High-precision well-ground real-time geometric guidance closed-loop control method and control system

Abstract

The invention provides a high-precision well-ground real-time geometric guidance closed-loop control method and system, and the method comprises the steps: regenerating a new to-be-drilled track according to a design track when the deviation value of a continuous actual drilling track and the design track is greater than a given threshold value, and enabling the joint of the to-be-drilled track and the design track to be completely connected; the track to be drilled and the obtained drilling operation parameters are input into a rotary steering tool control instruction prediction model together, a new control instruction is generated, the control instruction is sent to an underground central control unit, and closed-loop control is formed; the drilling operation data comprises dynamic well deviation, dynamic direction, deflecting force size and direction of a steering tool, bit pressure, torque, rotating speed, displacement and drilling time, and the rotary steering operation instruction data comprises a downloading instruction list.The artificial intelligence prediction technology is applied to well track instruction control, so that the accuracy of control instructions is improved; therefore, the track control precision, quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of oil and gas exploration and development drilling, and in particular to a high-precision well-ground real-time geometric guidance closed-loop control method and control system. Background Art

[0002] In rotary steerable drilling, the actual borehole trajectory is usually calculated based on the downhole measurement while drilling data combined with the footage depth of the surface drill string using mature methods in the industry. Then, the deviation between the actual borehole trajectory on the ground and the expected borehole trajectory is used as the basis for trajectory adjustment decisions.

[0003] When it is determined that the borehole trajectory needs to be adjusted, usually, the trajectory is controlled by sending instructions from the ground to the well. Each time the trajectory needs to be adjusted, an instruction needs to be sent down, and each time the instruction is sent down, it takes from ten to dozens of minutes. Sometimes a well needs hundreds of instructions, which seriously affects the drilling efficiency. With the emergence of intelligent drill pipe technology, the downlink instructions can be sent down in real time around the clock, improving the control efficiency and making it possible to achieve real-time geometric guidance closed-loop drilling of the well on the ground.

[0004] However, in the self-feedback control of commands, how to timely feedback accurate control commands to the well is crucial to improving the control accuracy and trajectory quality of the trajectory. Some self-feedback control methods commonly used in industry are difficult to meet the requirements.

[0005] A real-time prediction method for steering instructions of a push-to-pull rotary steering tool with Chinese patent application number CN202310223456.3 can predict the next steering instruction by real-time feedback of inclination data, thereby optimizing the instruction feedback process. It uses the inclination data fed back in real time by the push-to-pull rotary steering tool and the steering instructions implemented during the steering tool operation to calculate the well inclination change rate and azimuth change rate in the well depth interval for each steering instruction, and then gives the steering force percentage and steering tool face angle required by the rotary steering tool based on the steering instruction model and the target point for the next drilling.

[0006] In the calculation of wellbore trajectory inclination, the well inclination change rate and azimuth change rate are very important reference data. In the current calculation method, the well inclination change rate and azimuth change rate are calculated using the well inclination angle change and azimuth change of the well section. However, in a well section, if the well inclination angle or azimuth angle does not change linearly with the well depth, then the well inclination change rate or azimuth change rate is not a constant. And when the well section is relatively long, the actual well inclination change rate and azimuth change rate have non-negligible deviations from the well inclination change rate and azimuth change rate of the constant value.

[0007] Therefore, the real-time prediction method of the push-type rotary steering tool guidance command is adopted to directly obtain the force and angle adjustment of the drill bit to the target point through the well inclination change rate and the azimuth change rate, which will directly lead to the defect of low accuracy of the control command obtained intelligently, resulting in low control accuracy and trajectory quality of the new trajectory during the adjustment process. Summary of the invention

[0008] The purpose of the present invention is to provide a high-precision real-time well-ground geometric guidance closed-loop control method and control system, apply new artificial intelligence prediction technology to wellbore trajectory prediction and control to improve the accuracy of control instructions, thereby improving the accuracy, quality and efficiency of trajectory control; at the same time, with the help of cabled drill pipe, the real-time transmission of control instructions is realized to improve the efficiency of instruction control.

[0009] The embodiment of the present invention is achieved as follows: A high-precision well-ground real-time geometric steering closed-loop control method, the control method comprising: When the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε, a new trajectory to be drilled is regenerated according to the designed trajectory, and the connection between the trajectory to be drilled and the designed trajectory is completely connected; The trajectory to be drilled and the acquired drilling operation parameters are input into the rotary steering tool control command prediction model to generate new control commands, which are sent to the downhole central control unit to dynamically adjust the resultant force and direction of the steering tool.

[0010] In a preferred embodiment of the present invention, the method for constructing the above-mentioned rotary steerable tool control instruction prediction model includes: Acquire completed drilling operation data and rotary steering operation instruction data, perform data preprocessing, and construct a valid data set; The pre-processed drilling operation data is used as input, and the rotary steering operation instruction data is used as output to construct a sample set; The sample set data is used to train, verify and test the machine learning regression model to obtain a mature rotary steerable tool control command prediction model.

[0011] In a preferred embodiment of the present invention, the above-mentioned drilling operation data include dynamic well inclination, dynamic azimuth, magnitude and direction of the deflecting force of the steering tool, drilling pressure, torque, rotation speed, displacement, and drilling time, and the rotary steering operation instruction data includes a downlink instruction table.

[0012] In a preferred embodiment of the present invention, the preprocessing method of the above dynamic well deviation and dynamic azimuth data is: Sorting the dynamic well inclination and dynamic azimuth data during the drilling operation of a single well in chronological order; A first sliding window with a width of W is formed with a drill rod length W; Assuming that the deflection capability of the steering tool is X° / Ym, calculate the median M1 of all measured values ​​in the first sliding window, and obtain the value determination interval [M1-X*W / Y, M1+X*W / Y]; Use the first sliding window to slide and judge and clean the dynamic well deviation and dynamic azimuth data of the entire well section, and delete the dynamic well deviation and dynamic azimuth data outside the judgment interval; The dynamic well deviation and dynamic azimuth data after cleaning are enhanced, and the enhanced dynamic well deviation and dynamic azimuth data are smoothed.

[0013] In a preferred embodiment of the present invention, the preprocessing method of the deflecting force magnitude and direction data of the above-mentioned guide tool is: According to different instructions, the magnitude and direction data of the deflection force corresponding to the instructions are filtered out; Arrange the deflection force magnitude and direction data according to the commanded start and end depths; Calculate the median M2 of all deflection force magnitude and direction data under the instruction condition; According to the fluctuation amplitude of the deflecting force size and direction data under the command conditions, the threshold adjustment proportional coefficient r1 is determined. The judgment interval of the deflecting force size and direction data is [M2- M2*r1, M2+ M2*r1], and the data outside the judgment interval is deleted; The magnitude and direction data of the deflecting force after cleaning are enhanced, and the enhanced magnitude and direction data of the deflecting force are smoothed.

[0014] In a preferred embodiment of the present invention, the preprocessing method of the above-mentioned drilling pressure, torque, speed, displacement, and drilling time data is: Arrange the drilling pressure, torque, speed, displacement and drilling time data in the order of depth; A second sliding window with a width of W is formed with a drill rod length W; Calculate the median M3 of the corresponding data type in the second sliding window; According to the fluctuation range of different types of data, the threshold adjustment proportional coefficient r2 is determined. The judgment interval of drilling pressure, torque, speed, displacement and drilling time data is [M3- M3*r2, M3+ M3*r2], and the data outside the judgment interval is deleted; The drilling pressure, torque, speed, displacement and drilling time data after cleaning are enhanced, and the enhanced drilling pressure, torque, speed, displacement and drilling time data are smoothed.

[0015] In a preferred embodiment of the present invention, the specific operation method of taking the pre-processed drilling operation data as input and taking the rotary steering operation instruction data as output is as follows: The input of each sample is: well inclination a at the drill bit, drill bit position b, drilling tool shape data c within n meters behind the drill bit, average values ​​d, e, f, g, h of drilling pressure, speed, torque, drilling time, and displacement within m meters in front of the drill bit, and well inclination change rate i and change direction j within m meters in front of the drill bit; The output of each sample is: the magnitude k and direction l of the resultant force of the guide tool within m meters in front of the drill bit.

[0016] In a preferred embodiment of the present invention, the processing method of the drilling tool shape data c within n meters behind the drill bit is: The drill tool shape data c within n meters behind the drill bit are subdivided into n meters of drill tool shape data c at intervals of 1 m to form n drill tool shape data.

[0017] In a preferred embodiment of the present invention, the method for using the above machine learning regression model for training, verification and testing is: The sample set is divided into a training set, a validation set and a test set. The test set is samples from one or more wells. The samples in the training set and the validation set are sample data from wells different from those in the test set. After shuffling the samples of multiple wells in the selected training set and validation set, 80% of the samples were randomly selected as the training set, and the remaining 20% ​​of the samples were used as the validation set.

[0018] In a preferred embodiment of the present invention, the above-mentioned machine learning regression model includes any one or more regression model combinations of SVM, BP, LSTM, and deep learning.

[0019] A high-precision well-ground real-time geometric steering closed-loop control system, the control system comprising: The drilling trajectory generation module is used to regenerate a new drilling trajectory according to the designed trajectory when the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε, and the drilling trajectory and the designed trajectory have at least m meters of identical joint segments; The rotary steering tool control instruction prediction module is used to input the trajectory to be drilled and the acquired drilling operation parameters into the rotary steering tool control instruction prediction model, generate new control instructions, and send the control instructions to the underground central control unit.

[0020] In a preferred embodiment of the present invention, the control system further comprises: A downhole measurement module, which measures and collects attitude information through a continuous attitude measurement sensor, and the continuous attitude measurement sensor is arranged at the rotary steering tool and close to the drill bit; Communication module: The communication module transmits attitude information to the ground system through the cabled drill pipe and receives control instructions from the ground; The actual drilling trajectory generation module, the ground system calculates the continuous actual drilling trajectory through the attitude information; The trajectory comparison module compares the continuous actual drilling trajectory with the designed trajectory and calculates the deviation Δ; when the deviation Δ is greater than the given threshold ε, the to-be-drilled trajectory generation module generates a new to-be-drilled trajectory, and uses the rotary steering tool control instruction prediction module to generate a new control instruction to be transmitted downhole; when the deviation Δ is less than or equal to the given threshold ε, the drilling parameters and the rotary steering tool control instructions are kept unchanged.

[0021] The beneficial effects of the embodiments of the present invention are: 1. When the deviation between the actual drilling trajectory and the designed trajectory is too large and needs to be corrected, artificial intelligence is used to redesign a new trajectory to be drilled, and then the new trajectory to be drilled and the drilling operation parameters are input into the rotary steering tool control command prediction model to form a control command, so that the obtained control command is more accurate; 2. After the new control instructions are issued, the attitude measurement sensor collects the drilling parameters of the guide device in real time and uploads them in real time through the cabled drill pipe to form a self-feedback process on the ground and underground, making the adjustment process of the entire guide tool a dynamic real-time adjustment process, rather than a simple point-to-point orientation and force adjustment. The adjustment process is more precise and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 This is a flow chart of a high-precision well-ground real-time geometric steering closed-loop control method according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the connection of a new trajectory to be drilled according to the first embodiment of the present invention; Figure 3 This is a flow chart of constructing a rotary steerable tool control instruction prediction model according to the first embodiment of the present invention; Figure 4 This is a schematic control block diagram of a high-precision well-ground real-time geometric steering closed-loop control system according to the second embodiment of the present invention; Figure 5 This is a schematic diagram of a downhole tool assembly according to a second embodiment of the present invention; Icons: downhole measurement module 101; communication module 102; actual drilling trajectory generation module 103; trajectory comparison module 104; to-be-drilled trajectory generation module 105; rotary steering tool control instruction prediction module 106. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0027] The described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, it will be appreciated by those skilled in the art that the technical scheme of the present application may be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the present application.

[0028] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0029] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0030] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0031] First embodiment See also Figure 1 This embodiment provides a high-precision well-ground real-time geometric steering closed-loop control method, the control method comprising: S1: The continuous attitude measurement sensor of the rotary steerable tool continuously completes attitude measurement and transmits the attitude measurement data to the ground system in real time and continuously through the cable drill pipe; S2: The ground system calculates the continuous actual drilling trajectory based on the attitude measurement data using the trajectory fitting algorithm commonly used in the industry; S3: Compare the continuous actual drilling trajectory with the designed trajectory, and calculate the deviation Δ according to a given calculation principle; for example, the deviation between the actual drilling trajectory and the designed trajectory at the current point, or the deviation between the actual drilling trajectory and the designed trajectory within a given distance, or any other feasible calculation method in the industry; S41: when the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε, a new trajectory to be drilled m meters ahead of the drill bit is regenerated according to the designed trajectory, where the value of m is determined by the artificial intelligence prediction model, so that the connection between the trajectory to be drilled and the designed trajectory is completely joined, that is, the well inclination and azimuth of the connection point between the trajectory to be drilled and the designed trajectory are completely consistent; S42: When the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is less than or equal to a given threshold ε, it is considered that the drilling requirement is met, and the drilling parameters and the control instructions of the rotary steerable tool are kept unchanged; S5: After the trajectory to be drilled in S41 is generated, the trajectory to be drilled and the acquired drilling operation parameters are input into the rotary steering tool control instruction prediction model to generate a new control instruction; S6: Feedback the new control command to the underground central control unit to dynamically adjust the resultant force and direction of the steering tool.

[0032] Repeat the above steps S1-6 to achieve high-precision well-ground self-feedback closed-loop control, and obtain the continuously changing actual drilling trajectory by obtaining the attitude measurement of the rotary steering tool in real time. Especially when the rotary steering tool needs to be corrected, please refer to Figure 2 In the trajectory splicing diagram, the black curve is the designed trajectory, and the blue dotted line is the actual drilling trajectory generated by fitting. When the deviation between the actual drilling trajectory and the designed trajectory is large, the regenerated trajectory to be drilled is not a linear trajectory from the drill bit to a certain point on the designed trajectory, but a curved variable trajectory with a constantly changing direction that can be adjusted according to real-time drilling data such as dynamic well inclination, dynamic azimuth, magnitude and direction of the deflecting force of the steering tool, drilling pressure, torque, rotation speed, displacement, drilling time, etc. Finally, the new trajectory to be drilled is combined with the designed trajectory, so that there is a trajectory junction point or trajectory junction section with completely consistent well inclination and azimuth to meet complex downhole drilling operations.

[0033] Specifically, the method for constructing the rotary steerable tool control instruction prediction model in S5 includes: S51: Acquire completed drilling operation data and rotary steering operation instruction data, perform data preprocessing, and construct a valid data set; This step is used to clean and enhance the drilled data. In the present technical solution, different methods are used to clean the outliers according to the characteristics of different data types, which can greatly reduce the interference of outliers on the prediction model and have a great impact on improving the prediction accuracy of instructions.

[0034] The main purpose of the cleaning method is to eliminate data whose feature quantities exceed the threshold. The specific threshold selection for each type of data can be adjusted according to the actual data.

[0035] Data enhancement is to fill in missing values ​​and encrypt and interpolate the data after cleaning along the well depth. The interpolation algorithm must meet the requirements of curve smoothness and have the characteristics of interpolation while providing data points, such as Akima interpolation method, cubic spline interpolation, etc.

[0036] The following is a description of the data preprocessing methods for each parameter type involved in the sample set. The data that need to be preprocessed include drilling data and rotary steering operation instruction data, wherein the drilling operation data at least includes dynamic well inclination, dynamic azimuth, the magnitude and direction of the deflecting force of the steering tool, drilling pressure, torque, speed, displacement, and drilling time, and the rotary steering operation instruction data includes a downlink instruction table.

[0037] Among them, ① the preprocessing method of dynamic well deviation and dynamic azimuth data is: The dynamic well deviation and dynamic azimuth data during the drilling operation of a single well are sorted in the order of occurrence time; A first sliding window with a width of W is formed with a drill rod length W. For example, in this embodiment, a conventional drill rod length of 10 meters is selected to form a 10*10 square sliding window; Assume that the deflection capability of the steering tool is X° / Ym. For example, the maximum deflection rate of the current downhole tool is 18° / 30m. Calculate the median M1 of all measured values ​​in the first sliding window to obtain the value determination interval [M1-X*W / Y, M1+X*W / Y]; The first sliding window is used to slide and judge and clean all the dynamic well deviation and dynamic azimuth data of the whole well section, and the dynamic well deviation and dynamic azimuth data outside the judgment interval of the well section corresponding to the sliding window are deleted, and the normal dynamic well deviation and dynamic azimuth data are retained, so as to improve the prediction accuracy; for the dynamic well deviation and dynamic azimuth data, since the drilling process is continuous, the changes of these two types of data are also continuous. According to this feature, the technical method uses the sliding window method to clean abnormal values, which can quickly and effectively complete the cleaning of abnormal data; Then, the dynamic well deviation and dynamic azimuth data after cleaning are enhanced, and the enhanced dynamic well deviation and dynamic azimuth data are smoothed. In this embodiment, all data enhancements use the cubic spline interpolation method. Among them, the selection of interpolation spacing is the key. If the spacing is too small, the amount of data will increase and redundant data will be easily generated. If the spacing is too large, the data continuity will be poor, which will affect the stability of the subsequent prediction model.

[0038] Since the change of wellbore trajectory itself is a slow process, taking the current maximum inclination rate of downhole tools as 18° / 30m as an example, the average change rate per meter is 0.6°. Calculated based on the well inclination change rate of 0.1°, the spacing is approximately 0.17m. In order to accurately describe the trajectory changes, the data interpolation spacing can be taken as 0.1m, which meets the actual engineering needs.

[0039] ② The cleaning method of the deflection force magnitude and direction data of the steering tool is: According to different instructions, the deflecting force size and direction data corresponding to the instructions are filtered out; the deflecting force size and direction data are arranged according to the starting and ending depths of the instructions; the median M2 of all deflecting force size and direction data under the instruction condition is calculated; according to the fluctuation amplitude of the deflecting force size and direction data under the instruction condition, the threshold adjustment proportional coefficient r1 is determined, and the judgment interval of the deflecting force size and direction data is [M2- M2*r1, M2+ M2*r1], and the data outside the judgment interval is deleted.

[0040] For the guide force size and direction data, due to the large differences between different control instructions, these two data show discontinuous characteristics during the instruction switching process, and it is difficult to clean multiple instruction data at the same time, but within an instruction interval, these two types of numbers basically fluctuate within a smaller range.

[0041] Based on this, this solution processes the data of each instruction interval separately, that is, the starting and ending depths of each instruction are taken as a processing unit. In this unit, the median M2 of all measured guiding force magnitudes or orientations is used as the benchmark, and a suitable threshold is selected according to the actual fluctuation range of the guiding force magnitude and orientation. For example, the threshold can be set as the ratio of the benchmark value, and the proportional coefficient r1 can be adjusted according to the actual effect, generally set to r=30%. Then, when the measured value q exceeds the interval [M2- M2*r1, M2+ M2*r1], it is judged as abnormal and deleted, and the rest is judged as normal value.

[0042] After performing the same processing on the units corresponding to each instruction interval, the cleaning of this type of data is completed. The thresholds of the guide force size and direction in each processing unit can be flexibly adjusted according to the actual effect to achieve the optimal cleaning effect.

[0043] ③The cleaning method of drilling pressure, torque, speed, displacement and drilling time data is as follows: The drilling pressure, torque, speed, displacement and drilling time data are arranged in order of depth; a second sliding window with a width of W is formed with the drill rod length W; the median M3 of the corresponding data type in the second sliding window is calculated; according to the fluctuation amplitude of different types of data, the threshold adjustment proportional coefficient r2 is determined, and the judgment interval of the drilling pressure, torque, speed, displacement and drilling time data is [M3-M3*r2, M3+ M3*r2], and the data outside the judgment interval is deleted.

[0044] For the data of drilling pressure, torque, rotation speed, displacement, drilling time, etc., they are all obtained by ground measurement, with few data anomalies, and the actual changes also have certain continuity characteristics. This embodiment adopts the sliding window method to clean abnormal values. The window length and benchmark are the same as the cleaning method of dynamic well deviation, that is, the window length W=10m, and the median M3 of the data in the window is used as the benchmark value. The selection of the threshold is the same as the cleaning method of the guide force size, and is set as the proportion of the benchmark value. The proportional coefficient r can be adjusted according to the actual effect, and can generally be set to r2=30%. When the measured value s exceeds the interval [M3-M3*r2, M3+M3*r2], it is judged as abnormal and deleted, and the rest is judged as normal value. According to this principle, the data abnormality cleaning of the entire well section is completed by sliding the window.

[0045] S52: After S51 is completed, the pre-processed drilling operation data is used as input, and the rotary steering operation instruction data is used as output to construct a sample set; the sample set construction in this step is highly related to the instruction output speed of the rotary steering drilling operation process, and is one of the key technical points of the present invention.

[0046] Specifically, the constructed sample set is as shown in Table 1:

[0047] Table 1 The input of each sample is: well inclination a at the drill bit, drill bit position b, drill tool shape data c within n meters behind the drill bit, average values ​​d, e, f, g, h of drilling pressure, rotation speed, torque, drilling time and displacement within m meters in front of the drill bit, and well inclination change rate i and change direction j within m meters in front of the drill bit; the output of each sample is: the average value of the resultant force k and direction l of the guide tool within m meters in front of the drill bit.

[0048] Among them, in terms of the drill tool shape within n meters behind the drill bit, n meters are subdivided at intervals of 1m to form n drill tool shape data, so as to describe the drill tool shape more carefully. It should be noted that all parameters of each sample only correspond to one rotary steering control instruction and cannot span two instructions.

[0049] S53: The sample set data is used to train, verify and test the machine learning regression model to obtain a mature rotary steerable tool control command prediction model.

[0050] The machine learning regression model is used for training, validation and testing as follows: The sample set is divided into training set, validation set and test set. The test set is samples from one or more wells. The samples in the training set and validation set are sample data from wells different from those in the test set. After shuffling the samples from multiple wells in the selected training set and validation set, 80% of the samples are randomly selected as the training set, and the remaining 20% ​​of the samples are used as the validation set.

[0051] In this embodiment, the sample set is divided into a training set, a validation set and a test set. To test the generalization of the prediction model, the test set can be selected as samples from one or several wells, while the training set and validation set are obtained from samples from other wells.

[0052] In terms of the division of training sets and validation sets, due to the particularity of drilling projects, the input and output parameters in samples of different operating sections are quite different. For example, the well inclination change rate in samples of the build-up section is usually larger, while the well inclination change rate in samples of the horizontal section is usually smaller. In order to make the training set cover samples with a wider parameter range, samples can be randomly selected from samples of multiple wells in a certain proportion (such as 80%) in a disordered manner as the training set, and the remaining 20% ​​of the samples are used as the validation set.

[0053] The construction principle of the training set, the validation set and the test set is closely related to the drilling operation process of the rotary steerable tool and is one of the key technical points of the present invention.

[0054] The machine learning regression model includes any one or more regression model combinations of SVM, BP, LSTM, and deep learning. For the training set and validation set constructed by the above method, a machine learning regression prediction method, such as SVM, BP, LSTM, deep learning, etc., is selected to train the training set composed of multiple samples. The parameter adjustment process in the specific training process needs to be carried out according to the characteristics of the specific algorithm. Through training, the control instruction prediction model of the rotary steering tool, that is, the combined force size and direction instruction prediction model of the steering tool, is obtained. The more wells that have been drilled for constructing the training set, the more samples there are, and the more accurate the prediction model is. In addition, the prediction model can be dynamically updated, that is, the drilled depth data of the well being drilled can be added to the training set to predict the control instructions in the subsequent drilling process of the well.

[0055] Therefore, in the actual drilling process, the actual drilling data is organized into input parameters as shown in Table 1 and input into the prediction model, so that the best control instructions can be obtained in real time, avoiding the problem of inaccurate control instructions caused by insufficient human experience, thereby reducing the number of downlink instructions and improving drilling efficiency and wellbore trajectory quality.

[0056] The control instruction prediction model is in implicit form and can be expressed by f, that is, (k,l)=f(a,b,c,d,e,f,g,h,i,j).

[0057] Taking the LSTM machine learning prediction method as an example, the training set constructed by the drilling data is trained, and the training model is optimized using the validation set to obtain the optimal model f1 under the current sample. The prediction example is shown in Table 2 below:

[0058] Table 2 This embodiment provides a high-precision real-time well geometry steering closed-loop control method, which can use new artificial intelligence prediction means to predict control instructions. In the prediction process, an efficient and accurate sample set is established as training, verification and test data for the prediction model. The instruction prediction is accurate and can effectively avoid the problem of inaccurate control instructions caused by lack of human experience. The intelligence level of the rotary steering tool can be improved, thereby improving the trajectory control accuracy, quality and efficiency.

[0059] Second embodiment A high-precision well-ground real-time geometric steering closed-loop control system, which applies the control method in the first embodiment, see Figure 4 and Figure 5 , the control system includes: A downhole measurement module 101, which measures and collects attitude information through a continuous attitude measurement sensor, and the continuous attitude measurement sensor is arranged at the rotary steering tool and close to the drill bit; Communication module 102, which transmits attitude information to the ground system through the cabled drill pipe and receives control instructions transmitted from the ground; The actual drilling trajectory generation module 103, the ground system calculates the continuous actual drilling trajectory through the posture information; The trajectory comparison module 104 compares the continuous actual drilling trajectory with the designed trajectory and calculates the deviation Δ; The drilling trajectory generation module 105 is used to regenerate a new drilling trajectory m meters ahead of the drill bit according to the designed trajectory when the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε; The rotary steering tool control instruction prediction module 106 is used to input the trajectory to be drilled and the acquired drilling operation parameters into the rotary steering tool control instruction prediction model, generate new control instructions, and send the control instructions to the downhole central control unit.

[0060] When the deviation Δ is less than or equal to a given threshold ε, the drilling parameters and the rotary steerable tool control instructions are kept unchanged.

[0061] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0062] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0063] The units described as separate components may or may not be physically separated. As a unit, it can be appreciated by those skilled in the art that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0064] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0065] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A high-precision real-time geometric steering closed-loop control method for wells, characterized in that: The control method comprises: When the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε, a new trajectory to be drilled is regenerated according to the designed trajectory so that the connection between the trajectory to be drilled and the designed trajectory is completely connected; The trajectory to be drilled and the acquired drilling operation parameters are input into a rotary steering tool control instruction prediction model to generate a new control instruction, and the control instruction is sent to a downhole central control unit to dynamically adjust the resultant force size and direction of the steering tool; The method for constructing the rotary steerable tool control instruction prediction model comprises: Acquire completed drilling operation data and rotary steering operation instruction data, perform data preprocessing, and construct a valid data set; The pre-processed drilling operation data is used as input, and the rotary steering operation instruction data is used as output to construct a sample set; The sample set data is trained, verified and tested in a machine learning regression model to obtain a mature rotary steerable tool control command prediction model; The drilling operation data includes dynamic well inclination, dynamic azimuth, magnitude and direction of the deflecting force of the steering tool, drilling pressure, torque, rotation speed, displacement, and drilling time, and the rotary steering operation instruction data includes a downlink instruction table.

2. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 1 is characterized in that: The preprocessing method of the dynamic well deviation and dynamic azimuth data is as follows: Sorting the dynamic well inclination and dynamic azimuth data during the drilling operation of a single well in chronological order; A first sliding window with a width of W is formed with a drill rod length W; Assuming that the deflection capability of the steering tool is X° / Ym, calculate the median M1 of all the measured values ​​in the first sliding window to obtain the value determination interval [M1-X*W / Y, M1+X*W / Y]; Use the first sliding window to slide and judge and clean the dynamic well deviation and dynamic azimuth data of the entire well section, and delete the dynamic well deviation and dynamic azimuth data outside the judgment interval; The dynamic well deviation and dynamic azimuth data after cleaning are enhanced, and the enhanced dynamic well deviation and dynamic azimuth data are smoothed.

3. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 1 is characterized in that: The preprocessing method of the deflecting force magnitude and direction data of the steering tool is as follows: According to different instructions, the magnitude and direction data of the deflecting force corresponding to the instructions are filtered out; Arranging the deflecting force magnitude and direction data according to the starting and ending depths of the instructions; Calculate the median M2 of all deflection force magnitude and direction data under the instruction condition; According to the fluctuation amplitude of the deflecting force size and direction data under the command conditions, the threshold adjustment proportional coefficient r1 is determined, and the judgment interval of the deflecting force size and direction data is [M2- M2*r1, M2+ M2*r1], and the data outside the judgment interval is deleted; The cleaned deflecting force magnitude and direction data are enhanced, and the enhanced deflecting force magnitude and direction data are smoothed.

4. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 1 is characterized in that: The preprocessing method of the drilling pressure, torque, speed, displacement and drilling time data is as follows: Arrange the drilling pressure, torque, speed, displacement, and drilling time data in order of depth; A second sliding window with a width of W is formed with a drill rod length W; Calculate the median M3 of the corresponding data type in the second sliding window; According to the fluctuation range of different types of data, the threshold adjustment ratio coefficient r2 is determined. The determination interval of the drilling pressure, torque, speed, displacement and drilling time data is [M3- M3*r2, M3+ M3*r2], and the data outside the determination interval is deleted; The drilling pressure, torque, speed, displacement and drilling time data after cleaning are enhanced, and the enhanced drilling pressure, torque, speed, displacement and drilling time data are smoothed.

5. The high-precision real-time well geometry steering closed-loop control method according to claim 1 is characterized in that: The specific operation method of taking the pre-processed drilling operation data as input and taking the rotary steering operation instruction data as output is as follows: The input of each sample is: well inclination a at the drill bit, drill bit position b, drilling tool shape data c within n meters behind the drill bit, average values ​​d, e, f, g, h of drilling pressure, speed, torque, drilling time, and displacement within m meters in front of the drill bit, and well inclination change rate i and change direction j within m meters in front of the drill bit; The output of each sample is: the magnitude k and direction l of the resultant force of the guide tool within m meters in front of the drill bit.

6. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 5 is characterized in that: The processing method of the drilling tool shape data c within n meters behind the drill bit is: The drill tool shape data c within n meters behind the drill bit are subdivided into n meters of drill tool shape data c at intervals of 1 m to form n drill tool shape data.

7. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 1 is characterized in that: The method of using the machine learning regression model for training, verification and testing is as follows: Dividing the sample set into a training set, a validation set and a test set, wherein the test set is samples from one or more wells, and the samples in the training set and the validation set are sample data from wells different from those in the test set; After shuffling the selected training set and multiple well samples of the validation set, 80% of the samples were randomly extracted as the training set, and the remaining 20% ​​of the samples were used as the validation set.

8. The high-precision real-time well-ground geometric steering closed-loop control method according to claim 1 is characterized in that: The machine learning regression model includes any one or more regression model combinations of SVM, BP, LSTM, and deep learning.

9. A high-precision real-time well-ground geometric guidance closed-loop control system, characterized in that: The control system comprises: A drilling trajectory generation module is used to regenerate a new drilling trajectory m meters ahead of the drill bit according to the designed trajectory when the deviation Δ between the continuous actual drilling trajectory and the designed trajectory is greater than a given threshold ε; The rotary steering tool control instruction prediction module is used to input the trajectory to be drilled and the acquired drilling operation parameters into the rotary steering tool control instruction prediction model, generate new control instructions, and send the control instructions to the underground central control unit.

10. The high-precision well-ground real-time geometric steering closed-loop control system according to claim 9, characterized in that: The control system further comprises: A downhole measurement module, wherein the downhole measurement module measures and collects attitude information through a continuous attitude measurement sensor, wherein the continuous attitude measurement sensor is arranged at the rotary steering tool and close to the drill bit; A communication module, which transmits the attitude information to a ground system through a cabled drill pipe and receives control instructions transmitted from the ground; A real drilling trajectory generation module, wherein the ground system calculates a continuous real drilling trajectory through the posture information; The trajectory comparison module compares the continuous actual drilling trajectory with the designed trajectory and calculates the deviation Δ; when the deviation Δ is greater than a given threshold ε, the to-be-drilled trajectory generation module generates a new to-be-drilled trajectory, and uses the rotary steering tool control instruction prediction module to generate a new control instruction to be transmitted downhole; when the deviation Δ is less than or equal to the given threshold ε, the drilling parameters and the rotary steering tool control instructions are kept unchanged.

Images