Downhole drilling build-up rate data cleaning and build-up rate prediction method

Abstract

According to the downhole drilling build-up rate data cleaning and build-up rate prediction method, the speed and precision of build-up rate prediction can be remarkably improved, and effective help is provided for field engineers to control the future attitude of a drilling tool. The invention relates to an underground drilling build-up rate data cleaning method, which comprises the following steps of: pre-cleaning: cleaning and removing error instruction data and rotation steering data related to instruction data mapping for acquired instruction data and rotation steering data related to instruction data mapping; cleaning is carried out again, abnormal value cleaning is carried out on the rotary steering data through a sliding window, and abnormal value processing is carried out through a Z-score method in abnormal value cleaning; and performing data enhancement on the recleaned rotation steering data by using a Makima interpolation method, and forming a data set by using the processed instruction data and the rotation steering data.

Description

Technical Field

[0001] The invention relates to the field of drilling engineering, and in particular to a method for cleaning and predicting downhole drilling buildup rate data. Background Art

[0002] In drilling engineering, rotary steerable downhole tools are widely used in various wells. They have the ability to improve drilling efficiency and accurately control the direction and position of the wellbore, and can control the direction and angle of the wellbore during the drilling process. These tools can be widely used in horizontal wells, horizontal well section jumps, push wells, and steerable drilling under complex geological conditions. However, due to the complexity of the downhole environment, the state of the drill bit cannot be accurately controlled, which may lead to the inability to adjust the drill bit posture in time, and thus miss the drilling target area.

[0003] In the prior art, there are many methods for predicting the inclination rate of the drilling process, such as a method and system for predicting the inclination rate of the drilling process based on genetic algorithm optimization, with patent application number: CN201911319688.9, which includes the following processing steps: 1. Acquiring data; 2. Pearson correlation coefficient analysis and processing; 3. Normalization processing; 4. Using genetic algorithm to optimize BP neural network algorithm to build a prediction model for the overall data. In data processing, only the Pearson correlation coefficient analysis method is used to remove data with a small correlation coefficient to obtain the initial data set. Due to the complexity of the drilling operation process and the many interference factors, the acquired data has some anomalies or missing. If these data are used for modeling, the prediction results may be biased, and only correlation analysis is used to exclude irrelevant data. There are still a large number of errors, anomalies, missing or repeated data, which will reduce the processing speed and processing accuracy of the model, resulting in inaccurate inclination rate. Summary of the invention

[0004] The purpose of the present invention is to provide a downhole drilling buildup rate data cleaning and buildup rate prediction method, which can significantly improve the speed and accuracy of buildup rate prediction and provide effective assistance for on-site engineers to control the future posture of drilling tools.

[0005] The embodiment of the present invention is achieved as follows:

[0006] A method for cleaning downhole drilling deflection rate data, the data cleaning method comprising the following steps:

[0007] Pre-cleaning, cleaning and removing erroneous instruction data and rotary steering data related to instruction data mapping from the acquired instruction data and instruction data mapping;

[0008] Then, the rotary steering data is cleaned by using a sliding window to clean outliers, and the Z-score method is used to process outliers;

[0009] The Makima interpolation method is used to enhance the data of the rotary steering data after cleaning, and the processed instruction data and rotary steering data are formed into a data set.

[0010] In a preferred embodiment of the present invention, the above-mentioned rotary steering data include: near-bit well deviation, magnitude of resultant horizontal force, magnitude of resultant vertical force, drilling time, drilling pressure, torque, rotation speed, and pump stroke rate.

[0011] In a preferred embodiment of the present invention, the step of using the Makima interpolation method to perform data enhancement on the cleaned rotary steerable data comprises:

[0012] Calculate the slope between data points of the same type of rotational steerable data;

[0013] Use quadratic extrapolation to add virtual points at both ends of the sequence of data points;

[0014] When the slope at a data point changes suddenly, the Makima interpolation method is used to adjust the slope;

[0015] A cubic polynomial is constructed using the adjusted slope, and the slope is weighted based on the cubic polynomial so that the data points are continuous and smoothly connected;

[0016] For a query point within the range of known data points, find the data point interval corresponding to the query point, and obtain the enhanced result through the cubic polynomial calculation of the interval.

[0017] In a preferred embodiment of the present invention, the steps of using the Z-score method to process outliers include:

[0018] Calculate the mean μ and standard deviation σ of all data in the sliding window, and calculate Z = (x-μ)σ -1 ;

[0019] If |Z|>threshold, the data x is judged as an outlier.

[0020] In a preferred embodiment of the present invention, let the discrete data points (x t ,y t ), the slope between data points is calculated as:

[0021] δ i =(y i+1 -y i ) / (x i+1 -x i ), where δ i For data points (x i , xi+1 ) between .

[0022] In a preferred embodiment of the present invention, the calculation method of the quadratic extrapolation method is:

[0023] Edge points x1 and x n Required slope δ -1 , δ0 and δ n , δ n-1 , using the quadratic extrapolation method to calculate it as δ0 = 2δ1-δ2, δ -1 =2δ0-δ1 and δ n =2δ n-1 -δ n-2 , δ n+1 =2δ n -δ n-1 .

[0024] 7. According to the method for cleaning downhole drilling slope data of claim 3, the method for weighting the slope based on a cubic polynomial comprises:

[0025] Slope δ i-1 and δ i The weighted value calculation formula between is: Where w1 = |δ i+1 -δ i |+(|δ i+1 +δ i |)2 -1 , w2=|δ i-1 -δ i-2 |+(|δ i-1 +δ i-2 |) 2-1 .

[0026] A method for predicting the build-up rate of downhole drilling, using any of the above-mentioned data cleaning methods for predicting the build-up rate of downhole drilling.

[0027] In a preferred embodiment of the present invention, the prediction method comprises the following steps:

[0028] Construct a slope prediction model;

[0029] Read real-time rotary steering data, organize the rotary steering data into input format and import it into the slope prediction model;

[0030] The build rate model prediction model calculates and outputs the predicted build rate.

[0031] In a preferred embodiment of the present invention, the method for constructing a build slope prediction model comprises the following steps:

[0032] Acquire historical drilling data, sort and calculate the historical drilling data to obtain rotary steering data, map the command data and the corresponding rotary steering data to form a rotary steering operation database;

[0033] Clean and enhance the data in the rotary steerable operations database;

[0034] Based on the cleaned and enhanced telescopic rotary steerable operation database data, the Pearson correlation coefficient between various data and the slope rate is calculated, and the data with high correlation coefficients are selected to construct the training set, verification set and test set;

[0035] Integrate various types of adaptive hyperparameter adjustment methods to form a model hyperparameter tuning method library, and select appropriate adaptive hyperparameter adjustment methods for different types of data for parameter adjustment;

[0036] The LSTM neural network model is trained using the adjusted training set, validation set, and test set data to obtain an optimized downhole drilling inclination rate prediction model.

[0037] The beneficial effect of the embodiment of the present invention is: the data cleaning method of the present invention is used to perform multiple cleaning, enhancement, and correlation analysis on the command data and rotary steering data in the drilling process, and the near-drill bit inclination, the magnitude of the resultant horizontal force, the magnitude of the resultant vertical force, drilling time, drilling pressure, torque, rotation speed, and pump stroke combined data are selected as the input features of the inclination rate prediction model. Combined with the artificial intelligence algorithm and the neural network learning model, the inclination rate prediction model is constructed, which can improve the prediction speed and accuracy of the future inclination rate of the drilling tool, and assist on-site engineers in controlling the future posture of the drilling tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] 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.

[0039] Figure 1 This is a flow chart of a method for predicting the downhole drilling build-up rate according to an embodiment of the present invention;

[0040] Figure 2 A flow chart of a method for constructing a slope prediction model according to an embodiment of the present invention;

[0041] Figure 3 A flow chart of a data cleaning method according to an embodiment of the present invention;

[0042] Figure 4A flow chart of a method for performing data enhancement using the Makima interpolation method on the cleaned rotary steerable data according to an embodiment of the present invention;

[0043] Figure 5 Schematic diagram of a downhole drilling build-up rate prediction system according to an embodiment of the present invention.

[0044] Icon: 101-building slope rate prediction model construction unit; 102-building slope rate prediction model calling unit. DETAILED DESCRIPTION

[0045] 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.

[0046] 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.

[0047] 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.

[0048] 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.

[0049] In addition, described feature, structure or characteristic can 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, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0050] 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.

[0051] 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.

[0052] The main purpose of the present invention is to design a downhole drill tool inclination rate prediction method for accurately grasping the drill tool status and predicting the downhole drill tool inclination rate according to the drill tool status, thereby helping drilling engineers to better control the drill tool posture and achieve drilling goals.

[0053] First embodiment

[0054] See also Figure 1 This embodiment provides a method for predicting the build-up rate of downhole drilling, which cleans historical downhole drilling tool instrument data and well site data, and combines artificial intelligence methods to build a model, thereby predicting the build-up rate of downhole drilling tools. The prediction method includes the following steps:

[0055] S1: Construct slope prediction model;

[0056] S2: Read the real-time rotary steering data, organize the rotary steering data into an input format and import it into the slope rate prediction model; the slope rate prediction model calculates and outputs the predicted slope rate.

[0057] The key to the accuracy of prediction lies in the construction of the model and the selection and processing of training data. This example will focus on these two aspects. For details, please refer to Figure 2 ,The method for constructing the slope rate prediction model includes the following steps:

[0058] S11: Acquire historical drilling data, sort and calculate the historical drilling data to obtain rotary steering data, map the command data and the corresponding rotary steering data to form a rotary steering operation database, which includes original drilling data and calculated data, wherein the original drilling data includes well depth, well inclination, actual resultant force size, actual resultant force direction, downlink command, drilling time, drilling pressure, torque, rotation speed, pump stroke speed; the calculated data includes actual horizontal force size, actual vertical force size, and slope rate, see Table 1.

[0059] Table 1 Data sheet information

[0060]

[0061] In this step, by extracting useful features again based on the original drilling data, it is convenient to help the model better understand the nature of the data, discover the patterns and laws hidden behind the data, and make correct predictions.

[0062] In this embodiment, based on the well depth, well inclination, actual resultant force size, actual resultant force direction, downlink instruction, drilling time, drilling pressure, torque, rotation speed, and pump stroke speed, new characteristic indicators of actual horizontal force size, actual vertical force size, and inclination rate are generated for each sample point through parameter combination and interaction. The 10 parameters of well depth, well inclination, drilling pressure, drilling time, torque, rotation speed, pump stroke speed, actual horizontal force size, actual vertical force size, and inclination rate 8m behind the drill bit are selected as the input features of the inclination rate prediction model.

[0063] Since the various original drilling data are not independent of each other, they have a relationship of mutual influence and mutual restriction. Compared with a single parameter, the new characteristic index composed of multiple parameters in this embodiment can better reflect the changes in conditions during the drilling process.

[0064] S12: Clean and enhance the data in the rotary steering operation database. The data cleaning and enhancement steps directly affect the accuracy of sample features, model learning, and the accuracy of output prediction results. The processing process will be described in detail in the subsequent description.

[0065] S13: Based on the cleaned and enhanced telescopic rotary steering operation database data, the Pearson correlation coefficient between various data and the build-up rate is calculated, and the data with high correlation coefficients are selected to construct the training set, validation set and test set. This step adds correlation analysis and data weighting changes on the basis of general data set division. Since there are many factors affecting the build-up rate, the correlation coefficient between each factor and the actual build-up rate is obtained through known drilling data, and the sensitivity of the actual build-up rate to each influencing factor is analyzed, thereby reducing the error caused by the order of magnitude of each factor to improve the prediction accuracy of the build-up rate.

[0066] In this embodiment, the division ratio of the training set, the validation set, and the test set may be 7:1:2.

[0067] S14: Integrate various types of adaptive hyperparameter adjustment methods to form a model hyperparameter tuning method library, select appropriate adaptive hyperparameter adjustment methods for different types of data to adjust parameters, select the best hyperparameter combination for each algorithm, and compare and verify the prediction results between different models; use the adjusted training set, validation set and test set data to train the LSTM neural network model to obtain an optimized downhole drilling inclination rate prediction model, and obtain the optimal prediction model. Using different hyperparameters for each type of data verifies the feasibility of the artificial intelligence optimization algorithm and neural network disclosed in this embodiment in processing the inclination rate prediction optimization problem.

[0068] In this embodiment, when processing complex high-dimensional input data, if only LSTM neural network is used for processing, its convergence speed is fast but the accuracy is low. For this reason, intelligent optimization algorithms are used in the early steps of data cleaning, data set division, and hyperparameter adjustment. Its convergence speed is slow but the accuracy is high, which can overcome the defect of using only neural network to realize slope prediction.

[0069] Based on the constructed data set, various adaptive hyperparameter adjustment methods (grid search, random search, Bayesian search, genetic algorithm, etc.) are integrated to form a model hyperparameter tuning method library. The long short-term memory recurrent neural network (LSTM) is used to train the adaptive hyperparameter prediction model to obtain the best model and use it for the prediction of the inclination rate of downhole drilling tools, so as to quickly feedback the future posture information of the downhole drilling tools and assist on-site engineers to timely adjust the drilling tool-related instructions.

[0070] In the field of machine learning, there are two ways to deal with outliers: direct deletion and filling. In this embodiment, a new design is made for the data cleaning step. For details, please refer to Figure 3 , the data cleaning method in S12 includes the following steps:

[0071] S121: Pre-cleaning, cleaning the acquired instruction data and the rotary steering data related to the instruction data mapping to remove erroneous instruction data and the rotary steering data related to the instruction data mapping, so as to remove the erroneous instruction data and delete the drilling data in the process of the erroneous instruction data.

[0072] S122: Re-cleaning: using a sliding window to clean the rotary steering data for outliers. The outlier cleaning uses a Z-score method to process outliers, so as to remove outliers from the correct rotary steering data during the continuous drilling process.

[0073] Different drilling parameters usually have different orders of magnitude and units. If the original data is used directly, the parameters with higher values ​​may dominate the results, while the parameters with smaller values ​​will have a weakened impact on the results. Based on this, the Z-score method is used to process outliers during the re-cleaning process. The operation steps include:

[0074] Select a sliding window to capture data. Usually, a square window with a sliding window size of 11 is selected.

[0075] Calculate the mean μ and standard deviation σ of all data in the sliding window, and calculate Z = (x-μ)σ -1 , x is a data point;

[0076] If |Z|>threshold, the data x is judged as an outlier. Usually the threshold is selected as 3, that is, |Z|>3, and x is identified as an outlier.

[0077] The well inclination value range after processing in this step is 0° to 180°. By changing the distribution structure of the original data, the data changes continuously, approaches the center and does not contain sudden abnormal values. In particular, the actual combined force data needs to be decomposed into the actual horizontal force and the actual vertical force, and then the Z-score method is used for abnormal value processing. For other data, the Z-score method is used to remove abnormal data that exceeds the set threshold and data that is not within the normal range, and the data is standardized.

[0078] Compared with data cleaning of original data, using instruction data and rotary steering data related to instruction data mapping as data sets in the cleaning process, and performing data cleaning on the sub-data sets corresponding to each instruction respectively can better fit the characteristics of the data itself and reduce the information bias caused by box plot method data cleaning.

[0079] S123: The cleaned rotary steering data is enhanced using the Makima interpolation method, and the processed instruction data and rotary steering data are formed into a data set.

[0080] For details, see Figure 4 , the Makima interpolation method is used to enhance the data of the re-cleaned rotary steering data. Compared with the traditional interpolation method, the interpolation surface is smoother and closer to the drilling change trend. The specific operation steps include:

[0081] S1231: Calculate the slopes between data points of the same type of rotational steering data. First, calculate the slopes between each data point. These slopes are used to construct a local cubic polynomial. Suppose the discrete data point (x t ,y t ), the slope between data points is calculated as: δ i =(yi+1 -y i ) / (x i+1 -x i ), where δ i For data points (x i , x i+1 ) between .

[0082] S1232: Use quadratic extrapolation to add virtual points at both ends of the data point sequence to ensure that the slope of the boundary points can also be calculated correctly. Specifically, at the edge points x1 and x n Required slope δ -1 , δ0 and δ n , δ n-1 , but the slope of these points of the input data is not available, so it is calculated using quadratic extrapolation as: δ0 = 2δ1-δ2, δ -1 =2δ0-δ1 and δ n =2δ n-1 -δ n-2 , δ n+1 =2δ n -δ n-1 .

[0083] S1233: When the slope of a data point changes suddenly, the Makima interpolation method is used to adjust the slope; a cubic polynomial is constructed using the adjusted slope, and the slope is weighted based on the cubic polynomial to make the data points continuous and smoothly connected.

[0084] Between each data point, cubic polynomials are constructed using adjusted slopes that are continuous at the data points and smoothly connect between the data points. i The derivative of is through the five points around χ i-2 , x i-1 , x i , x i+1 , x i+2 The calculated value is defined as [x i-1 , x i ) and [x i , x i+1 ) has a slope δ i-1 and δ i The weighted value between:

[0085]

[0086] The weight formula is: i =|δ i+1 -δ i |+(|δ i+1 +δ i |)2 -1 , w2=|δi-1 -δ i-2 |+(|δ i-1 +δ i-2 |)2 -1 ,

[0087] When δ i =δ i+1 = 0, the additional 2 -1 (|δ i+1 +δ i |) and 2 -1 (|δ i-1 +δ i-2 |) will make d i =0, that is, y i =y i+1 =y i+2 Therefore, when the data remains the same continuously at more than two consecutive nodes, no overshoot occurs.

[0088] For any query point (x q ), find the string containing (x q ) data point interval, and then use the weighted calculation method to construct a cubic polynomial in this interval to calculate (y q ).

[0089] Special case handling: If the data is constant y i-2 =y i-1 =y i =y i+1 =y i+2 , then all four slopes are equal to zero, and we get d i =NaN, for the special case of constant data set d i =0.

[0090] S1234: For a query point within the known data point range, find the data point interval corresponding to the query point, and obtain the enhanced result through the cubic polynomial calculation of the interval, and finally form a data set in the format of drilling pressure, drilling time, rotation speed, pump stroke rate, torque, near-bit inclination, resultant horizontal force, resultant vertical force, and slope rate 8m before the bit with a well depth interval of 0.1m.

[0091] In the data set division step in S13, a correlation coefficient processing step is also introduced to calculate the Pearson correlation coefficient between each type of data and the slope rate, and arrange the correlation coefficients. For example, a correlation coefficient of 0.8-1 is a super strong correlation factor, a correlation coefficient of 0.7-0.8 is a strong correlation factor, a correlation coefficient of 0.6-0.7 is a relatively strong correlation factor, a correlation coefficient of 0.5-0.6 is a weak correlation factor, a correlation coefficient of 0.3-0.5 is a relatively weak correlation factor, and a correlation coefficient of less than 0.3 is a non-correlation factor.

[0092] The weighting coefficients are assigned in turn according to the strength of the influencing factors, for example, the super strong correlation factor is 0.9, the strong correlation factor is 0.75, the relatively strong correlation factor is 0.65, the weak correlation factor is 0.55, the relatively weak correlation factor is 0.25, and the no correlation factor is 0. In this way, irrelevant factors can be proposed. With the addition of more and more high-influence factors and the elimination of low-influence factors, the prediction accuracy of the slope prediction model will be further improved, and the prediction speed will also be further improved, see Table 2.

[0093]

[0094] In summary, the inclination rate prediction method in this embodiment is adopted. By performing multiple cleaning, enhancement, and correlation analysis on the command data and rotary steering data in the drilling process, the near-drill bit inclination, the magnitude of the resultant horizontal force, the magnitude of the resultant vertical force, drilling time, drilling pressure, torque, rotation speed, and pump stroke rate are selected as the input features of the inclination rate prediction model. Combined with the artificial intelligence algorithm and the neural network learning model, an inclination rate prediction model is constructed, which can improve the prediction speed and accuracy of the future inclination rate of the drilling tool and assist on-site engineers in controlling the future posture of the drilling tool.

[0095] Second embodiment

[0096] This embodiment provides a downhole drilling buildup rate prediction system based on the downhole drilling buildup rate prediction method of the first embodiment, which is mainly used for downhole rotary steerable drilling tool operations in oil drilling projects and assists in downhole drilling tool posture analysis. It includes:

[0097] The build rate prediction model building unit 101 collects the original drilling data to form the rotary steering data consisting of the downhole drilling tools and the well site data, then cleans and enhances the rotary steering, analyzes and builds the model data set through the feature engineering method, and obtains a mature build rate prediction model after model training and optimization;

[0098] The build rate prediction model calling unit 102 reads the drilling data at a certain moment, processes the data into a data format that conforms to the build rate prediction model input, and then imports the data into the model, and the model calculates and outputs the predicted build rate.

[0099] 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.

[0100] 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.

[0101] 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.

[0102] 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.

[0103] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a grid device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0104] 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 method for cleaning downhole drilling deflection rate data, characterized in that: The data cleaning method comprises the following steps: Pre-cleaning, cleaning the acquired instruction data and the rotary steering data related to the instruction data mapping to remove erroneous instruction data and the rotary steering data related to the instruction data mapping; Cleaning again, using a sliding window to clean the rotational steering data for outliers, wherein the outlier cleaning uses a Z-score method to process outliers; The Makima interpolation method is used to enhance the data of the rotary steering data after cleaning, and the processed instruction data and rotary steering data are formed into a data set.

2. The method for cleaning downhole drilling deflection rate data according to claim 1, characterized in that: The rotary steering data include: well deviation near the drill bit, magnitude of resultant horizontal force, magnitude of resultant vertical force, drilling time, drilling pressure, torque, rotation speed, and pump stroke rate.

3. The downhole drilling deflection rate data cleaning method according to claim 1, characterized in that: The step of using the Makima interpolation method to perform data enhancement on the cleaned rotary steerable data comprises: calculating the slope between data points of the same type of the rotationally steerable data; Using quadratic extrapolation to add virtual points at both ends of the sequence of data points; When the slope at a data point changes suddenly, the Makima interpolation method is used to adjust the slope; constructing a cubic polynomial using the adjusted slope, weighting the slope based on the cubic polynomial so that the data points are continuous and smoothly connected; For a query point within a known data point range, a data point interval corresponding to the query point is found, and an enhancement result is obtained by calculating a cubic polynomial of the interval.

4. The method for cleaning downhole drilling buildup rate data according to claim 1, characterized in that: The step of using the Z-score method to process outliers includes: Calculate the mean μ and standard deviation σ of all data in the sliding window, and calculate Z = (x-μ)σ -1 ; If |Z|>threshold, the data x is judged as an outlier.

5. The method for cleaning downhole drilling inclination rate data according to claim 3, characterized in that: Discrete data points (x i ,y i ), the slope calculation formula between the data points is: δ i =(y i+1 -y i ) / (x i+1 -x i ), where δ i For the data point (x i , x i+1 ) between .

6. The downhole drilling deflection rate data cleaning method according to claim 3, characterized in that: The calculation method of the quadratic extrapolation method is: Edge points x1 and x n Required slope δ -1 , δ0 and δ n , δ n-1 , using the quadratic extrapolation method to calculate it as δ0 = 2δ1-δ2, δ -1 =2δ0-δ1 and δ n =2δ n-1 -δ n-2 , δ n+1 =2δ n -δ n-1 .

7. The method for cleaning downhole drilling inclination rate data according to claim 3, characterized in that: The method for weighting the slope based on the cubic polynomial comprises: Slope δ i-1 and δ i The weighted value calculation formula between is: Where w1=|δ i+1 -δ i |+(|δ i+1 +δ i |)2 -1 , w2=|δ i-1 -δ i-2 |+(|δ i-1 +δ i-2 |)2 -1 .

8. A method for predicting the build-up rate of downhole drilling, characterized in that: The data cleaning method described in any one of claims 1 to 7 is used for predicting the build-up rate of downhole drilling.

9. The method for predicting downhole drilling build-up rate according to claim 8, characterized in that: The prediction method comprises the following steps: Construct a slope prediction model; Reading real-time rotary steering data, arranging the rotary steering data into an input format and then importing it into the build rate prediction model; The build rate model prediction model calculates and outputs the predicted build rate.

10. The method for predicting downhole drilling build-up rate according to claim 8, characterized in that: The method for constructing a build rate prediction model comprises the following steps: Acquire historical drilling data, sort and calculate the historical drilling data to obtain rotary steering data, map the instruction data and the corresponding rotary steering data to form a rotary steering operation database; Cleaning and enhancing the data in the rotary steerable operation database; Based on the cleaned and enhanced telescopic rotary steerable operation database data, the Pearson correlation coefficient between various data and the slope rate is calculated, and the data with high correlation coefficients are selected to construct the training set, verification set and test set; Integrate various types of adaptive hyperparameter adjustment methods to form a model hyperparameter tuning method library, and select appropriate adaptive hyperparameter adjustment methods for different types of data for parameter adjustment; The LSTM neural network model is trained using the training set, validation set and test set data after parameter adjustment to obtain an optimized downhole drilling inclination rate prediction model.

Images