Method and device for monitoring the risk of a stuck drill string

Abstract

This invention provides a method and apparatus for monitoring stuck pipe risks in drilling, relating to the field of deep well drilling technology. The method includes acquiring a time-series dataset; performing risk monitoring on the time-series dataset based on a preset stuck pipe risk monitoring model to obtain stuck pipe risk monitoring results. The apparatus executes the above method. The stuck pipe risk monitoring method and apparatus provided by this invention can accurately and timely monitor stuck pipe risks.

Description

Technical Field

[0001] This invention relates to the field of deep well drilling technology, specifically to a method and device for monitoring the risk of stuck drill pipe. Background Technology

[0002] Oil and gas drilling is the main means of obtaining oil and gas resources, and it is an engineering project full of uncertainties. Due to the harsh downhole conditions, complex downhole situations (overflow, leakage, stuck pipe, etc.) often occur, among which stuck pipe is a frequent occurrence during drilling operations.

[0003] Currently, stuck pipe incidents account for 40% to 50% of all complex drilling situations, severely hindering safe, economical, and efficient drilling. Analysis of field data and communication with field drilling engineers revealed that stuck pipe can occur during tripping and drilling, but the severity varies. For example, some highly deviated wells and horizontal wells frequently experience stuck pipe during tripping, but these incidents are often not severe and can be successfully resolved using anti-sticking agents and movable drill strings.

[0004] In the drilling process, once the drill string is completely stuck, the only way to free it is by explosion, requiring a new sidetracking plan to be designed. This seriously affects drilling safety and efficiency. Therefore, there is an urgent need for an intelligent early warning method to facilitate timely prediction of the risk of stuck drill, so that on-site engineers can adjust drilling parameters and process measures in a timely manner to avoid the risk. Summary of the Invention

[0005] To address the problems in the prior art, embodiments of the present invention provide a method and apparatus for monitoring drilling stuck pipe risks, which can at least partially solve the problems existing in the prior art.

[0006] On the one hand, this invention proposes a method for monitoring the risk of stuck pipe in drilling, comprising:

[0007] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0008] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0009] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0010] The acquisition of the time series dataset includes:

[0011] Obtain the initial preset parameters and actual production data of the well; the initial preset parameters include the initial friction coefficient and drilling design parameters, and the actual production data includes the logging data, bottom hole drilling pressure and bottom hole torque;

[0012] Based on the tubing mechanical friction torque model, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are calculated to obtain the calculated values ​​of axial force and torque at the surface location;

[0013] The calculated values ​​of axial force and torque at the ground location are verified based on the logging data. The initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient. The mapping relationship between the optimized friction coefficient and the logging data, as well as the logging data and the optimized friction coefficient, are determined as the integrated data.

[0014] The calculation of the bottom hole drilling pressure, bottom hole torque, and initial preset parameters based on the tubing string mechanical friction torque model to obtain the calculated values ​​of axial force and torque at the surface location includes:

[0015] Based on the mechanical friction torque model of the tubing string, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are iteratively calculated from the initial position at the bottom of the well towards the surface to obtain the calculated values ​​of axial force and torque at the surface position.

[0016] Specifically, the step of iteratively calculating the bottom hole drilling pressure, bottom hole torque, and initial preset parameters from the initial position at the bottom of the well based on the tubing mechanical friction torque model, to obtain the calculated values ​​of axial force and torque at the surface position, includes:

[0017] The difference between the axial force calculated from two adjacent position iterations is calculated sequentially to obtain the first difference result, and the difference between the torque calculated from two adjacent position iterations is calculated sequentially to obtain the second difference result.

[0018] Compare the sum of the first difference result and the second difference result with the first preset error. If the comparison result is determined to be less than the first preset error, then the axial force calculation value and torque calculation value at this time are determined as the axial force calculation value and torque calculation value at the ground position.

[0019] The logging data includes measured values ​​of axial force and torque at the surface location; correspondingly, the step of verifying the calculated values ​​of axial force and torque at the surface location based on the logging data and the initial friction coefficient, and optimizing the initial friction coefficient based on the verification results to obtain an optimized friction coefficient, includes:

[0020] The first absolute value of the difference between the calculated value of the axial force at the ground location and the measured value of the axial force at the ground location is calculated, and the second absolute value of the difference between the calculated value of the torque at the ground location and the measured value of the torque at the ground location is calculated.

[0021] If it is determined that the sum of the first absolute value and the second absolute value is less than the second preset error, then the friction coefficient at this time is determined as the optimized friction coefficient.

[0022] The drilling sticking risk monitoring method also includes:

[0023] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is greater than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is greater than the measured value of the torque at the ground position, then the friction coefficient is reduced.

[0024] The reduced friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, the bottom hole torque and the initial preset parameters are recalculated based on the tubing mechanical friction torque model to obtain the axial force and torque calculation values ​​at the ground position, as well as subsequent steps.

[0025] The drilling sticking risk monitoring method also includes:

[0026] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is less than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is less than the measured value of the torque at the ground position, then the friction coefficient is increased.

[0027] The increased friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are recalculated based on the tubing mechanical friction torque model to obtain the axial force and torque calculation values ​​at the ground position, as well as subsequent steps.

[0028] The autoencoder includes a time-series encoder and a time-series decoder; correspondingly, the monitoring of abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data includes:

[0029] The optimized friction coefficient in the time series dataset is preprocessed to obtain the preprocessed time series dataset.

[0030] The preprocessed time series dataset is slide-segmented according to a preset window step size to obtain multiple window data;

[0031] Based on the time encoder, dimensionality reduction processing is performed on each window data to obtain a low-dimensional latent vector corresponding to each window data.

[0032] Based on the time-series decoder, each low-dimensional latent vector is restored and reconstructed to obtain reconstructed data corresponding to each window data.

[0033] Each window of data and its corresponding reconstructed data are compared. Window data whose comparison result is greater than the preset reconstruction error are identified as abnormal fluctuation data, and the logging data corresponding to the abnormal fluctuation data are identified as abnormal feature data.

[0034] The assessment of the risk of stuck pipe caused by the abnormal fluctuation data and the abnormal characteristic data includes:

[0035] Before the current moment, sort each group of abnormal fluctuation data and abnormal feature data in chronological order to obtain the target time series dataset and the preset weights set from low to high for the values ​​corresponding to each group of abnormal fluctuation data and abnormal feature data.

[0036] For each group of abnormal fluctuation data and abnormal feature data in the target time series dataset, drilling sticking risk monitoring is performed to obtain multiple drilling sticking risk probabilities;

[0037] The probability of multiple well sticking risks is optimized and calculated according to the preset weights, and the optimized calculation results are used as the well sticking risk monitoring results.

[0038] The drilling sticking risk monitoring method also includes:

[0039] The target time series dataset is subjected to predictive processing to obtain the optimized friction coefficient and the variation law of the logging data over time;

[0040] Based on the aforementioned change pattern, predict the target optimized friction coefficient and target logging data for a specified future time.

[0041] The drilling sticking risk monitoring method also includes:

[0042] If it is determined that the probability of the drilling sticking risk is lower than the preset probability threshold, then the drilling sticking risk is monitored based on the target optimized friction coefficient and the target logging data according to the preset drilling sticking risk monitoring model to obtain the drilling sticking risk probability.

[0043] Based on the preset drilling sticking risk probability range into which the drilling sticking risk probability falls, a corresponding level of drilling sticking risk warning message is generated.

[0044] On the one hand, the present invention proposes a drilling stuck pipe risk monitoring device comprising:

[0045] An acquisition unit is used to acquire a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0046] The monitoring unit is used to perform risk monitoring on the time series dataset based on a preset drilling stuck pipe risk monitoring model, and obtain drilling stuck pipe risk monitoring results;

[0047] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0048] In another aspect, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the following method:

[0049] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0050] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0051] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0052] This invention provides a computer-readable storage medium, comprising:

[0053] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the following method:

[0054] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0055] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0056] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0057] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the following method:

[0058] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0059] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0060] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0061] The present invention provides a drilling stuck risk monitoring method and apparatus, which acquires a time-series dataset. The time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficient. Risk monitoring is performed on the time-series dataset based on a preset drilling stuck risk monitoring model to obtain drilling stuck risk monitoring results. The preset drilling stuck risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to evaluate the drilling stuck risk caused by the abnormal fluctuation data and the abnormal feature data, enabling accurate and timely drilling stuck risk monitoring. Attached Figure Description

[0062] 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:

[0063] Figure 1 This is a schematic flowchart of a drilling stuck pipe risk monitoring method provided in an embodiment of the present invention.

[0064] Figure 2 This is a flowchart illustrating a drilling stuck pipe risk monitoring method provided in another embodiment of the present invention.

[0065] Figure 3 This is a flowchart illustrating a drilling stuck pipe risk monitoring method provided in another embodiment of the present invention.

[0066] Figure 4 This is a schematic diagram illustrating the change in friction coefficient during drilling depth, provided in an embodiment of the present invention.

[0067] Figure 5 This is a schematic diagram of the drilling stuck pipe risk monitoring device provided in an embodiment of the present invention.

[0068] Figure 6 A schematic diagram of the physical structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0069] 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.

[0070] Figure 1 This is a flowchart illustrating a drilling stuck pipe risk monitoring method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the drilling sticking risk monitoring method provided in this embodiment of the invention includes:

[0071] Step S1: Obtain the time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data of optimized friction coefficient.

[0072] Step S2: Perform risk monitoring on the time series dataset based on the preset drilling stuck pipe risk monitoring model to obtain drilling stuck pipe risk monitoring results;

[0073] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0074] In step S1 above, the device acquires a time-series dataset; the time-series dataset includes logging data arranged sequentially according to the chronological order of logging data acquisition time and integrated data for optimizing the friction coefficient. The device can be a computer device that executes this method. The acquisition, storage, use, and processing of data in this application's technical solution all comply with relevant regulations. The time interval between two adjacent acquisition times can be set independently according to actual conditions.

[0075] Well logging data refers to downhole and surface data recorded in real time or periodically during the drilling process, used to monitor drilling status and analyze formation characteristics. In this embodiment of the invention, to predict abnormal fluctuations in the friction coefficient, the well logging data collected includes at least the following data dimensions: drill string assembly parameters, wellbore trajectory parameters, drilling condition parameters, drilling fluid performance parameters, formation parameters, etc. The following is an illustrative explanation:

[0076] Drill string assembly parameters include data such as drill string dimensions, drill string stiffness, and drill string type.

[0077] Wellbore trajectory parameters include data such as well inclination angle, azimuth angle, and wellbore curvature.

[0078] Drilling operating parameters include data such as drilling pressure, rotational speed, displacement, and pump pressure.

[0079] Drilling fluid performance parameters include data such as density, viscosity, solid content, and fluid loss.

[0080] Formation parameters include data such as formation lithology, formation pressure, and formation stability.

[0081] The optimized friction coefficient is obtained by continuously adjusting the friction coefficient, such as... Figure 2 As shown.

[0082] The acquisition of the time series dataset includes:

[0083] Acquire initial preset parameters and actual production data for drilling; the initial preset parameters include the initial friction coefficient and drilling design parameters, and the actual production data includes the logging data, bottom hole drilling pressure, and bottom hole torque; the initial friction coefficient can be a manually set friction coefficient, and the drilling design parameters can include the wellbore trajectory, drill string assembly, and wellbore structure, wherein the wellbore trajectory can include well depth, well inclination angle, and azimuth angle; the drill string assembly can include drill string outer diameter, drill string inner diameter, and drill string length, etc.; the wellbore structure can include casing depth and well diameter, etc.

[0084] Based on the tubing string mechanical friction-torque model, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are calculated to obtain the calculated values ​​of axial force and torque at the surface location. The tubing string mechanical friction-torque model is an important tool for studying the stress, deformation, and stability of tubing strings (such as drill strings, casing, and tubing) in oil and gas wells, geothermal wells, and other engineering projects. By establishing a reasonable mechanical model, the mechanical behavior of the tubing string under different working conditions can be predicted. Figure 2 As shown, the bottom hole drilling pressure, torque, friction coefficient, wellbore trajectory, drill string assembly, and wellbore structure can be used as input data for the tubing string mechanical friction torque model. The output data of the tubing string mechanical friction torque model are the calculated values ​​of axial force and torque at the surface location.

[0085] like Figure 3 The diagram shows the data processing flow within the tubular mechanical friction torque model. A detailed explanation follows:

[0086] (1) Start: The process begins.

[0087] (2) Bottom hole drilling pressure (WOB) and bottom hole torque (TOB): Obtain the bottom hole drilling pressure and bottom hole torque.

[0088] (3) Divide the wellbore into n units, each unit corresponding to a range of well depth values. Calculate the initial position i = n from the bottom of the well and start the calculation from the bottom of the well.

[0089] (4)F i+1 =WOB, MT i+1 =TOB: Sets the initial axial force and initial torque.

[0090] (5)F a1 =F i+1 MT a1 =MT i+1 Record the initial axial force and initial torque.

[0091] (6) Determine the buckling state of the tubing: The calculation method for contact force is different for different buckling states of the tubing.

[0092] (7) Calculate the contact force n ti : The contact force can be calculated using conventional methods.

[0093] (8) Differential equations for axial force and torque transmission: Establish the differential equations for axial force and torque transmission.

[0094] (9) Calculate F i MT i : Calculate the axial force and torque at the next iteration position.

[0095] (10)F a2 =F iMT a2 =MT i Record the axial force and torque at the position of the next iteration.

[0096] (11)|F a1 -F a2 |+|MT a2 -MT a1 |<ε: Determine whether the changes in axial force and torque are less than the first preset error.

[0097] If not, return to step (6) to continue iterative calculation; if yes, continue the next iteration of position traversal calculation until the traversal from the bottom of the well to the surface, at which point the process ends.

[0098] Due to the complex operating conditions at drilling sites, factors such as drilling rig vibration and driller's manual operation can affect the frictional torque of the drill string, which in turn is reflected in the friction coefficient between the drill string and the wellbore (casing). Figure 4 As shown, although the friction coefficient fluctuated somewhat in the 6000m-6100m well section, it generally showed a gradually increasing trend, with most values ​​between 0.1 and 0.4. Near 6100m, the friction coefficient sharply increased from around 0.35 to 0.75, a very drastic change, indicating that stuck pipe was imminent or had already occurred. In actual operation, after reviewing the well log, it was confirmed that the well experienced a stoppage torque and drill string jamming near 6100m, demonstrating that the friction coefficient has a certain characterizing effect on the trend of stuck pipe risk.

[0099] Therefore, in order to accurately predict the risk of stuck pipe during drilling operations in subsequent steps, this embodiment of the invention requires obtaining the friction coefficient as a data basis.

[0100] The process of inverting logging data using a pre-built mechanical friction and torque model of the drill string mainly includes two execution parts: one is to calculate and obtain mechanical parameters such as drill string friction, axial force, lateral force, and torque to achieve accurate characterization of drill string stress; the other is to use the relevant data information of these drill string stress characterizations to invert the friction coefficient.

[0101] The calculated values ​​of axial force and torque at the ground location are verified based on the logging data. The initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient. The mapping relationship between the optimized friction coefficient and the logging data, as well as the logging data and the optimized friction coefficient, are determined as the integrated data.

[0102] The logging data includes measured values ​​of axial force and torque at the surface location; correspondingly, the calculated values ​​of axial force and torque at the surface location are verified based on the logging data and the initial friction coefficient, and the initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient, including:

[0103] The first absolute value of the difference between the calculated value of the axial force at the ground location and the measured value of the axial force at the ground location is calculated, and the second absolute value of the difference between the calculated value of the torque at the ground location and the measured value of the torque at the ground location is calculated.

[0104] If it is determined that the sum of the first absolute value and the second absolute value is less than the second preset error, then the friction coefficient at this time is determined as the optimized friction coefficient.

[0105] The drilling sticking risk monitoring method also includes:

[0106] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated axial force at the ground location is greater than the measured axial force at the ground location, and the calculated torque at the ground location is greater than the measured torque at the ground location, then the friction coefficient is reduced. Due to the correlation between axial force and torque, when the sum of the two absolute values ​​is greater than or equal to the second preset error, there will be no situation where the calculated axial force at the ground location is greater than the measured axial force at the ground location, and the calculated torque at the ground location is less than the measured torque at the ground location, nor will there be a situation where the calculated axial force at the ground location is less than the measured axial force at the ground location, and the calculated torque at the ground location is greater than the measured torque at the ground location. The friction coefficient can be reduced using a bisection method. The standardized numerical range of the friction coefficient is 0-1. For example, if the original friction coefficient is 0.6, then the reduced friction coefficient will be 0.3.

[0107] The reduced friction coefficient is used to replace the original friction coefficient. The friction-torque model based on tubing mechanics is re-executed to calculate the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters, obtaining the calculated values ​​of axial force and torque at the surface location, as well as subsequent steps. The original friction coefficient calculated in the first iteration is the initial friction coefficient, and the original friction coefficients thereafter are the friction coefficients obtained in the previous iteration.

[0108] The drilling sticking risk monitoring method also includes:

[0109] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is less than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is less than the measured value of the torque at the ground position, then the friction coefficient is increased.

[0110] The increased friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, bottom hole torque, and initial preset parameters are recalculated using the tubing string mechanics friction-torque model to obtain the calculated axial force and torque values ​​at the surface location, as well as subsequent steps. Referring to the example above, if the original friction coefficient is 0.6, the increased friction coefficient will be 0.8.

[0111] The above execution logic process is as follows: Figure 2 As shown.

[0112] In step S2 above, the device performs risk monitoring on the time series dataset based on a preset drilling stuck risk monitoring model to obtain drilling stuck risk monitoring results; the time series dataset can be input into the preset drilling stuck risk monitoring model, and the output of the preset drilling stuck risk monitoring model can be used as the drilling stuck risk monitoring results.

[0113] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0114] An autoencoder is an unsupervised neural network consisting of a temporal encoder and a temporal decoder. It is adapted to the time-series datasets to be processed in the embodiments of this invention. Both the temporal encoder and the temporal decoder are temporal networks. Since stuck drill pipe is a small-sample and temporal problem, the optimal model is obtained by training the autoencoder using a large number of normal, risk-free samples (friction coefficients and logging data).

[0115] The autoencoder includes a time-series encoder and a time-series decoder; correspondingly, the monitoring of abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data includes:

[0116] The optimized friction coefficients in the time series dataset are preprocessed to obtain a preprocessed time series dataset. Each time series dataset contains optimized friction coefficients and logging data. In this embodiment, the optimized friction coefficients are first normalized to scale the data to between 0 and 1, eliminating the influence of dimensions, thus obtaining the processed time series dataset.

[0117] The preprocessed time series dataset is slide-segmented according to a preset window step size to obtain multiple window data; for example, continuous time series data is divided into time windows of fixed length (such as 1 minute or 5 minutes), and each window is used as a sample.

[0118] Based on the time encoder, dimensionality reduction processing is performed on each window data to obtain a low-dimensional latent vector corresponding to each window data.

[0119] Based on the time-series decoder, each low-dimensional latent vector is restored and reconstructed to obtain reconstructed data corresponding to each window data.

[0120] Each window of data is compared with its corresponding reconstructed data. Window data with a comparison result greater than a preset reconstruction error are identified as abnormal fluctuation data, and the logging data corresponding to the abnormal fluctuation data are identified as abnormal feature data. During training, the autoencoder only encounters normal data and learns its inherent distribution patterns. For abnormal data (such as abnormal fluctuations in friction coefficient), the autoencoder has not seen this pattern before, and its reconstruction error is significantly higher than that of normal data. Therefore, based on the reconstruction error exceeding a certain threshold, abnormal data points are detected.

[0121] The core idea of ​​using the reconstruction error of an autoencoder to determine outlier data points is that the autoencoder can reconstruct normal data with low error by learning the feature distribution of normal data, while the reconstruction error of outlier data increases significantly due to feature pattern mismatch. Therefore, outlier points can be identified by setting a reconstruction error threshold.

[0122] For example, in the monitoring scenario of optimizing friction coefficient and logging data in the embodiments of the present invention, for normal data, since the autoencoder has learned the characteristics of normal data, its reconstruction error is small. However, for abnormal data, since the feature pattern is much different from that of normal data, the autoencoder cannot accurately reconstruct the data, resulting in a significant increase in reconstruction error.

[0123] As can be seen, reconstruction error is an indicator that measures the difference between input data and reconstructed data. For example, in optimizing friction coefficient prediction: for normal samples, the optimized friction coefficient changes according to normal patterns, and the autoencoder can accurately learn its features, resulting in a small reconstruction error; for abnormal samples, the optimized friction coefficient exhibits abnormal fluctuations (such as sudden increases or decreases), and the autoencoder cannot effectively reconstruct it, leading to a reconstruction error significantly higher than that of normal samples. Therefore, by setting a threshold (such as based on the distribution of reconstruction errors of normal samples), when the reconstruction error of a data point exceeds this threshold, it can be identified as an outlier.

[0124] For example, embodiments of this application provide a training process for an autoencoder, including the following:

[0125] The model is built using Python and the "LSTM()" and "Bidirectional()" functions from the Keras library. The encoder is defined as the function φ:R. m →R n The decoder is defined as the function ψ:R n →R m An autoencoder aims to learn an approximate expression of an identity function, as shown in the following equation:

[0126] argmin φ,ψ ||X-ψ(φ(X))|| 2

[0127] During the encoding phase, the autoencoder uses the function shown below to encode the input vector X∈R. m Mapped to intermediate vector V∈R n And m>n.

[0128] V = σ(WX + b)

[0129] Where W∈R n×m It is a weight matrix, b∈R n Here, is the bias vector, and σ is the activation function of the coding layer. During the decoding phase, the autoencoder converts the intermediate vector V∈R using the following two formulas. n Decoded into an output vector And m > n. Calculate the reconstructed output vector. The reconstruction error between the original input vector X and the original input vector X.

[0130]

[0131]

[0132] in, It is a weight matrix. It is a bias vector. It is the activation function of the coding layer.

[0133] Building upon this, the autoencoder improves the accuracy of data reconstruction by minimizing the reconstruction error. This is achieved by using optimization algorithms such as gradient descent to learn an appropriate weight matrix W. and the bias vector b and Make Minimize. Through the above steps, the construction and training of the autoencoder are complete.

[0134] The assessment of the risk of stuck pipe caused by the abnormal fluctuation data and the abnormal characteristic data includes:

[0135] Before the current moment, sort each group of abnormal fluctuation data and abnormal feature data in chronological order to obtain the target time series dataset and the preset weights set from low to high for the values ​​corresponding to each group of abnormal fluctuation data and abnormal feature data.

[0136] For each group of abnormal fluctuation data and abnormal feature data in the target time series dataset, drilling sticking risk monitoring is performed to obtain multiple drilling sticking risk probabilities;

[0137] The probability of multiple well sticking risks is optimized and calculated according to the preset weights, and the optimized calculation results are used as the well sticking risk monitoring results.

[0138] For training the classifier, embodiments of the present invention may employ, but are not limited to, logistic regression, random forest, and support vector machine (SVM) to train a classifier for predicting the probability of stuck pipe risk through sample labeling. When processing abnormal fluctuation data corresponding to friction coefficients and abnormal feature data corresponding to logging data, this classifier considers the temporal characteristics of the drilling process (such as the trend of friction variation over time), and uses methods such as SHAP values ​​or LIME to interpret the model prediction results, assisting in decision-making and thereby improving the accuracy of predicting the probability of stuck pipe risk.

[0139] The target time series dataset obtained in chronological order is denoted as {C1, C2, C3…}, where Cj represents the j-th group of abnormal fluctuation data and abnormal feature data. Since C3 is closest to the current time, C2 is next closest, and C1 is furthest from the current time, the preset weight value W3 corresponding to C3 is greater than the preset weight value W2 corresponding to C2, and the preset weight value W2 corresponding to C2 is greater than the preset weight value W1 corresponding to C1.

[0140] Since the target time series dataset processed in this embodiment of the invention is equivalent to processing multiple abnormal data points that are ordered sequentially in time, in terms of the drilling jam risk that needs to be predicted at the current moment, the closer an abnormal data point is to the current moment, the greater its impact on the drilling jam risk predicted at the current moment, and vice versa.

[0141] For example, if the probability of stuck pipe risk corresponding to C1 is P1, the probability of stuck pipe risk corresponding to C2 is P2, and the probability of stuck pipe risk corresponding to C3 is P3, then the optimized calculation formula is:

[0142] Drilling stuck pipe risk monitoring result = P1×W1+P2×W2+P3×W3.

[0143] The drilling sticking risk monitoring method also includes:

[0144] The target time series dataset is subjected to predictive processing to obtain the optimized friction coefficient and the variation law of the logging data over time;

[0145] Based on the aforementioned change pattern, predict the target optimized friction coefficient and target logging data for a specified future time.

[0146] Since the autoencoder monitors abnormal data points in drilling operations, it does not directly monitor future moments, but rather identifies abnormal patterns in the current data. Therefore, even if the probability of stuck pipe risk output by the model does not require triggering a risk warning, in some modified embodiments, it is possible to comprehensively consider whether the abnormal data points monitored by the autoencoder will bring a high degree of stuck pipe risk at a specified future moment.

[0147] In short, on the timeline, the probability of stuck drilling risk output by the model can be called the "current stuck drilling risk". If the "current stuck drilling risk" is relatively low (i.e., it is determined that no risk warning needs to be triggered), but as time goes on, even if the "current stuck drilling risk" is relatively small, if there is a high "hidden danger", it will inevitably have an adverse impact on drilling operations in the future.

[0148] Therefore, this embodiment of the invention is further extended to perform an operation that predicts the risk of drill bit jamming at a specified future time. This "future drill bit jamming risk" is then fed back to the present. If the risk of drill bit jamming at the specified future time is relatively high, the feedback indicates that there is a relatively high degree of "potential danger" at present. This is equivalent to predicting the degree of "potential danger" at present. In a real-world scenario, this indicates that if no warning or correction is issued now, and the situation continues, there is a high probability that drill bit jamming will occur at the specified future time. Therefore, a risk warning should be triggered at present to intervene.

[0149] For example, a specific future time can be defined as a time range relatively close to the current time. If the risk to the future specified time is high, a risk warning should be triggered at the current time. This is equivalent to timely detection of early risk signs, thereby enabling more timely avoidance of stuck pipe risks and improving the safety of drilling operations. This requires predicting the target optimized friction coefficient and target logging data corresponding to the future specified time.

[0150] The target time series dataset is input into a pre-trained LSTM unit for processing, thereby learning and optimizing the variation of friction coefficient and logging data over time, and then obtaining the target friction coefficient and target logging data corresponding to a specified time in the future.

[0151] Autoencoders excel at identifying anomalies but cannot directly predict future values. Prediction requires the model to understand the dynamic changes and trends of time-series data, while autoencoders are essentially static models. Therefore, if we want to predict the target friction coefficient and target logging data at a specified future time, this embodiment of the invention employs models that need to capture temporal correlations, such as, but not limited to, recurrent neural networks (RNNs), long short-term memory networks (LSTMs), or gated recurrent units (GRUs).

[0152] Then, based on the target friction coefficient and target logging data at the specified future time, a corresponding time series dataset is formed, with each dataset containing the target friction coefficient and target logging data at the same future time.

[0153] The drilling sticking risk monitoring method also includes:

[0154] If the probability of stuck pipe risk is determined to be lower than a preset probability threshold, then the target optimized friction coefficient and target logging data are monitored for stuck pipe risk based on the preset stuck pipe risk monitoring model to obtain the stuck pipe risk probability. The preset probability threshold can be set independently according to the actual situation. The monitoring of stuck pipe risk for the target optimized friction coefficient and target logging data can be referred to the above description of monitoring stuck pipe risk for time series datasets before the current time.

[0155] Based on the preset drilling sticking risk probability range into which the drilling sticking risk probability falls, a corresponding level of drilling sticking risk warning message is generated. The preset drilling sticking risk probability range can be set according to actual conditions, but is typically 0-20%, 20%-40%, 40%-60%, 60%-80%, or 80%-100%, corresponding to levels one, two, three, four, and five, respectively. If the drilling sticking risk probability falls within 20%-40%, a level two drilling sticking risk warning message is generated.

[0156] The drilling sticking risk monitoring method provided in this invention acquires a time-series dataset. The time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficients. Risk monitoring is performed on the time-series dataset based on a preset drilling sticking risk monitoring model to obtain drilling sticking risk monitoring results. The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficients and abnormal feature data corresponding to the logging data. The classifier is used to evaluate the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data, enabling accurate and timely drilling sticking risk monitoring.

[0157] Furthermore, obtaining the time series dataset includes:

[0158] Obtain the initial preset parameters and actual production data of the well; the initial preset parameters include the initial friction coefficient and drilling design parameters, and the actual production data includes the logging data, bottom hole drilling pressure and bottom hole torque; the above embodiments can be referred to for explanation, and will not be repeated here.

[0159] The drilling pressure, bottom hole torque, and initial preset parameters are calculated based on the tubing mechanical friction torque model to obtain the calculated values ​​of axial force and torque at the surface location; the above embodiments can be referred to for explanation, and will not be repeated here.

[0160] The calculated values ​​of axial force and torque at the surface location are verified based on the logging data. The initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient. The mapping relationship between the optimized friction coefficient and the logging data, as well as the relationship between the logging data and the optimized friction coefficient, is determined as the integrated data. This can be referred to the above embodiment for further explanation and will not be repeated here.

[0161] Furthermore, the calculation of the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters based on the tubing string mechanical friction torque model yields the calculated values ​​of the axial force and torque at the surface location, including:

[0162] Based on the aforementioned tubing mechanical friction torque model, starting from the initial position at the bottom of the well, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are iteratively calculated towards the surface to obtain the calculated values ​​of axial force and torque at the surface position. This can be referred to the above embodiment for further explanation and will not be repeated here.

[0163] Furthermore, based on the tubing string mechanical friction torque model, the bottom hole drilling pressure, bottom hole torque, and initial preset parameters are iteratively calculated from the initial position at the bottom of the well towards the surface to obtain the calculated values ​​of axial force and torque at the surface position, including:

[0164] The difference between the axial force calculated from two adjacent position iterations is calculated sequentially to obtain the first difference result, and the difference between the torque calculated from two adjacent position iterations is calculated sequentially to obtain the second difference result; the above embodiments can be referred to for explanation, and will not be repeated here.

[0165] The sum of the first difference result and the second difference result is compared with a first preset error. If the comparison result is determined to be less than the first preset error, then the calculated axial force value and torque value at this time are determined as the calculated axial force value and torque value at the ground position. This can be referred to the above embodiment for explanation, and will not be repeated here.

[0166] Further, the logging data includes measured values ​​of axial force and torque at the surface location; correspondingly, the step of verifying the calculated values ​​of axial force and torque at the surface location based on the logging data and the initial friction coefficient, and optimizing the initial friction coefficient based on the verification results to obtain an optimized friction coefficient, includes:

[0167] The first absolute value of the difference between the calculated value of the axial force at the ground location and the measured value of the axial force at the ground location, and the second absolute value of the difference between the calculated value of the torque at the ground location and the measured value of the torque at the ground location are calculated; these can be referred to the above embodiments for explanation, and will not be repeated here.

[0168] If it is determined that the sum of the first absolute value and the second absolute value is less than the second preset error, then the friction coefficient at this time is determined as the optimized friction coefficient. This can be referred to the above embodiments for further explanation, and will not be repeated here.

[0169] Furthermore, the drilling sticking risk monitoring method also includes:

[0170] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is greater than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is greater than the measured value of the torque at the ground position, then the friction coefficient is reduced; this can be referred to the above embodiment for explanation, and will not be repeated here.

[0171] The reduced friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, bottom hole torque, and initial preset parameters are recalculated using the tubing string mechanics friction torque model. This yields the calculated axial force and torque values ​​at the surface location, as well as subsequent steps. Refer to the above embodiment for further details.

[0172] Furthermore, the drilling sticking risk monitoring method also includes:

[0173] If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is less than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is less than the measured value of the torque at the ground position, then the friction coefficient is increased; this can be referred to the above embodiment for explanation, and will not be repeated here.

[0174] The increased friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, bottom hole torque, and initial preset parameters are recalculated using the tubing string mechanical friction torque model to obtain the calculated values ​​of axial force and torque at the surface location, as well as subsequent steps. Refer to the above embodiment for further details.

[0175] Furthermore, the autoencoder includes a time-series encoder and a time-series decoder; correspondingly, the monitoring of abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data includes:

[0176] The optimized friction coefficients in the time series dataset are preprocessed to obtain the preprocessed time series dataset; this can be referred to the above embodiments for explanation, and will not be repeated here.

[0177] The preprocessed time series dataset is slide-segmented according to a preset window step size to obtain multiple window data; the above embodiments can be referred to for explanation, and will not be repeated here.

[0178] The time encoder is used to perform dimensionality reduction on each window data to obtain a low-dimensional latent vector corresponding to each window data; the above embodiments can be referred to for explanation, and will not be repeated here.

[0179] The time-series decoder is used to restore and reconstruct each low-dimensional latent vector to obtain reconstructed data corresponding to each window of data; this can be referred to the above embodiments for explanation, and will not be repeated here.

[0180] Each window of data is compared with its corresponding reconstructed data. Window data whose comparison result is greater than a preset reconstruction error is identified as abnormal fluctuation data, and the logging data corresponding to the abnormal fluctuation data is identified as abnormal feature data. This can be referred to the above embodiment for further explanation, and will not be repeated here.

[0181] Furthermore, the assessment of the drilling sticking risk caused by the abnormal fluctuation data and the abnormal characteristic data includes:

[0182] Before the current moment, each group of abnormal fluctuation data and abnormal feature data is sorted in chronological order to obtain the target time series dataset and the preset weights set from low to high for the values ​​corresponding to each group of abnormal fluctuation data and abnormal feature data; the above embodiments can be referred to for explanation, and will not be repeated here.

[0183] For each group of abnormal fluctuation data and abnormal feature data in the target time series dataset, drilling sticking risk monitoring is performed to obtain multiple drilling sticking risk probabilities; the above embodiments can be referred to for explanation, and will not be repeated here.

[0184] The probabilities of multiple stuck pipe risks are optimized and calculated according to the preset weights, and the optimized calculation results are used as the stuck pipe risk monitoring results. This can be referred to the above embodiments for explanation, and will not be repeated here.

[0185] Furthermore, the drilling sticking risk monitoring method also includes:

[0186] The target time series dataset is subjected to predictive processing to obtain the optimized friction coefficient and the variation law of the logging data over time; the above embodiments can be referred to for explanation, and will not be repeated here.

[0187] Based on the described changing patterns, predict the target optimized friction coefficient and target logging data for a specified future time. Refer to the above embodiments for further details; they will not be repeated here.

[0188] Furthermore, the drilling sticking risk monitoring method also includes:

[0189] If the probability of stuck pipe in drilling is determined to be lower than a preset probability threshold, then the target optimized friction coefficient and target logging data are monitored for stuck pipe risk based on the preset stuck pipe risk monitoring model to obtain the stuck pipe risk probability; the above embodiments can be referred to for explanation, and will not be repeated here.

[0190] Based on the preset drilling sticking risk probability range into which the drilling sticking risk probability falls, a corresponding level of drilling sticking risk warning message is generated. This can be referred to the above embodiment for explanation, and will not be repeated here.

[0191] Figure 5 This is a schematic diagram of the structure of a drilling stuck pipe risk monitoring device provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the drilling stuck pipe risk monitoring device provided in this embodiment of the invention includes an acquisition unit 501 and a monitoring unit 502, wherein:

[0192] The acquisition unit 501 is used to acquire a time-series dataset; the time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficient; the monitoring unit 502 is used to perform risk monitoring on the time-series dataset based on a preset drilling sticking risk monitoring model to obtain drilling sticking risk monitoring results; wherein, the preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier, the autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data; the classifier is used to evaluate the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0193] Specifically, the acquisition unit 501 in the device is used to acquire a time-series dataset; the time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficient; the monitoring unit 502 is used to perform risk monitoring on the time-series dataset based on a preset drilling sticking risk monitoring model to obtain drilling sticking risk monitoring results; wherein, the preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier, the autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data; the classifier is used to evaluate the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0194] The drilling sticking risk monitoring device provided in this embodiment of the invention acquires a time-series dataset. The time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficient. Risk monitoring is performed on the time-series dataset based on a preset drilling sticking risk monitoring model to obtain drilling sticking risk monitoring results. The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to evaluate the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data, enabling accurate and timely drilling sticking risk monitoring.

[0195] The embodiments of the present invention provide a drilling stuck pipe risk monitoring 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.

[0196] Figure 6 This is a schematic diagram of the physical structure of a computer device provided in an embodiment of the present invention, such as... Figure 6 As shown, the computer device includes: a memory 601, a processor 602, and a computer program stored in the memory 601 and executable on the processor 602. When the processor 602 executes the computer program, it implements the following method:

[0197] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0198] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0199] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0200] This embodiment discloses a computer program product, which includes a computer program that, when executed by a processor, implements the following method:

[0201] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0202] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0203] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0204] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the following method:

[0205] Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient;

[0206] Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained.

[0207] The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data.

[0208] Compared with existing technologies, the drilling sticking risk monitoring method provided in this invention acquires a time-series dataset. The time-series dataset includes logging data arranged sequentially according to the acquisition time of the logging data and integrated data of optimized friction coefficients. Risk monitoring is performed on the time-series dataset based on a preset drilling sticking risk monitoring model to obtain drilling sticking risk monitoring results. The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder monitors abnormal fluctuation data corresponding to the optimized friction coefficients and abnormal feature data corresponding to the logging data. The classifier assesses the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data, enabling accurate and timely drilling sticking risk monitoring.

[0209] 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.

[0210] 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.

[0211] 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.

[0212] 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.

[0213] 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.

[0214] 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 monitoring the risk of stuck pipe in drilling, characterized in that, include: Obtain a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient; Based on a preset drilling sticking risk monitoring model, the time series dataset is monitored for risk, and the drilling sticking risk monitoring results are obtained. The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data. The autoencoder includes a time-series encoder and a time-series decoder; correspondingly, the monitoring of abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data includes: The optimized friction coefficient in the time series dataset is preprocessed to obtain the preprocessed time series dataset. The preprocessed time series dataset is slide-segmented according to a preset window step size to obtain multiple window data; Based on the time encoder, dimensionality reduction processing is performed on each window data to obtain a low-dimensional latent vector corresponding to each window data. Based on the time-series decoder, each low-dimensional latent vector is restored and reconstructed to obtain reconstructed data corresponding to each window data. Each window of data and its corresponding reconstructed data are compared separately. Window data whose comparison results are greater than the preset reconstruction error are identified as abnormal fluctuation data, and the logging data corresponding to the abnormal fluctuation data are identified as abnormal feature data. The assessment of the risk of stuck pipe caused by the abnormal fluctuation data and the abnormal characteristic data includes: Before the current moment, sort each group of abnormal fluctuation data and abnormal feature data in chronological order to obtain the target time series dataset and the preset weights set from low to high for the values ​​corresponding to each group of abnormal fluctuation data and abnormal feature data. For each group of abnormal fluctuation data and abnormal feature data in the target time series dataset, drilling sticking risk monitoring is performed to obtain multiple drilling sticking risk probabilities; The probability of multiple well sticking risks is optimized and calculated according to the preset weights, and the optimized calculation results are used as the well sticking risk monitoring results.

2. The drilling sticking risk monitoring method according to claim 1, characterized in that, The acquisition of the time series dataset includes: Obtain the initial preset parameters and actual production data of the well; the initial preset parameters include the initial friction coefficient and drilling design parameters, and the actual production data includes the logging data, bottom hole drilling pressure and bottom hole torque; Based on the tubing mechanical friction torque model, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are calculated to obtain the calculated values ​​of axial force and torque at the surface location; The calculated values ​​of axial force and torque at the ground location are verified based on the logging data. The initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient. The mapping relationship between the optimized friction coefficient and the logging data, as well as the logging data and the optimized friction coefficient, are determined as the integrated data.

3. The drilling sticking risk monitoring method according to claim 2, characterized in that, The calculation of the bottom hole drilling pressure, bottom hole torque, and initial preset parameters based on the tubing string mechanical friction torque model yields the calculated values ​​of axial force and torque at the surface location, including: Based on the mechanical friction torque model of the tubing string, the bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are iteratively calculated from the initial position at the bottom of the well towards the surface to obtain the calculated values ​​of axial force and torque at the surface position.

4. The drilling sticking risk monitoring method according to claim 3, characterized in that, The method involves iteratively calculating the bottom hole drilling pressure, bottom hole torque, and initial preset parameters from the initial position at the bottom of the well, based on the tubing mechanical friction torque model, to obtain the calculated values ​​of axial force and torque at the surface position, including: The difference between the axial force calculated from two adjacent position iterations is calculated sequentially to obtain the first difference result, and the difference between the torque calculated from two adjacent position iterations is calculated sequentially to obtain the second difference result. Compare the sum of the first difference result and the second difference result with the first preset error. If the comparison result is determined to be less than the first preset error, then the axial force calculation value and torque calculation value at this time are determined as the axial force calculation value and torque calculation value at the ground position.

5. The drilling sticking risk monitoring method according to claim 2, characterized in that, The logging data includes measured values ​​of axial force and torque at the surface location; correspondingly, the calculated values ​​of axial force and torque at the surface location are verified based on the logging data and the initial friction coefficient, and the initial friction coefficient is optimized based on the verification results to obtain an optimized friction coefficient, including: The first absolute value of the difference between the calculated value of the axial force at the ground location and the measured value of the axial force at the ground location is calculated, and the second absolute value of the difference between the calculated value of the torque at the ground location and the measured value of the torque at the ground location is calculated. If it is determined that the sum of the first absolute value and the second absolute value is less than the second preset error, then the friction coefficient at this time is determined as the optimized friction coefficient.

6. The drilling sticking risk monitoring method according to claim 5, characterized in that, The drilling sticking risk monitoring method also includes: If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is greater than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is greater than the measured value of the torque at the ground position, then the friction coefficient is reduced. The reduced friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, the bottom hole torque and the initial preset parameters are recalculated based on the tubing mechanical friction torque model to obtain the axial force and torque calculation values ​​at the ground position, as well as subsequent steps.

7. The drilling sticking risk monitoring method according to claim 5, characterized in that, The drilling sticking risk monitoring method also includes: If it is determined that the sum of the first absolute value and the second absolute value is greater than or equal to the second preset error, and the calculated value of the axial force at the ground position is less than the measured value of the axial force at the ground position, and the calculated value of the torque at the ground position is less than the measured value of the torque at the ground position, then the friction coefficient is increased. The increased friction coefficient is used to replace the original friction coefficient. The bottom hole drilling pressure, the bottom hole torque, and the initial preset parameters are recalculated based on the tubing mechanical friction torque model to obtain the axial force and torque values ​​at the ground position, as well as subsequent steps.

8. The drilling sticking risk monitoring method according to claim 1, characterized in that, The drilling sticking risk monitoring method also includes: The target time series dataset is subjected to predictive processing to obtain the optimized friction coefficient and the variation law of the logging data over time; Based on the aforementioned change pattern, predict the target optimized friction coefficient and target logging data for a specified future time.

9. The drilling sticking risk monitoring method according to claim 8, characterized in that, The drilling sticking risk monitoring method also includes: If it is determined that the probability of the drilling sticking risk is lower than the preset probability threshold, then the drilling sticking risk is monitored based on the target optimized friction coefficient and the target logging data according to the preset drilling sticking risk monitoring model to obtain the drilling sticking risk probability. Based on the preset drilling sticking risk probability range into which the drilling sticking risk probability falls, a corresponding level of drilling sticking risk warning message is generated.

10. A drilling stuck pipe risk monitoring device, characterized in that, include: An acquisition unit is used to acquire a time series dataset; the time series dataset includes logging data arranged in chronological order of logging data acquisition time and integrated data with optimized friction coefficient; The monitoring unit is used to perform risk monitoring on the time series dataset based on a preset drilling stuck pipe risk monitoring model, and obtain drilling stuck pipe risk monitoring results; The preset drilling sticking risk monitoring model includes a pre-trained autoencoder and a classifier. The autoencoder is used to monitor abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data. The classifier is used to assess the drilling sticking risk caused by the abnormal fluctuation data and the abnormal feature data. The autoencoder includes a time-series encoder and a time-series decoder; correspondingly, the monitoring of abnormal fluctuation data corresponding to the optimized friction coefficient and abnormal feature data corresponding to the logging data includes: The optimized friction coefficient in the time series dataset is preprocessed to obtain the preprocessed time series dataset. The preprocessed time series dataset is slide-segmented according to a preset window step size to obtain multiple window data; Based on the time encoder, dimensionality reduction processing is performed on each window data to obtain a low-dimensional latent vector corresponding to each window data. Based on the time-series decoder, each low-dimensional latent vector is restored and reconstructed to obtain reconstructed data corresponding to each window data. Each window of data and its corresponding reconstructed data are compared separately. Window data whose comparison results are greater than the preset reconstruction error are identified as abnormal fluctuation data, and the logging data corresponding to the abnormal fluctuation data are identified as abnormal feature data. The assessment of the risk of stuck pipe caused by the abnormal fluctuation data and the abnormal characteristic data includes: Before the current moment, sort each group of abnormal fluctuation data and abnormal feature data in chronological order to obtain the target time series dataset and the preset weights set from low to high for the values ​​corresponding to each group of abnormal fluctuation data and abnormal feature data. For each group of abnormal fluctuation data and abnormal feature data in the target time series dataset, drilling sticking risk monitoring is performed to obtain multiple drilling sticking risk probabilities; The probability of multiple well sticking risks is optimized and calculated according to the preset weights, and the optimized calculation results are used as the well sticking risk monitoring results.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 9.

13. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 9.

Images