Directional drilling data prediction method and device

Abstract

The invention provides a directional drilling data prediction method and device, and relates to the technical field of petroleum drilling engineering. The method comprises the following steps: acquiring drilling data; processing the drilling data based on a preset directional drilling direction prediction model, and predicting to obtain a final hole drift angle, an azimuth angle, a vertical depth and a geodetic coordinate of lower-section drilling; wherein the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on bit pressure and rotating speed change and a second loss function corresponding to the prediction data. The device executes the method. According to the directional drilling data prediction method and device provided by the embodiment of the invention, the drilling precision and efficiency can be improved, and then the oil and gas recovery efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of oil drilling engineering technology, specifically to a method and apparatus for predicting directional drilling data. Background Technology

[0002] Directional drilling technology, by controlling the wellbore trajectory, enables precise drilling of horizontal and extended-reach wells, significantly improving oil and gas recovery rates. However, existing directional drilling methods rely heavily on engineers' experience and geological data analysis, which introduces subjectivity and uncertainty. This method struggles to guarantee drilling accuracy and efficiency in complex formations and variable well conditions, necessitating more scientific and systematic predictive models to optimize drilling strategies.

[0003] While existing technologies have made some innovative developments in directional drilling tools, they still face common problems in directional drilling modeling methods. These include the difficulty in establishing theoretical models that accurately reflect field conditions due to the multi-factor influence on directional drilling effectiveness, such as formation, drill string, rock breaking, and vibration. Existing methods lack systematic analysis and modeling of the coupling relationships among these multiple factors, leading to significant uncertainty in the formulation of directional drilling plans. Particularly in complex well conditions, existing methods struggle to provide reliable predictions and guidance, impacting drilling efficiency and safety. Summary of the Invention

[0004] To address the problems in the prior art, embodiments of the present invention provide a method and apparatus for predicting directional drilling data, which can at least partially solve the problems existing in the prior art.

[0005] On the one hand, this invention proposes a method for predicting directional drilling data, comprising:

[0006] Obtain drilling data;

[0007] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0008] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0009] The comprehensive loss function is represented by the following expression:

[0010] ι=λ PDE ι PDE +λ data ι data

[0011]

[0012] l data =||uu new || 2

[0013] Where ι is the comprehensive loss function, λ PDE λ is the first weighting coefficient. data ι is the second weighting coefficient. PDE Let ι be the first loss function. data The second loss function is defined as h, which is the drilling pressure transfer coefficient used to characterize the drilling pressure loss from the drilling platform to the drill bit. The wellbore trajectory coordinates u = f(WOB, R) are functions of drilling pressure WOB and rotational speed R, used to characterize the actual values ​​of the directional drilling data. new These are the predicted values ​​of directional drilling data output by the model.

[0014] The directional drilling data prediction method further includes the following steps before training the deep learning network model based on the comprehensive loss function:

[0015] The model sample data is constrained based on the mechanical boundary constraints of directional drilling, and the constrained data is then input into the deep learning network model. The model sample data includes drill pressure sample data, well inclination angle sample data, azimuth angle sample data, and rotational speed sample data.

[0016] The directional drilling mechanical boundary constraints include a drill string assembly angle offset function related to the drill pressure sample data, the well inclination angle sample data, and the azimuth angle sample data, and a drill string assembly displacement function related to the drill pressure sample data, the well inclination angle sample data, the azimuth angle sample data, and the rotation speed sample data.

[0017] The constraint on the model sample data based on the mechanical boundary constraints of directional drilling includes:

[0018] The model sample data are constrained according to the mechanical boundary constraints of directional drilling expressed in the following expression:

[0019]

[0020] θ t (x,0)=0

[0021]

[0022] ω t (x,0)=0

[0023] θ(0,t)=θ(π,t)=ω(0,t)=ω(π,t)=0

[0024] Where θ(x,0) is the angular offset function of the drill string assembly, representing the angular offset of the bottom drill string assembly's force-bearing point at t=0, and ω(x,0) is the displacement function of the drill string assembly, representing the displacement of the bottom drill string assembly's force-bearing point at t=0. t (x,0) is the value of θ(x,0) at time t=0 after taking the partial derivative with respect to t, ω t (x,0) is the value of ω(x,0) at time t=0 after taking the partial derivative with respect to t, x∈(0,π) is the position of the bottom drill string assembly force point, θ(0,t) is the angular offset of the drill string assembly at the initial position t, θ(π,t) is the angular offset of the drill string assembly at the final position t, ω(0,t) is the displacement of the drill string assembly at the initial position t, θ(π,t) is the displacement of the drill string assembly at the final position t, and t is the mechanical loading time.

[0025] The method further includes, prior to constraining the model sample data based on the mechanical boundary constraints of directional drilling, the method further includes:

[0026] The dataset is dimensionality-reduced based on a pre-constructed geological static and dynamic mechanical theoretical model that reflects the directional drilling process, and the retained data is used as the model sample data.

[0027] The geological static and dynamic mechanical theoretical model is represented by the following expression:

[0028]

[0029] Where Path represents the trajectory of the wellbore coordinates, H is the vertical depth in meters, N is the north coordinate in meters, E is the east coordinate in meters, and α is the well inclination angle in degrees. 1. Azimuth angle, in degrees; κ is the rock drillability coefficient, dimensionless; WOB is the pressure on drill bit, in tons (t); R is the drill bit rotation speed, in RPM; Q is the pump displacement, in meters (m³). 3 / min; P is pump pressure, in MPa; B is amplitude, in mm; ROP is mechanical drilling speed, in m / s; F α F represents the dip angle of the strata, in degrees. d The dip direction of the strata is expressed in degrees (°); F t γ represents the formation strike in degrees, γ represents the formation resistivity in Ω·m, AC represents the sonic transit time in µs / m, and n represents the coordinate trajectory point of each wellbore.

[0030] Specifically, before performing dimensionality reduction processing on the dataset based on a pre-constructed geological static and dynamic mechanical theoretical model that reflects the directional drilling process, the directional drilling data prediction method further includes:

[0031] The initial dataset is cleaned to handle missing and outlier values, resulting in the dataset in question.

[0032] On one hand, the present invention proposes a directional drilling data prediction device, comprising:

[0033] The acquisition unit is used to acquire drilling data;

[0034] The prediction unit is used to process the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth and geodetic coordinates of the next drilling section.

[0035] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0036] In another aspect, embodiments of the present invention provide an electronic device, including: a processor, a memory, and a bus, wherein,

[0037] The processor and the memory communicate with each other via the bus;

[0038] The memory stores program instructions that can be executed by the processor, and the processor can execute the following methods by calling the program instructions:

[0039] Obtain drilling data;

[0040] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0041] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0042] This invention provides a non-transitory computer-readable storage medium, comprising:

[0043] The non-transitory computer-readable storage medium stores computer instructions that cause the computer to perform the following methods:

[0044] Obtain drilling data;

[0045] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0046] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0047] The directional drilling data prediction method and apparatus provided in this invention acquire drilling data; process the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section; wherein, the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotational speed, and a second loss function corresponding to the predicted data, which can improve drilling accuracy and efficiency, thereby improving oil and gas recovery rate. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0049] Figure 1 This is a flowchart illustrating a directional drilling data prediction method provided in an embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of the structure of a directional drilling data prediction device provided in an embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other.

[0053] Figure 1 This is a flowchart illustrating a directional drilling data prediction method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the directional drilling data prediction method provided in this embodiment of the invention includes:

[0054] Step S1: Obtain drilling data.

[0055] Step S2: Process the drilling data based on the preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth and geodetic coordinates of the next drilling section;

[0056] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0057] In step S1 above, the device acquires drilling data. The device can be a computer or similar device that performs the method. It should be noted that the data involved in this embodiment of the invention is user-authorized. The drilling data may include the following:

[0058] H is the vertical depth, in meters; α is the inclination angle, in degrees. 1. Azimuth angle, in degrees; κ is the rock drillability coefficient, dimensionless; WOB is the pressure on drill bit, in tons (t); R is the drill bit rotation speed, in RPM; Q is the pump displacement, in meters (m³). 3 / min; P is pump pressure, in MPa; ROP is mechanical drilling speed, in m / s; F α F represents the dip angle of the strata, in degrees. d The dip direction of the strata is expressed in degrees (°); F t The strike of the strata is indicated in degrees (°).

[0059] In step S2 above, the device processes the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section. The drilling data can be input into the preset directional drilling direction prediction model, and the output of the preset directional drilling direction prediction model can be used as the obtained directional drilling data. Referring to this data can improve drilling accuracy and efficiency, thereby improving oil and gas recovery rate.

[0060] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. This comprehensive loss function combines a first loss function corresponding to a physical model based on changes in drill pressure and rotational speed, and a second loss function corresponding to the predicted data. The process is described below in the order of data collection and processing, model training, and validation:

[0061] Data collection includes interfaces for static geological data, static drilling parameters, historical measurement-while-drilling data, and historical integrated logging data, and supports communication protocols such as WIS0, WISML, and LAS.

[0062] It also includes data cleaning functions required for data preprocessing, enabling the handling of missing and outlier values. For missing values, interpolation or mean methods are used to fill in the gaps; for outliers, they are removed or corrected based on threshold ranges.

[0063] The collected data types are diverse, but the data items and dimensions required for modeling vary depending on the lithology, formation type, and drilling technology. By using data dimensionality reduction and data quality optimization methods based on the modeling objectives, data standardization, core data extraction, and data dimensionality reduction are achieved, thereby improving the efficiency and accuracy of model solving.

[0064] By constructing a geological-multi-field static mechanics-dynamics theoretical model for the directional drilling process, i.e., a geological static-dynamic mechanics theoretical model, the collected data are initially screened.

[0065] The geological static and dynamic mechanical theoretical model can be represented by the following expression:

[0066]

[0067] Where Path represents the trajectory of the wellbore coordinates, H is the vertical depth in meters, N is the north coordinate in meters, E is the east coordinate in meters, and α is the well inclination angle in degrees. 1. Azimuth angle, in degrees; κ is the rock drillability coefficient, dimensionless; WOB is the pressure on drill bit, in tons (t); R is the drill bit rotation speed, in RPM; Q is the pump displacement, in meters (m³). 3 / min; P is pump pressure, in MPa; B is amplitude, in mm; ROP is mechanical drilling speed, in m / s; F α F represents the dip angle of the strata, in degrees. d The dip direction of the strata is expressed in degrees (°); F t γ represents the formation strike in degrees, γ represents the formation resistivity in Ω·m, AC represents the sonic transit time in µs / m, and n represents the coordinate trajectory point of each wellbore.

[0068] Different formations, lithologies, drilling conditions, and drilling techniques result in varying influencing variables. Taking composite drilling in soft, unpenetrated formations as an example, drill string amplitude (B), resistivity (γ), and sonic transit time (AC) are neglected in the theoretical model calculations. Specifically, experimental data observations are conducted based on the theoretical model, and the observed values ​​are used to assist manual judgment to determine the data items to be ignored, thus achieving data dimensionality reduction. Subsequent removal of such data improves the accuracy of model predictions.

[0069] The structure of the deep learning neural network-based directional drilling direction prediction model constrained by a physical model consists of an input layer, multiple hidden layers, a loss function layer, and an output layer; these are described below:

[0070] The input layer may include directional drilling mechanical boundary constraints serially linked to a feature data matrix constructed from model sample data. This feature data matrix can be a 12×1200 matrix, where each row represents one of 12 types of model input data:

[0071] H is the vertical depth, in meters; α is the inclination angle, in degrees. 1. Azimuth angle, in degrees; κ is the rock drillability coefficient, dimensionless; WOB is the pressure on drill bit, in tons (t); R is the drill bit rotation speed, in RPM; Q is the pump displacement, in meters (m³). 3 / min; P is pump pressure, in MPa; ROP is mechanical drilling speed, in m / s; F α F represents the dip angle of the strata, in degrees. d The dip direction of the strata is expressed in degrees (°); F t The strike of the strata is indicated in degrees (°).

[0072] The column represents a drilling depth of 1200 meters, and can be constructed according to the granularity per meter to obtain a matrix of 1200 columns.

[0073] By constraining the model sample data represented by the feature data matrix through directional drilling mechanical boundary constraints, it is possible to constrain the angular offset increment and displacement increment of the upper and lower boundaries of the bottom drill string assembly, which can effectively improve the convergence and computational performance of the training model.

[0074] The mechanical boundary constraints for directional drilling, wherein the constraints on the model sample data based on the mechanical boundary constraints for directional drilling include:

[0075] The model sample data are constrained according to the mechanical boundary constraints of directional drilling expressed in the following expression:

[0076]

[0077] θ t (x,0)=0

[0078]

[0079] ω t (x,0)=0

[0080] θ(0,t)=θ(π,t)=ω(0,t)=ω(π,t)=0

[0081] Where θ(x,0) is the angular offset function of the drill string assembly, representing the angular offset of the bottom drill string assembly's force-bearing point at t=0, and ω(x,0) is the displacement function of the drill string assembly, representing the displacement of the bottom drill string assembly's force-bearing point at t=0. t(x,0) is the value of θ(x,0) at time t=0 after taking the partial derivative with respect to t, ω t (x,0) is the value of ω(x,0) at time t=0 after taking the partial derivative with respect to t, x∈(0,π) is the position of the bottom drill string assembly force point, θ(0,t) is the angular offset of the drill string assembly at the initial position t, θ(π,t) is the angular offset of the drill string assembly at the final position t, ω(0,t) is the displacement of the drill string assembly at the initial position t, θ(π,t) is the displacement of the drill string assembly at the final position t, and t is the mechanical loading time.

[0082] The positional changes of the drill string assembly can be constrained by the angular offset function and the displacement function of the drill string assembly, which are reflected in the angular offset and radial displacement of the drill string assembly. The force-bearing position of the bottom drill string assembly only moves on the drill string body, i.e., on a simply supported beam. Therefore, the position x of the bottom drill string assembly force-bearing point moves within the range of (0, π). For each force-bearing position x, the influence of the force loading time on the angular offset and radial displacement of this position is considered. In summary, the mechanical boundary constraints of directional drilling can comprehensively constrain the drilling force based on time, spatial position, angular offset of the drill string assembly, displacement of the drill string assembly, and the changes of the angular offset and displacement of the drill string assembly with respect to t (reflected in the partial derivative with respect to t).

[0083] Since the drill string assembly angle offset function is related to the drill pressure sample data, well inclination angle sample data, and azimuth angle sample data, and the drill string assembly displacement function is related to the drill pressure sample data, well inclination angle sample data, azimuth angle sample data, and rotation speed sample data, all of which are data that need to be input into the model for training, the above constraints can indirectly constrain the numerical range of some of the model input data, thereby greatly reducing the amount of invalid data in model training and improving model training efficiency and model prediction accuracy.

[0084] The model sample data includes not only drill pressure sample data, well inclination angle sample data, azimuth angle sample data, and rotation speed sample data, but also drilling data as sample data, i.e., the above 12 types of data.

[0085] The number of hidden layers can be calculated based on the number of nodes in the input layer (12) and the number of nodes in the output layer (4). The specific calculation method is a mature technology in this field and will not be elaborated here.

[0086] The output layers are well inclination angle, azimuth angle, vertical depth, and geodetic coordinates. These four types of data can effectively assist drilling operations, improve drilling accuracy and efficiency, and thus increase oil and gas recovery rates.

[0087] The loss function layer mainly includes a loss function. Unlike conventional loss functions, this comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the predicted data.

[0088] The comprehensive loss function is expressed as follows:

[0089] ι=λ PDE ι PDE +λ data ι data

[0090]

[0091] l data =‖uu new || 2

[0092] Where l is the comprehensive loss function, λ PDE λ is the first weighting coefficient. data ι is the second weighting coefficient. PDE Let ι be the first loss function. data The second loss function is defined as h, which is the drilling pressure transfer coefficient used to characterize the drilling pressure loss from the drilling platform to the drill bit. The wellbore trajectory coordinates u = f(WOB, R) are functions of drilling pressure WOB and rotational speed R, used to characterize the actual values ​​of the directional drilling data. new This represents the predicted values ​​of the directional drilling data output by the model. h is a value that can be set empirically. The wellbore trajectory coordinates can be further expressed as:

[0093] u (x,y,z) =f(WOB,R)

[0094] Where (x, y, z) represent the three-dimensional spatial coordinates of the wellbore trajectory. The specific expressions for the functions of drilling pressure (WOB) and rotational speed (R) can be implemented using conventional techniques and will not be elaborated further. The above ι PDE The values ​​do not change dynamically during model training; instead, they are reflected by adjusting the corresponding first weight coefficient. PDE The influence of the above u new This will reflect the accuracy of the model's output, thereby making ι data The values ​​will change dynamically as the model trains, and the second weight coefficient will further reflect this. data Its influence.

[0095] The method of this invention is through ι PDE The first and second partial derivatives in the model comprehensively consider the influence characteristics of drilling pressure and rotation speed on the wellbore trajectory coordinate changes, effectively reflecting the dynamic influence of drilling pressure and rotation speed on the wellbore trajectory coordinate changes.

[0096] Finally, if the comparison result of the comprehensive loss function and the preset threshold is less than the preset error, the model training can be terminated and subsequent model validation can be performed.

[0097] Historical data was used to train the directional drilling direction prediction model, and the parameters were optimized using stochastic gradient descent. Cross-validation was used to evaluate the model performance and prevent overfitting. The model learning rate, the number of hidden layer nodes, and the loss function coefficients were adjusted based on the validation results.

[0098] Dynamic data from the drilling process is continuously acquired, such as measurement-while-drilling (MWD) data and logging-while-drilling (LMD) data. The real-time optimization function updates model parameters by inputting real-time data into the model, performs online training based on new data, improves the model's regional adaptability, and enables the model to self-iterate and upgrade.

[0099] The directional drilling data prediction method provided in this invention offers a novel approach to problem-solving through a hybrid modeling method driven by both physical and machine learning models. By utilizing historical and real-time data, combined with deep learning neural network technology, a more accurate directional drilling prediction model can be established. However, in the current field of directional drilling, single data-driven modeling methods lack systematic and scientific research and application. Therefore, this invention proposes a directional drilling modeling method based on physical constraints and deep learning neural network technology. This method, building upon theoretical models, leverages historical data to achieve multi-factor coupling, providing a scientific basis for improving the prediction effectiveness of directional drilling directions and rationalizing drilling strategies.

[0100] The directional drilling data prediction method provided in this invention acquires drilling data; processes the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section; wherein, the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotational speed, and a second loss function corresponding to the predicted data, which can improve drilling accuracy and efficiency, thereby improving oil and gas recovery rate.

[0101] Furthermore, the comprehensive loss function is expressed as follows:

[0102] ι=λ PDE ι PDE +λ data ι data

[0103]

[0104] ι data =||uunew || 2

[0105] Where ι is the comprehensive loss function, λ PDE λ is the first weighting coefficient. data ι is the second weighting coefficient. PDE Let ι be the first loss function. data The second loss function is defined as h, which is the drilling pressure transfer coefficient used to characterize the drilling pressure loss from the drilling platform to the drill bit. The wellbore trajectory coordinates u = f(WOB, R) are functions of drilling pressure WOB and rotational speed R, used to characterize the actual values ​​of the directional drilling data. new This refers to the predicted values ​​of directional drilling data output by the model. Refer to the above examples for further details; further elaboration is unnecessary.

[0106] Furthermore, before training the deep learning network model based on the comprehensive loss function, the directional drilling data prediction method also includes:

[0107] The model sample data is constrained based on the mechanical boundary constraints of directional drilling, and the constrained data is input into the deep learning network model. The model sample data includes drill pressure sample data, well inclination angle sample data, azimuth angle sample data, and rotation speed sample data. The above embodiments can be referred to for explanation, and will not be repeated here.

[0108] The directional drilling mechanical boundary constraints include a drill string assembly angle offset function related to the drill pressure sample data, the well inclination angle sample data, and the azimuth angle sample data, and a drill string assembly displacement function related to the drill pressure sample data, the well inclination angle sample data, the azimuth angle sample data, and the rotational speed sample data. These can be referred to the above embodiments for further explanation and will not be repeated here.

[0109] Furthermore, the constraint on the model sample data based on the mechanical boundary constraints of directional drilling includes:

[0110] The model sample data are constrained according to the mechanical boundary constraints of directional drilling expressed in the following expression:

[0111]

[0112] θ t (x,0)=0

[0113]

[0114] ω t (x,0)=0

[0115] θ(0,t)=θ(π,t)=ω(0,t)=ω(π,t)=0

[0116] Where θ(x,0) is the angular offset function of the drill string assembly, representing the angular offset of the bottom drill string assembly's force-bearing point at t=0, and ω(x,0) is the displacement function of the drill string assembly, representing the displacement of the bottom drill string assembly's force-bearing point at t=0. t (x,0) is the value of θ(x,0) at time t=0 after taking the partial derivative with respect to t, ω t (x,0) represents the partial derivative of ω(x,0) with respect to t at time t=0, x∈(0,π) represents the position of the bottom drill string assembly's stress point, θ(0,t) represents the angular offset of the drill string assembly at the initial position t, θ(π,t) represents the angular offset of the drill string assembly at the final position t, ω(0,t) represents the displacement of the drill string assembly at the initial position t, θ(π,t) represents the displacement of the drill string assembly at the final position t, and t represents the mechanical loading time. Refer to the above embodiment for further explanation; further details are omitted.

[0117] Furthermore, before constraining the model sample data based on the mechanical boundary constraints of directional drilling, the method further includes:

[0118] The dataset is dimensionality-reduced based on a pre-constructed geological static and dynamic mechanical theoretical model that reflects the directional drilling process, and the retained data is used as the model sample data. This can be referred to the above embodiment for further explanation, and will not be repeated here.

[0119] Furthermore, the geological static and dynamic mechanical theoretical model is represented by the following expression:

[0120]

[0121] Where Path represents the trajectory of the wellbore coordinates, H is the vertical depth in meters, N is the north coordinate in meters, E is the east coordinate in meters, and α is the well inclination angle in degrees. 1. Azimuth angle, in degrees; κ is the rock drillability coefficient, dimensionless; WOB is the pressure on drill bit, in tons (t); R is the drill bit rotation speed, in RPM; Q is the pump displacement, in meters (m³). 3 / min; P is pump pressure, in MPa; B is amplitude, in mm; ROP is mechanical drilling speed, in m / s; F α F represents the dip angle of the strata, in degrees. d The dip direction of the strata is expressed in degrees (°); F t The formation strike is represented by degrees (°), γ is the formation resistivity (Ω·m), AC is the sonic transit time (µs / m), and n represents the coordinate trajectory point of each wellbore. Refer to the above examples for further details.

[0122] Furthermore, before performing dimensionality reduction on the dataset based on a pre-constructed geological static and dynamic mechanical theoretical model that reflects the directional drilling process, the directional drilling data prediction method also includes:

[0123] The initial dataset is cleaned to remove missing and outlier values, resulting in the dataset in question. This can be referred to the above embodiments for further explanation and will not be repeated here.

[0124] Figure 2 This is a schematic diagram of the structure of a directional drilling data prediction device provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the directional drilling data prediction device provided in this embodiment of the invention includes an acquisition unit 201 and a prediction unit 202, wherein:

[0125] The acquisition unit 201 is used to acquire drilling data; the prediction unit 202 is used to process the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section; wherein, the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the predicted data.

[0126] Specifically, the acquisition unit 201 in the device is used to acquire drilling data; the prediction unit 202 is used to process the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section; wherein, the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotational speed, and a second loss function corresponding to the predicted data.

[0127] The directional drilling data prediction device provided in this embodiment of the invention acquires drilling data; processes the drilling data based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section; wherein, the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotational speed, and a second loss function corresponding to the predicted data, which can improve drilling accuracy and efficiency, thereby improving oil and gas recovery rate.

[0128] The embodiments of the present invention provide a directional drilling data prediction device that can be used to execute the processing flow of the above-described method embodiments. Its functions will not be repeated here, but can be referred to the detailed description of the above-described method embodiments.

[0129] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, the electronic device includes: a processor 301, a memory 302, and a bus 303;

[0130] The processor 301 and the memory 302 communicate with each other via the bus 303.

[0131] The processor 301 is used to call program instructions in the memory 302 to execute the methods provided in the above-described method embodiments, including, for example:

[0132] Obtain drilling data;

[0133] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0134] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0135] This embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer can perform the methods provided in the above-described method embodiments, such as including:

[0136] Obtain drilling data;

[0137] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0138] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0139] This embodiment provides a computer-readable storage medium storing a computer program that causes the computer to execute the methods provided in the above-described method embodiments, including, for example:

[0140] Obtain drilling data;

[0141] The drilling data is processed based on a preset directional drilling direction prediction model to predict the final well inclination angle, azimuth angle, vertical depth, and geodetic coordinates of the next drilling section.

[0142] The preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function. The comprehensive loss function is obtained by combining a first loss function corresponding to the physical model based on changes in drilling pressure and rotation speed, and a second loss function corresponding to the prediction data.

[0143] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0145] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0147] In the description of this specification, the references to terms such as "an embodiment," "a specific embodiment," "some embodiments," "for example," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0148] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are 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 within the scope of protection of the present invention.

Claims

1. A method for directional drilling data prediction, characterized in that, The method comprises the following steps: obtaining drilling data; processing the drilling data based on a preset directional drilling direction prediction model to predict a final hole inclination angle, azimuth angle, vertical depth and geodetic coordinates of a lower section drilling; wherein the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotary speed, and a second loss function corresponding to prediction data.

2. The directional drilling data prediction method of claim 1, wherein, The comprehensive loss function is represented by the following expression: i = λ PDE i PDE + λ data i data i data = ||u - u new || 2 wherein, l is the comprehensive loss function, λ PDE is a first weight coefficient, λ data is a second weight coefficient, l PDE is the first loss function, l data is the second loss function, h is a weight on bit transmission coefficient, used to represent the weight on bit loss from the drilling platform to the drill bit, wellbore trajectory coordinate value u = f(WOB, R) is a function about the weight on bit WOB and the rotary speed R, used to represent the actual value of the directional drilling data, u new is the predicted value of the directional drilling data output by the model.

3. The directional drilling data prediction method of claim 1, wherein, Before training the deep learning network model based on the comprehensive loss function, the directional drilling data prediction method further comprises the following steps: constraining model sample data based on a directional drilling mechanics boundary constraint condition, and inputting the constrained data into the deep learning network model; the model sample data comprises drilling pressure sample data, hole inclination angle sample data, azimuth angle sample data and rotary speed sample data; wherein the directional drilling mechanics boundary constraint condition comprises a drilling tool assembly angle offset function related to the drilling pressure sample data, the hole inclination angle sample data and the azimuth angle sample data, and a drilling tool assembly displacement function related to the drilling pressure sample data, the hole inclination angle sample data, the azimuth angle sample data and the rotary speed sample data.

4. The directional drilling data prediction method of claim 3, wherein, The constraining of the model sample data based on the directional drilling mechanics boundary constraint condition comprises the following steps: constraining the model sample data based on the directional drilling mechanics boundary constraint condition represented by the following expression: θ t (x,0) = 0 ω t (x,0) = 0 θ(0,t)=θ(π,t)=ω(0,t)=ω(π,t)=0 Wherein, θ(x, 0) is the drill string combination angle offset function, indicating the angle offset of the bottom drill string combination stress point at t=0 moment, ω(x, 0) is the drill string combination displacement function, indicating the displacement offset of the bottom drill string combination stress point at t=0 moment, θ t (x, 0) is the value of θ(x, 0) after t is differentiated at t=0 moment, ω t (x, 0) is the value of ω(x, 0) after t is differentiated at t=0 moment, x∈(0, π), is the position of the bottom drill string combination stress point, θ(0, t) is the drill string combination angle offset of x at the initial position t moment, θ(π, t) is the drill string combination angle offset of x at the terminal position t moment, ω(0, t) is the drill string combination displacement of x at the initial position t moment, θ(π, t) is the drill string combination displacement of x at the terminal position t moment, and t is the mechanical loading time.

5. The directional drilling data prediction method of claim 3, wherein, Before the constraining of the model sample data based on the directional drilling mechanics boundary constraint condition, the method further comprises the following steps: performing dimensionality reduction processing on a data set based on a pre-constructed geologic static and dynamic mechanics theoretical model that can reflect a directional drilling process, and taking the retained data as the model sample data.

6. The directional drilling data prediction method of claim 5, wherein, The geologic static and dynamic mechanics theoretical model is represented by the following expression: Wherein, Path represents the wellbore coordinate point track, H is the vertical depth, the unit is m; N is the north coordinate, the unit is m; E is the east coordinate, the unit is m; α is the inclination angle, the unit is °; is the azimuth angle, the unit is °; κ is the rock drillability coefficient, dimensionless; WOB is the drilling pressure, the unit is t; R is the drilling tool rotating speed, the unit is RPM; Q is the pump displacement, the unit is m 3 / min; P is the pump pressure, the unit is MPa; B is the amplitude, the unit is mm; ROP is the mechanical drilling speed, the unit is m / s, F α is the formation dip angle, the unit is °; F d is the formation dip direction, the unit is °; F t is the formation strike, the unit is °, γ is the formation resistivity, the unit is Ω·m; AC is the acoustic time difference, the unit is us / m, n represents each wellbore coordinate point.

7. The directional drilling data prediction method of claim 5, wherein, Before the dimensionality reduction processing on the data set based on the pre-constructed geologic static and dynamic mechanics theoretical model that can reflect the directional drilling process, the directional drilling data prediction method further comprises the following steps: performing data cleaning processing on an initial data set to realize data missing value and abnormal value processing, and obtaining the data set.

8. A directional drilling data prediction apparatus, characterized by, The method comprises the following steps: an obtaining unit, configured to obtain drilling data; a prediction unit, configured to process the drilling data based on a preset directional drilling direction prediction model to predict a final hole inclination angle, azimuth angle, vertical depth and geodetic coordinates of a lower section drilling; wherein the preset directional drilling direction prediction model is obtained by training a deep learning network model based on a comprehensive loss function; the comprehensive loss function is obtained by combining a first loss function corresponding to a physical model based on changes in drilling pressure and rotary speed, and a second loss function corresponding to prediction data.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7.

Images