, it is determined to be a forward adjustment, and the rotation angle is the weighted average of the deviation value and the initial rotation angle increment; if the deviation exceeds 180 degrees, 360 degrees minus the deviation value is taken as the reverse adjustment angle, and the direction is counterclockwise; Based on the fluctuation frequency of the historical azimuth offset and the change rate of the center of gravity coordinates of the current dynamic triangle, the initial period parameters are corrected. That is, when the change rate of the center of gravity coordinates exceeds the preset threshold, the adjustment period is shortened to improve the response speed. When the offset fluctuation tends to be stable, the period is extended to reduce the mechanical load to obtain the corrected period parameters.

[0095] In the embodiment of the present invention, the specific calculation process of step S4 above can be as follows: The first step is to calculate the angle deviation: Initial deviation calculation: Subtract the azimuth correction value obtained in step S3 (representing the ideal angle after comprehensive current state, target and historical compensation) from the tool face azimuth (current actual angle) collected in real time to obtain the preliminary angle deviation. For example, if the correction value is 45°, the real-time azimuth is 30°, and the initial deviation is +15°; if the correction value is 350°, the real-time azimuth is 10°, and the initial deviation is -20° (or 340°, which requires subsequent processing of circular characteristics).

[0096] Step 2: Process the circumferential characteristics and determine the effective adjustment deviation Circular angle characteristics: Since the azimuth angle is a 360° cycle, direct subtraction may result in a deviation of more than 180° (for example, the deviation between 350° and 10° is 340°, but the actual minimum adjustment angle is 20°, which is the shortest path clockwise or counterclockwise).

[0097] Minimum angle difference calculation: Take the minimum angle difference in the clockwise direction as the effective adjustment deviation. The specific rules are as follows: If the absolute value of the initial deviation is ≤180°, the deviation is directly used as the effective adjustment deviation; If the absolute value of the initial deviation is greater than 180°, subtract the absolute value of the deviation from 360° to obtain the minimum angular difference in the clockwise direction (for example, when the deviation is -340°, the effective adjustment deviation is 20°, and the direction is clockwise).

[0098] Step 3: Determine the forward and reverse directions and the corrected rotation angle Determine the forward and reverse direction: Forward adjustment: If the correction value is greater than the real-time azimuth angle and the effective adjustment deviation is within the range of 0-180° (including 180°), it is determined to be in the forward direction (i.e. clockwise rotation).

[0099] Reverse adjustment: If the correction value is less than the real-time azimuth, or the effective adjustment deviation exceeds 180° (in this case, 360°-deviation value is taken as the actual adjustment angle, and the direction is counterclockwise).

[0100] Calculate the rotation angle: When rotating forward, the rotation angle is the weighted average of the effective adjustment deviation value and the initial rotation angle increment in the dynamic adjustment instruction (for example, the deviation value accounts for 60% of the weight and the initial increment accounts for 40% of the weight. The specific weight can be dynamically set according to the historical adjustment effect) to ensure that the adjustment range takes into account the current deviation and the preset parameters.

[0101] When reversing, the rotation angle is 360° minus the effective adjustment deviation value (for example, when the deviation is 340°, the reversal angle is 20°), and the direction is counterclockwise to avoid taking a long detour for adjustment (for example, from 350° to 10°, reversing 20° is more efficient than rotating 340° forward).

[0102] Step 4: Correct the cycle parameters Analyze historical drift and current rate of change: Historical offset fluctuation frequency: Review the historical azimuth offset data recorded in step S1 and calculate the recent offset fluctuation frequency (such as the number of times the offset changes exceed a certain threshold per unit time), reflecting the stability of the tool face angle.

[0103] Barycentric coordinate change rate: Calculate the position change rate of the dynamic triangle barycentric coordinates in two adjacent adjustment cycles (that is, the change amplitude of the x and y components of the barycentric coordinates per unit time), reflecting the dynamic trend of the current tool face adjustment.

[0104] Cycle adjustment strategy: Quick response scenario: If the rate of change of the center of gravity coordinate exceeds the preset threshold (such as the angle change per unit time> 5°), or the historical offset fluctuation frequency is high (such as more than 3 times per minute), it means that the tool face angle is unstable and the adjustment cycle needs to be shortened (such as from the initial cycle of 10 seconds to 5 seconds) to improve the system response speed and correct the deviation in time.

[0105] Stable scenario: If the rate of change of the center of gravity coordinate is low (for example, the angle change per unit time is less than 2°), and the historical offset fluctuation tends to be stable (for example, the offset change is less than 1° for 5 consecutive cycles), it means that the current adjustment is close to the target. Extend the adjustment cycle (for example, from 10 seconds to 15 seconds) to reduce the mechanical load caused by the frequent movement of the drill string and reduce equipment wear.

[0106] In the embodiments of the present invention, circular characteristic processing is used to avoid invalid large-angle rotation (such as a 340° deviation converted to a 20° reversal), thereby improving adjustment efficiency. The weighted average rotation angle takes into account the current deviation and preset parameters, balances the adjustment accuracy and mechanical load, and dynamically adjusts the cycle according to historical fluctuation patterns and the current rate of change, thereby responding in a timely manner under rapidly changing conditions, reducing the action frequency under stable conditions, and improving system robustness and equipment life.

[0107] In another preferred embodiment of the present invention, the above step S5: controlling the drill string action based on the modified dynamic adjustment instruction so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment, may include: The corrected drill string action parameters are obtained from step S4, including the forward and reverse directions (clockwise / counterclockwise), the corrected rotation angle (such as 20° forward rotation) and the adjustment period (such as 5 seconds), and the above parameters are converted into control signals (such as motor drive instructions) of the drill string drive system to drive the drill string to rotate in the specified direction and angle and complete the action within the preset period. For example, if the correction is "counterclockwise rotation of 15°, period of 8 seconds", the motor is controlled to rotate counterclockwise at a constant speed and stop after 8 seconds.

[0108] The current tool face azimuth (denoted as θ) is collected in real time through an inertial navigation system (such as MWD measurement while drilling tool) 2 ), which reflects the actual angle after the drill string moves; the newly collected azimuth angle θ 2 Angle θ with target tool face t Compare and calculate the current deviation △θ=θ 2 -θt (Taking into account the circular characteristics, take the minimum angle difference within the range of 0-360°.) For example, if the target angle is 90° and the current angle is 85°, the deviation is +5°; if the current angle is 350° and the target angle is 10°, the minimum deviation is 20° (counterclockwise).

[0109] Step 3: Determine whether to enter the next round of adjustment Accuracy check: Compare the current deviation △θ with the preset accuracy threshold (such as 1°): If △θ≤threshold, it is considered that the tool face has approached the target azimuth, the closed-loop adjustment is terminated, and the stable tracking state is entered; If △θ>threshold, it means that adjustment is still needed and enter the next round of iteration.

[0110] Data recording: The azimuth angle θ after this adjustment 2 , deviation △θ, actual rotation angle, cycle parameters, etc. are stored in the history database to provide a historical offset reference for subsequent adjustments (such as the compensation calculation of the third target point in step S3).

[0111] Step 4: Closed-loop iterative adjustment (if the deviation does not meet the standard) Restart the dynamic adjustment process: with the current azimuth angle θ 2 As the new real-time position point (first target point), the target angle θ t Still at the theoretical position point (second target point), calculate the third target point (offset compensation point) in combination with the latest historical offset (including the current deviation △θ), and repeat the process of steps S3-S4: Construct a new dynamic triangle and calculate the barycentric coordinates as the new azimuth correction value; The forward and reverse angles and cycle parameters are adjusted according to the new correction values ​​to generate the next dynamic adjustment instruction.

[0112] Iterative optimization logic: Each iteration is based on the latest real-time data and historical deviation rules, dynamically adjusting the compensation amplitude and response period to form a closed loop of "execution-measurement-correction-re-execution" to gradually reduce the deviation.

[0113] When the azimuth deviations collected for N consecutive times (such as 3 times) are all ≤ the preset threshold, or the rate of change of the center of gravity coordinates within the adjustment period approaches 0, it is determined that the tool face has stabilized near the target azimuth, and active adjustment is stopped and switched to real-time monitoring mode. After the adjustment is terminated, the tool face azimuth still needs to be monitored in real time. If the deviation suddenly increases due to factors such as formation disturbance (such as exceeding 2 times the threshold), a new round of closed-loop adjustment is automatically triggered to ensure that the tool face continues to be stable at the target angle.

[0114] In an embodiment of the present invention, through the cycle of "execution instruction → real-time feedback → deviation calculation → parameter correction", dynamic calibration of the tool face angle is achieved to avoid the blindness of a single adjustment and adapt to complex downhole environments (such as friction changes and drill string vibrations). The historical offset of each iteration is accumulated and recorded to optimize the compensation amplitude (such as the offset distance of the third target point) and cycle parameters of subsequent adjustments, so that the system has self-optimization capabilities and reduces manual intervention. The preset accuracy threshold and termination conditions ensure that the adjustment process is neither too frequent (reducing mechanical losses) nor responding quickly when disturbances occur (such as restarting the closed loop when the deviation suddenly increases), thereby balancing control accuracy and system stability.

[0115] like Figure 2 As shown, an embodiment of the present invention further provides an intelligent drilling guide system for a petroleum drilling tool face, comprising: An acquisition module is used to acquire the azimuth data of the downhole tool face in real time and simultaneously record the historical azimuth offset; A generating module, used for processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; A correction module is used to determine the first target point, the second target point and the third target point in the process of tracking the target azimuth according to the generated dynamic adjustment instruction; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; A processing module, used for correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; The control module is used to control the drill string action based on the corrected dynamic adjustment instruction, so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

[0116] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent drilling method for a petroleum drilling tool face, characterized in that: The method comprises: Step S1: acquiring the azimuth data of the downhole tool face in real time, and synchronously recording the historical azimuth offset; Step S2: Processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; Step S3: According to the dynamic adjustment instruction generated in step S2, in the process of tracking the target azimuth, determine the first target point, the second target point and the third target point; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; Step S4: correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; Step S5: Based on the corrected dynamic adjustment instruction, the drill string movement is controlled so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

2. The intelligent drilling method for oil drilling tool face according to claim 1, characterized in that: Step S1: Acquire the azimuth data of the downhole tool face in real time and simultaneously record the historical azimuth offset, including: The tool face azimuth data is collected in real time through a downhole sensor array, and the azimuth data is preprocessed to generate a preprocessed azimuth data stream; The historical azimuth data is stored in a preset time window length, and the historical azimuth offset is calculated.

3. The intelligent drilling method for oil drilling tool face according to claim 2, characterized in that: Step S2: Process the azimuth data in real time, and generate a dynamic adjustment instruction in combination with a preset target tool face angle. The dynamic adjustment instruction includes the forward and reverse angles and cycle parameters of the drill string, including: The real-time azimuth data stream, the historical azimuth offset and the preset target angle outputted in step S1 are inputted into a pre-trained artificial intelligence model, wherein the artificial intelligence model is a hybrid architecture based on a long short-term memory network and a convolutional neural network, and is used to learn the mapping relationship between the dynamic change of the azimuth and the control parameters of the drill string; The artificial intelligence model is used to extract the spatiotemporal characteristics of the azimuth data, and the real-time dynamic response of the drill string is predicted in combination with the current drilling parameters to obtain a prediction result; Generate an initial dynamic adjustment instruction according to the prediction result, wherein the initial dynamic adjustment instruction includes a start angle threshold for forward and reverse rotation of the drill string, a rotation angle increment, and a rotation cycle parameter, wherein the start angle threshold for forward and reverse rotation of the drill string is dynamically corrected by the amplitude and direction of the historical offset; The initial dynamic adjustment instruction and the target angle constraint conditions are subjected to multi-objective optimization, wherein the optimization conditions include the azimuth tracking accuracy, the mechanical load balance of the drill string and the energy consumption threshold of the adjustment action, and finally a feasible solution that meets the constraints is generated as the dynamic adjustment instruction.

4. The intelligent drilling method for oil drilling tool face according to claim 3, characterized in that: The initial dynamic adjustment instruction and the target angle constraint conditions are subjected to multi-objective optimization, wherein the optimization conditions include the azimuth tracking accuracy, the mechanical load balance of the drill string, and the energy consumption threshold of the adjustment action, and finally a feasible solution that satisfies the constraints is generated as a dynamic adjustment instruction, including: Based on the forward and reverse angle thresholds, rotation angle increments and cycle parameters in the initial dynamic adjustment instructions, a multi-objective evaluation function including azimuth tracking accuracy error, drill string axial torque fluctuation value and single adjustment energy consumption is constructed, where the weight coefficient of each objective is dynamically allocated according to the real-time working conditions downhole. The multi-objective evaluation function is iteratively solved by using a constrained particle swarm optimization algorithm, wherein the constraints include: the azimuth tracking accuracy error does not exceed a preset threshold, the drill string mechanical load fluctuation value is within a safe range, and the energy consumption of a single adjustment is lower than the rated power of the equipment; In the iterative process, the candidate solution set with the smallest azimuth tracking accuracy error is retained first, and the final instruction parameter combination of the mechanical load fluctuation value and the energy consumption comprehensive score is screened in the candidate solution set to generate the preliminary optimization instruction; Performing downhole dynamic response simulation verification on the preliminary optimization instruction, simulating the trajectory of tool face azimuth change after executing the preliminary optimization instruction through a preset drill string dynamics model, and determining it as a feasible solution if the azimuth in the simulation result converges to the target angle and meets the load and energy consumption constraints; If the simulation verification fails, the weight coefficient of the multi-objective evaluation function is adjusted according to the error direction, and new optimization instructions are re-iterated until a dynamic adjustment instruction that meets all constraints is output to obtain the final optimized dynamic adjustment instruction.

5. The intelligent drilling method for oil drilling tool face according to claim 4, characterized in that: Based on the forward and reverse angle thresholds, rotation angle increments and cycle parameters in the initial dynamic adjustment instructions, a multi-objective evaluation function including azimuth tracking accuracy error, drill string axial torque fluctuation value and single adjustment energy consumption is constructed, where the weight coefficient of each objective is dynamically allocated according to the real-time working conditions downhole, including: Real-time collection of downhole working parameters, including drilling pressure, drill string speed, formation lithology hardness coefficient and mud flow resistance, as the basis for weight allocation; Calculate the initial weight ratios of azimuth tracking accuracy error, torque fluctuation and energy consumption in the multi-objective evaluation function according to the operating condition parameters; Dynamically correcting the initial weight ratio through a pre-trained fuzzy logic rule base to obtain a corrected weight coefficient, wherein the rule base is constructed based on a mapping relationship between operating parameters and weight adjustments in historical drilling data; The corrected weight coefficients are respectively integrated with the azimuth tracking accuracy error, the drill string axial torque fluctuation value and the single adjustment energy consumption to form weighted parameter items corresponding to each objective. A multi-objective evaluation function matching the current downhole working conditions is constructed through the combination of various weighted parameter items.

6. The intelligent drilling method for oil drilling tool face according to claim 5, characterized in that: Step S3: According to the dynamic adjustment instruction generated in step S2, in the process of tracking the target azimuth, determine the first target point, the second target point and the third target point; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value, including: The current real-time azimuth of the tool face is converted into a real-time position point in the polar coordinate system as the first target point. The wellhead center is taken as the coordinate origin, and the unit radius is taken to construct the azimuth coordinate plane. The coordinates of the first target point corresponding to the real-time azimuth are (cosθ1, sinθ1), where θ1 is the tool face azimuth collected in real time; The preset target tool face angle is converted into a theoretical position point in the polar coordinate system as the second target point. The coordinates of the theoretical position point are (cosθ t , sinθ t ), where θ t is the preset target tool face angle; The third target point, i.e., the offset compensation point, is calculated based on the historical azimuth offset, and the direction of the line connecting the real-time position point and the theoretical position point is determined. The offset distance is determined in a proportional relationship according to the magnitude of the historical azimuth offset along the reverse extension line of the line. The real-time position point is used as the starting point, and the distance proportional to the magnitude of the historical offset is extended in the reverse direction of the line to obtain the coordinates of the third target point. The proportional relationship is pre-set according to the mapping relationship between the offset and the adjustment amplitude in the historical drilling data. The coordinates of the three target points in the polar coordinate system are converted into rectangular coordinates, a dynamic triangle is constructed with the three rectangular coordinate points, the arithmetic mean of the coordinates of the three vertices of the dynamic triangle is calculated as the centroid coordinates, the centroid coordinates are converted into angle values ​​in the polar coordinate system, and the angle values ​​are used as azimuth correction values.

7. The intelligent drilling method for a petroleum drilling tool face according to claim 6, characterized in that: Step S4: correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction, including: Calculate the angle deviation between the azimuth correction value and the current tool face real-time azimuth; According to the circular characteristics of the azimuth angle, the minimum angle difference in the clockwise direction is taken as the effective adjustment deviation; According to the preset forward and reverse angle increment range in the dynamic adjustment instruction, the forward and reverse direction of the drill string and the corrected rotation angle are determined in combination with the effective adjustment deviation: if the correction value is greater than the real-time azimuth and the deviation is within the range of 0-180 degrees, it is determined to be a forward adjustment, and the rotation angle is the weighted average of the deviation value and the initial rotation angle increment; if the deviation exceeds 180 degrees, 360 degrees minus the deviation value is taken as the reverse adjustment angle, and the direction is counterclockwise; Based on the fluctuation frequency of the historical azimuth offset and the change rate of the center of gravity coordinates of the current dynamic triangle, the initial period parameters are corrected. That is, when the change rate of the center of gravity coordinates exceeds the preset threshold, the adjustment period is shortened to improve the response speed. When the offset fluctuation tends to be stable, the period is extended to reduce the mechanical load to obtain the corrected period parameters.

8. An intelligent drilling guide system for a petroleum drilling tool face, the system implementing the method according to any one of claims 1 to 7, characterized in that: include: An acquisition module is used to acquire the azimuth data of the downhole tool face in real time and simultaneously record the historical azimuth offset; A generating module, used for processing the azimuth angle data in real time, and generating a dynamic adjustment instruction in combination with a preset target tool face angle, wherein the dynamic adjustment instruction includes a forward and reverse angle and a cycle parameter of the drill string; A correction module is used to determine the first target point, the second target point and the third target point in the process of tracking the target azimuth according to the generated dynamic adjustment instruction; construct a dynamic triangle through the first target point, the second target point and the third target point, and use the centroid coordinates of the dynamic triangle as the azimuth correction value; A processing module, used for correcting the forward and reverse rotation angles and cycle parameters of the drill string according to the azimuth correction value to obtain a corrected dynamic adjustment instruction; The control module is used to control the drill string action based on the corrected dynamic adjustment instruction, so that the downhole tool face continuously approaches the target azimuth through closed-loop iterative adjustment.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images