Drilling overflow risk prediction model establishment method and device, and prediction method and device

Abstract

The present specification relates to the technical field of oil and gas well drilling, and provides a drilling overflow risk prediction model establishment method and device, a prediction method and device. The method comprises: receiving drilling feature data set, formation pressure profile data set and overflow risk data set corresponding to a target well; constructing a logging while drilling sub-data set in the drilling feature data set and the formation pressure profile data set into a first training sample set and training a first neural network model; predicting and updating the formation pressure profile data set in advance according to the first neural network model; constructing a drilling sub-data set, a logging sub-data set, the updated formation pressure profile data set and the overflow risk data set in the drilling feature data set into a second training sample set to train a second neural network, and taking the trained second neural network model as a drilling overflow risk prediction model. Through the embodiments of the present specification, the problem of strong subjectivity and low accuracy of the existing drilling overflow risk prediction can be solved.

Description

Technical Field

[0001] This manual relates to the field of oil and gas well drilling technology, and in particular to a method and apparatus for establishing a drilling overflow risk prediction model, as well as a prediction method and apparatus. Background Technology

[0002] As oil and gas exploration and development continue to deepen, the focus is shifting towards unconventional oil and gas, deep formations, and ultra-deep formations. However, with the increasing burial depth of oil and gas structures, the exploration and development process is often characterized by high temperatures, complex formation pressure systems, and narrow safe drilling fluid density windows. If drilling techniques are not properly implemented, complex downhole situations such as overflows and gas intrusion can easily occur, causing significant losses. Therefore, it is necessary to accurately calculate and predict formation pressure in order to rationally design drilling fluid density and predict overflow risks during the drilling process.

[0003] However, traditional formation pressure prediction is based on empirical formulas, which suffer from low accuracy and poor reliability under complex formation conditions. Furthermore, the empirical parameters in these formulas are often manually assigned, leading to strong subjectivity, limited applicability, large prediction errors, and potentially inappropriate drilling fluid density design, significantly increasing the risks of blowouts and lost circulation. Traditional blowout risk prediction relies on real-time monitoring of logging parameters using integrated logging instruments to diagnose complex blowouts. However, threshold setting depends on the professional experience of technicians, is labor-intensive, and highly subjective, resulting in limited accuracy and issues of missed or false alarms. Although artificial intelligence technology has been widely applied in oil and gas exploration and development for operational condition diagnosis and parameter optimization in recent years, its stability and accuracy remain lacking due to the absence of constraints from geological mechanisms and traditional empirical knowledge. Therefore, an intelligent method for predicting drilling blowout risk is urgently needed to address the problems of high subjectivity and low accuracy in existing drilling blowout risk prediction methods. Summary of the Invention

[0004] Given that current drilling overflow risk prediction is highly subjective, has low accuracy, and suffers from missed alarms and false alarms, this solution is proposed to overcome or at least partially solve these problems.

[0005] On the one hand, the purpose of some embodiments of this specification is to provide a method for establishing a drilling overflow risk prediction model, the method comprising:

[0006] Receive the drilling characteristic dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well;

[0007] The logging-while-drilling sub-dataset from the drilling feature dataset and the formation pressure profile dataset are used to construct the first training sample set;

[0008] The first neural network is trained using the first training sample set to obtain the first neural network model;

[0009] The formation pressure profile dataset is predicted and updated in advance based on the first neural network model;

[0010] The drilling sub-dataset, logging sub-dataset, advanced updated formation pressure profile dataset, and overflow risk dataset in the drilling feature dataset are used to construct a second training sample set;

[0011] The second neural network is trained using the second training sample set, and the trained second neural network model is used as a drilling overflow risk prediction model.

[0012] Furthermore, after receiving the formation pressure profile dataset corresponding to the target well, the process further includes:

[0013] The drilling feature dataset and the formation pressure profile dataset are used as third training samples to construct a third training sample set.

[0014] The third neural network is trained using the third training sample set to obtain the third neural network model;

[0015] The formation pressure profile dataset is updated in real time based on the third neural network model.

[0016] Furthermore, the step of training the third neural network using the third training sample set to obtain the third neural network model further includes:

[0017] The drilling feature dataset from the third training sample set is input into the first sub-neural network to obtain the drilling time series feature dataset corresponding to the drilling feature dataset.

[0018] The drilling time-series feature dataset is input into the second sub-neural network to obtain the training results;

[0019] The accuracy was calculated based on the training results and the formation pressure profile dataset in the third training sample set.

[0020] If the accuracy is less than the first threshold, then the third neural network is retrained;

[0021] If the accuracy is not less than the first threshold, then the training of the third neural network is completed, and the third neural network model is obtained.

[0022] Furthermore, the first sub-neural network is a sequential neural network; the second sub-neural network is a feedforward neural network.

[0023] Furthermore, after constructing the first training sample set, the process further includes:

[0024] The correlation between each first sub-feature dataset of the logging-while-drilling sub-dataset in the first training sample set and the formation pressure profile dataset is calculated using the following formula:

[0025]

[0026] Among them, R 2 (X i ,Y p X represents the correlation between the i-th type of first sub-feature dataset and the formation pressure profile dataset. i For the i-th type of first sub-feature dataset, Y p For formation pressure profile dataset, V 2 (X i ,Y p V represents the distance covariance between the i-th type of first sub-feature dataset and the formation pressure profile dataset. 2 (X i V represents the distance variance of the i-th type of first sub-feature dataset. 2 (Y p ) represents the distance variance of the formation pressure profile dataset;

[0027] Filter the first sub-feature dataset whose relevance is greater than the second threshold;

[0028] The logging-while-drilling sub-dataset is updated using the filtered first sub-feature dataset;

[0029] After constructing the second training sample set, the following is further included:

[0030] The correlation between each second sub-feature dataset of the input dataset in the second training sample set and the overflow risk dataset is calculated according to the following formula:

[0031]

[0032] Where, r 2 (X j ,Y q X represents the correlation between the j-th type of second sub-feature dataset and the overflow risk dataset. j For the j-th type of second sub-feature dataset, Y q For the overflow risk dataset, V 2 (X j ,Y q Let V be the distance covariance between the j-th type of second sub-feature dataset and the overflow risk dataset. 2 (X j V represents the distance variance of the j-th type of second sub-feature dataset. 2 (Y q) represents the distance variance of the overflow risk dataset; wherein, the input dataset includes the drilling sub-dataset, the logging sub-dataset, and the advanced updated formation pressure profile dataset;

[0033] Filter the second sub-feature dataset whose relevance is greater than the third threshold;

[0034] The input dataset is updated using the filtered second sub-feature dataset.

[0035] Furthermore, the first neural network includes an encoder and a decoder:

[0036] The encoder includes a first attention unit and a first sequential neural network unit. The first attention unit is used to obtain the input parameter attention between each logging data during drilling based on each logging data during drilling in the logging data during drilling sub-data set and the input relationship between each logging data during drilling. The first sequential neural network unit is used to update the hidden state of the encoder in real time based on the input parameter attention between each logging data during drilling and the hidden state of the encoder at the previous moment.

[0037] The decoder includes a second attention unit, a second sequential neural network unit, and a nonlinear fitting unit. The second attention unit is used to determine the formation depth attention of the encoder's hidden state based on the real-time updated encoder's hidden state, and to determine the context vector based on the formation depth attention of the encoder's hidden state. The second sequential neural network unit is used to update the decoder's hidden state in real time based on the context vector and the decoder's hidden state at the previous time step. The nonlinear fitting unit is used to fit and predict the formation pressure based on the updated decoder's hidden state and the context vector at the corresponding time step.

[0038] Furthermore, the first attention unit and / or the second attention unit are constructed using the following formula:

[0039]

[0040]

[0041] in, h represents the attention score. t Let h represent the first state matrix. t T Let W represent the transpose of the first-state matrix, and let W represent the learnable matrix parameters. Let α represent the second state matrix. ts s' represents the attention weights, S represents the total number of states in the second state matrix, and s' represents the ordinal number of the second state matrix. Let s' represent the second state matrix.

[0042] On the other hand, some embodiments of this specification also provide a drilling overflow risk prediction model establishment apparatus, the apparatus comprising:

[0043] The first receiving module is used to receive the drilling feature dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well.

[0044] The first construction module is used to construct the first training sample set by combining the logging-while-drilling sub-dataset in the drilling feature dataset and the formation pressure profile dataset.

[0045] The first training module is used to train the first neural network using the first training sample set to obtain the first neural network model.

[0046] The first prediction module is used to predict and update the formation pressure profile dataset in advance based on the first neural network model.

[0047] The second construction module is used to construct a second training sample set from the drilling sub-dataset, logging sub-dataset, advanced updated formation pressure profile dataset and the overflow risk dataset in the drilling feature dataset.

[0048] The second training module is used to train the second neural network using the second training sample set, and the trained second neural network model is used as a drilling overflow risk prediction model.

[0049] Based on the same inventive concept, this embodiment also provides a method for predicting drilling overflow risk, the method comprising:

[0050] Receive the drilling feature dataset of the well to be predicted;

[0051] The drilling feature dataset of the well to be predicted is input into the drilling overflow risk prediction model trained using the method described in any of the preceding embodiments, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

[0052] On the other hand, some embodiments of this specification also provide a drilling overflow risk prediction device, the device comprising:

[0053] The second receiving module is used to receive the drilling feature dataset of the well to be predicted;

[0054] The second prediction module is used to input the drilling feature dataset of the well to be predicted into the drilling overflow risk prediction model trained using the method described in any of the preceding embodiments, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

[0055] On the other hand, some embodiments of this specification also provide a computer device, including a memory, a processor, and a computer program stored in the memory, which, when run by the processor, executes instructions for the methods described above.

[0056] On the other hand, some embodiments of this specification also provide a computer storage medium having a computer program stored thereon, which, when run by the processor of a computer device, executes instructions for the methods described above.

[0057] On the other hand, some embodiments of this specification also provide a computer program product, which includes a computer program that, when run by the processor of a computer device, executes instructions for the methods described above.

[0058] Some embodiments of this specification provide one or more technical solutions, which have at least the following technical effects:

[0059] The embodiments of this specification first construct a first training sample set using the logging-while-drilling sub-dataset and the formation pressure profile dataset, and train a first neural network model for advance updating of formation pressure profile data, thereby being able to predict the profile pressure data of the formation that the target well will soon drill into. Then, based on the profile pressure data of the formation that will soon be drilled, drilling data, logging data and overflow risk data, a drilling overflow risk prediction model is trained, thereby improving the completeness of the training data and the accuracy of drilling overflow risk prediction.

[0060] The above description is merely an overview of some embodiments of the technical solutions in this specification. In order to better understand the technical means of some embodiments of this specification and to implement them in accordance with the content of the specification, and to make the above and other objects, features and advantages of some embodiments of this specification more apparent and understandable, specific implementation methods of some embodiments of this specification are given below. Attached Figure Description

[0061] To more clearly illustrate some embodiments or technical solutions in the prior art of this specification, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort. In the drawings:

[0062] Figure 1 A schematic diagram of the implementation system of the drilling overflow risk prediction model establishment method in some embodiments of this specification is shown;

[0063] Figure 2A flowchart of a method for establishing a drilling overflow risk prediction model is shown in some embodiments of this specification;

[0064] Figure 3 This is a schematic diagram illustrating the steps of updating the formation pressure profile dataset using a third neural network in some embodiments of this specification;

[0065] Figure 4 This is a schematic diagram illustrating the steps of training a third neural network in some embodiments of this specification;

[0066] Figure 5 This is a schematic diagram of the structure of the first neural network in some embodiments of this specification;

[0067] Figure 6 This specification shows schematic diagrams of the structure of LSTM neural networks in some embodiments;

[0068] Figure 7 A flowchart of a drilling overflow risk prediction method is shown in some embodiments of this specification;

[0069] Figure 8 This is a schematic diagram of a device for establishing a drilling overflow risk prediction model in some embodiments of this specification;

[0070] Figure 9 This is a schematic diagram of a drilling overflow risk prediction device in some embodiments of this specification;

[0071] Figure 10 This is a schematic diagram of the computer device structure provided in some embodiments of this specification.

[0072] [Explanation of Labels in the Attached Image]

[0073] 101. Terminal;

[0074] 102. Server;

[0075] 801. First receiving module;

[0076] 802. First building block;

[0077] 803. First Training Module;

[0078] 804. First Prediction Module;

[0079] 805. Second building block;

[0080] 806. Second Training Module;

[0081] 901. Second receiving module;

[0082] 902. Second prediction module;

[0083] 1002. Computer equipment;

[0084] 1004, Processor;

[0085] 1006. Memory;

[0086] 1008. Drive mechanism;

[0087] 1010. Input / output interface;

[0088] 1012. Input devices;

[0089] 1014. Output devices;

[0090] 1016. Presentation device;

[0091] 1018. Graphical User Interface;

[0092] 1020. Network interface;

[0093] 1022. Communication link;

[0094] 1024. Communication bus. Detailed Implementation

[0095] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in some embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on some embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0096] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings herein are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices. It should be noted that the acquisition, storage, use, and processing of data in the technical solutions of this application comply with the relevant provisions of national laws and regulations.

[0097] like Figure 1The diagram illustrates an implementation system for a drilling overflow risk prediction model establishment method according to an embodiment of the present invention. The system may include a terminal 101 and a server 102. The terminal 101 and server 102 communicate via a network, which may include a Local Area Network (LAN), a Wide Area Network (WAN), the Internet, or a combination thereof, and is connected to a website, user equipment (e.g., a computing device), and a backend system. Operators can send drilling feature datasets, formation pressure profile datasets, and overflow risk datasets corresponding to the target well to the server 102 via the terminal 101. Upon receiving these datasets, the server 102 constructs training samples using the data in the datasets and performs calculations based on a preset neural network to obtain a trained neural network model. The trained neural network model is then sent back to the terminal 101, enabling operators to predict drilling overflow risks based on the trained neural network model.

[0098] In the embodiments of this specification, the server 102 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0099] In an optional embodiment, terminal 101 may be an electronic device, including but not limited to desktop computers, tablets, laptops, etc. Optionally, the operating system running on the electronic device may include, but is not limited to, Android, iOS, Linux, Windows, etc. Of course, terminal 101 is not limited to the aforementioned physical electronic devices; it may also be software running on the aforementioned electronic devices.

[0100] In addition, it should be noted that, Figure 1 The example shown is merely one application environment provided by this disclosure. In practical applications, it may include multiple terminals 101, and this specification does not impose any restrictions.

[0101] Figure 2This is a flowchart illustrating a method for establishing a drilling overflow risk prediction model according to an embodiment of the present invention. This specification provides the operational steps of the method described in the embodiments or flowchart, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or accompanying drawings can be executed sequentially or in parallel. Specifically, as shown... Figure 2 As shown, the method, executed by the aforementioned server side, may include:

[0102] S201: Receive the drilling characteristic dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well;

[0103] S202: Construct the first training sample set by combining the logging-while-drilling sub-dataset in the drilling feature dataset and the formation pressure profile dataset;

[0104] S203: Train the first neural network using the first training sample set to obtain the first neural network model;

[0105] S204: Predict and update the formation pressure profile dataset in advance based on the first neural network model;

[0106] S205: Construct a second training sample set from the drilling feature dataset, the well logging dataset, the advanced updated formation pressure profile dataset, and the overflow risk dataset.

[0107] S206: Use the second training sample set to train the second neural network, and use the trained second neural network model as a drilling overflow risk prediction model.

[0108] The embodiments of this specification first construct a first training sample set using the logging-while-drilling sub-dataset and the formation pressure profile dataset, and train a first neural network model for advance updating of formation pressure profile data, thereby being able to predict the profile pressure data of the formation that the target well will soon drill into. Then, based on the profile pressure data of the formation that will soon be drilled, drilling data, logging data and overflow risk data, a drilling overflow risk prediction model is trained, thereby improving the completeness of the training data and the accuracy of drilling overflow risk prediction.

[0109] Specifically, in some embodiments, the drilling feature dataset may include a logging-while-drilling sub-dataset, a drilling sub-dataset, and a logging sub-dataset. The drilling feature dataset may be based on adjacent wells and / or on the drilling dataset. Specifically, the logging-while-drilling sub-dataset may include wellbore, sonic transit time, natural gamma, density, compensated neutron, deep induced resistivity, medium induced resistivity, formation resistivity, etc. The drilling sub-dataset and the logging sub-dataset may include well depth, DC exponent, mechanical drilling rate, rotary table speed, drilling pressure, torque, riser pressure, inlet flow rate, outlet flow rate, inlet density, outlet density, equivalent density, inlet temperature, and outlet temperature. The datasets include inlet conductivity, outlet conductivity, hook load, pump speed, etc. Since these datasets are related to formation depth, for example, if the two types of data involve different formation depth ranges, it is necessary to align the formation depths of each type of data when constructing the various datasets. In addition, in some embodiments, there may be a problem of missing individual data points. In this case, interpolation fitting methods can be used to improve the data. In some embodiments, the second neural network can be a random forest, XGBoost algorithm, etc., which are robust to the data, that is, when there are 1-2 missing values, it can still make a good judgment.

[0110] In some embodiments, the overflow risk dataset can be obtained by analyzing the characterization parameters and variation patterns of overflow risk, specifically including:

[0111] ① First, determine the approximate time of overflow risk based on the drilling log and the increase in total pool volume;

[0112] ② Determine the exact time when the overflow risk will occur based on parameters such as riser pressure and hook load, which are time-sensitive.

[0113] ③ Verify the overflow risk based on the difference between import and export flows.

[0114] This accurately identifies the time point when overflow risk occurs. Then, based on this time point information, the overflow risk situation is processed using 0-1 one-hot encoding. The 0-1 one-hot encoding for overflow risk type data during normal drilling involves setting a 2D row vector, where one column represents normal drilling and the other represents overflow risk. When data point A represents normal drilling, the column indicating normal drilling is 1, and the column indicating overflow risk is 0. When data point B represents overflow risk, the column indicating normal drilling is 0, and the column indicating overflow risk is 1. The output overflow risk situation is encoded into a 0-1 code that can be recognized by a computer, thus obtaining the overflow risk dataset.

[0115] In some embodiments, since the amount of overflow risk data in the overflow risk dataset may be small or unevenly distributed, a hybrid sampling method combining oversampling and undersampling is needed to expand and balance the overflow risk dataset. Oversampling is used to repeatedly sample the data that has overflow risk, increasing the number of overflow risk data samples. Undersampling is used to incompletely sample the data that has not overflow risk, reducing the number of non-overflow risk data samples to balance the overflow risk sample size. The overflow risk data samples obtained from oversampling and the non-overflow risk data samples obtained from undersampling are then concatenated and merged to construct the overflow risk dataset, which is used to construct a second training sample set suitable for training the second neural network.

[0116] See attached document Figure 3 In some embodiments, after receiving the formation pressure profile dataset corresponding to the target well, the process may further include:

[0117] S301: Use the drilling feature dataset and the formation pressure profile dataset as third training samples to construct a third training sample set;

[0118] S302: Train the third neural network using the third training sample set to obtain the third neural network model;

[0119] S303: Update the formation pressure profile dataset in real time according to the third neural network model.

[0120] Specifically, in some embodiments, the formation pressure profile datasets of some adjacent wells are obtained after the drilling process has been completed, while the formation pressure profile datasets of some adjacent wells are obtained in real time during the drilling process. Traditional formation pressure prediction is based on empirical formulas. This method of formation pressure prediction suffers from low prediction accuracy, low accuracy, and poor reliability under complex formation conditions. Moreover, since the empirical parameters of the empirical formulas are mostly given manually, they have strong subjectivity, resulting in weak applicability and large prediction errors. Therefore, in some embodiments, drilling feature datasets and formation pressure profile datasets of adjacent wells and wells under drilling can be used to construct training samples. The third neural network can be constructed by cascading LSTM neural networks and BP neural networks. LSTM neural networks are a type of sequential neural network, which is suitable for formation data obtained from the formation deposition and compaction process. Specifically, formation deposition can be understood as the process of sediments continuously accumulating in the vertical direction. After sediment deposition, a series of diagenetic processes occur under the influence of temperature, pressure, fluid, and overlying load. The pressure generated by the overlying strata causes the sediments to be continuously mechanically compacted, leading to the expulsion of fluids from the sediments, a decrease in porosity, and consequently, a reduction in stratum thickness and an increase in density. Stratigraphic deposition is a lengthy process dependent on the deposition time, and because strata have porosity, pressure can be transmitted. Therefore, stratum pressure is a gradual process rather than a direct abrupt change, resulting in a correlation between any stratum and its adjacent strata. Thus, a sequential neural network is needed to consider the influence of preceding and following data. The trained third neural network model is then used to calculate stratum profile pressure, and the stratum pressure profile dataset is updated in real time during drilling, thereby reducing the subjectivity of constructing training samples and improving the accuracy of stratum pressure profile calculation.

[0121] See attached document Figure 4 In some embodiments, training a third neural network using the third training sample set to obtain a third neural network model may further include:

[0122] S401: Input the drilling feature dataset from the third training sample set into the first sub-neural network to obtain the drilling time series feature dataset corresponding to the drilling feature dataset;

[0123] S402: Input the drilling time-series feature dataset into the second sub-neural network to obtain the training result;

[0124] S403: Calculate the accuracy based on the training results and the formation pressure profile dataset in the third training sample set;

[0125] S404: If the accuracy is less than the first threshold, then the third neural network is retrained;

[0126] S405: If the accuracy is not less than the first threshold, then the training of the third neural network is completed, and the third neural network model is obtained.

[0127] In some embodiments, the third neural network is trained using a third training sample set. The drilling feature dataset in the third training sample set is usually the logging-while-drilling data in the well. This is because the drilling-logging-well parameters of adjacent wells cannot fully reflect the characteristics of the well, and may even interfere with the accuracy of the formation profile pressure prediction in the well. Furthermore, the logging data of the well can only be obtained after the drilling period is over. The drilling data (e.g., mechanical rotation speed, rotary drilling speed, etc.) is not closely related to the formation profile pressure data. The third neural network includes a first sub-neural network and a second sub-neural network connected in series. The third neural network is repeatedly trained iteratively based on the third training sample set until the accuracy of the training result corresponding to the target output (formation pressure profile dataset) meets the preset threshold requirement.

[0128] In some embodiments, the first sub-neural network may be a sequential neural network; the second sub-neural network may be a feedforward neural network.

[0129] Specifically, in some embodiments, when calculating the pressure of the advanced formation profile, based on the characteristics of the formation deposition and compaction process, a sequential neural network is selected as the first sub-neural network. The first sub-neural network can extract the drilling time-series feature dataset corresponding to the drilling feature dataset. A feedforward neural network (including fully connected feedforward neural networks and convolutional neural networks) is selected as the second sub-neural network for fitting regression prediction of the data. Common feedforward neural networks include BP neural networks, etc. The feedforward neural network can be regarded as a function. Through multiple composites of simple nonlinear functions, a complex mapping from the input space to the output space is achieved. It should be noted that the second sub-neural network can be one of the feedforward neural networks, which is not limited in this paper.

[0130] Furthermore, in some embodiments, after constructing the first training sample set, the following may be further included:

[0131] The correlation between each first sub-feature dataset of the logging-while-drilling sub-dataset in the first training sample set and the formation pressure profile dataset is calculated using the following formula:

[0132]

[0133] Among them, R 2 (X i ,Y p X represents the correlation between the i-th type of first sub-feature dataset and the formation pressure profile dataset. i For the i-th type of first sub-feature dataset, Y pFor formation pressure profile dataset, V 2 (X i ,Y p V represents the distance covariance between the i-th type of first sub-feature dataset and the formation pressure profile dataset. 2 (X i V represents the distance variance of the i-th type of first sub-feature dataset. 2 (Y p ) represents the distance variance of the formation pressure profile dataset;

[0134] Filter the first sub-feature dataset whose relevance is greater than the second threshold;

[0135] The logging-while-drilling sub-dataset is updated using the filtered first sub-feature dataset;

[0136] After constructing the second training sample set, the following is further included:

[0137] The correlation between each second sub-feature dataset of the input dataset in the second training sample set and the overflow risk dataset is calculated according to the following formula:

[0138]

[0139] Where, r 2 (X j ,Y q X represents the correlation between the j-th type of second sub-feature dataset and the overflow risk dataset. j For the j-th type of second sub-feature dataset, Y q For the overflow risk dataset, V 2 (X j ,Y q Let V be the distance covariance between the j-th type of second sub-feature dataset and the overflow risk dataset. 2 (X j V represents the distance variance of the j-th type of second sub-feature dataset. 2 (Y q ) represents the distance variance of the overflow risk dataset; wherein, the input dataset includes the drilling sub-dataset, the logging sub-dataset, and the advanced updated formation pressure profile dataset;

[0140] Filter the second sub-feature dataset whose relevance is greater than the third threshold;

[0141] The input dataset is updated using the filtered second sub-feature dataset.

[0142] In some embodiments, when constructing the first training sample set and / or the second training sample set, the correlation between the various types of input data and the target output is different. When the correlation of a type of book is less than the preset second threshold or third threshold, such data can be removed from the input data set, thereby reducing the time required for model training, improving training efficiency, reducing the complexity of model training, and eliminating the interference of low-correlation data types on the neural network training process.

[0143] In some embodiments, the first neural network may include an encoder and a decoder:

[0144] The encoder includes a first attention unit and a first sequential neural network unit. The first attention unit is used to obtain the input parameter attention between each logging data during drilling based on each logging data during drilling in the logging data during drilling sub-data set and the input relationship between each logging data during drilling. The first sequential neural network unit is used to update the hidden state of the encoder in real time based on the input parameter attention between each logging data during drilling and the hidden state of the encoder at the previous moment.

[0145] The decoder includes a second attention unit, a second sequential neural network unit, and a nonlinear fitting unit. The second attention unit is used to determine the formation depth attention of the encoder's hidden state based on the real-time updated encoder's hidden state, and to determine the context vector based on the formation depth attention of the encoder's hidden state. The second sequential neural network unit is used to update the decoder's hidden state in real time based on the context vector and the decoder's hidden state at the previous time step. The nonlinear fitting unit is used to fit and predict the formation pressure based on the updated decoder's hidden state and the context vector at the corresponding time step.

[0146] It can be understood that, in some embodiments, reference is made to the appendix. Figure 5 The first neural network is built using an encoder and decoder structure. The encoder includes a first attention unit and a first sequential neural network unit. The first attention unit is used to calculate the corresponding input parameter attention for various types of logging-while-drilling data. The decoder includes a second attention unit, a second sequential neural network unit, and a nonlinear fitting unit. The second attention unit is used to determine the formation depth attention of the encoder's hidden state based on the real-time updated encoder's hidden state, and to determine the context vector based on the formation depth attention of the encoder's hidden state. It can be understood that the first attention unit and the second attention unit are both weight parameter allocation mechanisms, with the goal of assisting the model in capturing important information (input logging-while-drilling data and formation depth), thereby obtaining more accurate prediction results of the pressure of the formation profile to be drilled.

[0147] In some embodiments, a first attention unit and / or a second attention unit can be constructed using the following formula:

[0148]

[0149]

[0150] in, h represents the attention score. t Represents the first state matrix. Let W represent the transpose of the first-state matrix, and let W represent the learnable matrix parameters. Let α represent the second state matrix. ts s' represents the attention weights, S represents the total number of states in the second state matrix, and s' represents the ordinal number of the second state matrix. Let s' represent the second state matrix.

[0151] It can be understood that, in some embodiments, when constructing the first attention unit and / or the second attention unit, the first state matrix refers to the model input or the output of the LSTM encoder, and the second state matrix refers to the output or hidden state of the LSTM encoder and LSTM decoder. W is usually initialized as a random matrix and then continuously updated as the model is trained.

[0152] Furthermore, in some embodiments, the formula for calculating the attention weight of the input parameter can be:

[0153]

[0154]

[0155] Among them, score input (x,h) represents the input parameter attention weight score, x T Let W represent the input parameters, W represent the learnable matrix parameters, h represent the first hidden state, and α represent the input parameters, W, h, and α. d Let represent the input parameters, attention weights, D represent the total number of first hidden states, i represent the ordinal number of the first hidden state, and h represent the number of first hidden states. i This represents the i-th first hidden state.

[0156] Furthermore, in some embodiments, the formula for calculating the formation depth attention weight can be:

[0157] score depth (h,g)=h T ·W·g

[0158]

[0159] Among them, score depth(h,g) represents the formation depth attention weight score, h T Let W represent the depth attention matrix, W represent the learnable matrix parameters, g represent the second hidden state, and β represent the depth attention matrix. d Let represent the depth-based attention weight, D represent the total number of first hidden states, i represent the ordinal number of the second hidden state, and g represent the depth-based attention weight. i This represents the i-th second hidden state.

[0160] In some embodiments, reference is made to the appendix. Figure 6 As shown, x (i.e., x) t ) is the input, f (i.e. f) t ) is the forget gate, i (i.e., i) t ) is the input gate, o (i.e., O) t ) is the output gate, C (i.e., C) t Let h be the state of the memory unit and h be the output or hidden state. h and g can be calculated by LSTM neurons. Specifically, the current input and the previous hidden state can be passed through the forget gate and the input gate in turn to update the state of the memory unit. Then, based on the updated state of the memory unit and the output gate, the hidden state can be updated to obtain the hidden state matrix h and g.

[0161] Furthermore, the forget gate can calculate the information to be forgotten based on the current input and the hidden state of the previous time step. The calculation formula is as follows:

[0162] f t =σ(W f (h t-1 ,x t )+b f )

[0163] The input gate determines the new information to add based on the current input and the hidden state of the memory unit from the previous time step. The calculation formula is as follows:

[0164] i t =σ(W i (h t-1 ,x t )+b i )

[0165] New Information N t The calculation formula is:

[0166] N t =tanh(W c (h t-1 ,x t )+b c )

[0167] The memory unit updates its state based on the previous state and the new information generated. The calculation formula is as follows:

[0168] C t =f t C t-1 +i t N t

[0169] The output gate merges the input from the previous time step with the hidden state of the memory cell, and updates the hidden state using the current time step's memory cell state. The calculation formula is as follows:

[0170] O t =σ(W o (h t-1 ,x t )+b o )

[0171] The current hidden state output is calculated using the following formula:

[0172] h t =O t tanh(C t )

[0173] Among them, f t The output of the neuron in the LSTM forgetting gate is σ, where σ is the sigmoid activation function and W is the output of the neuron. f h is the weight matrix of the neurons in the LSTM forgetting gate. t-1 x is the output of the previous time step. t For the current input, b f Let i be the bias vector of the neuron in the LSTM forget gate. t W is the output of the neuron in the LSTM input gate. i Let b be the weight matrix of the neurons in the LSTM input gate. i N is the bias vector of the neuron in the input gate of the LSTM. t New information is stored in the LSTM memory cells, tanh is the activation function, and W... c Let b be the weight matrix of the neurons in the LSTM memory unit. c C is the bias vector of the neurons in the LSTM memory unit. t C represents the current state of the LSTM memory cell. t-1 This represents the state of the LSTM memory cell at the previous time step, O t W represents the output of the neuron in the LSTM output gate. o Let b be the weight matrix of the neurons in the LSTM output gate. o h is the bias vector of the neuron in the output gate of the LSTM. t Output the current hidden state.

[0174] Based on the same inventive concept, and referring to the appendix Figure 7In some embodiments, the drilling overflow risk prediction method may include:

[0175] S701: Receive the drilling feature dataset of the well to be predicted;

[0176] S702: Input the drilling feature dataset of the well to be predicted into the drilling overflow risk prediction model trained using the method described in any of the preceding embodiments, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

[0177] It can be understood that, in some embodiments, after receiving the drilling feature dataset of the well to be predicted, the drilling overflow risk prediction model constructed by training the method described in any of the aforementioned embodiments is used to obtain the drilling overflow risk prediction result. Compared with the prediction results obtained by existing drilling rig overflow risk prediction methods, the subjective interference is avoided and the richness of the dataset is improved, thereby further improving the prediction accuracy.

[0178] It should be noted that although the operation of the method of the present invention has been described in a specific order in the above embodiments and figures, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0179] Corresponding to the above-described method for establishing a drilling overflow risk prediction model, some embodiments of this specification also provide a device for establishing a drilling overflow risk prediction model, see reference. Figure 8 As shown, in some embodiments, the apparatus may include:

[0180] The first receiving module 801 is used to receive the drilling feature dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well.

[0181] The first construction module 802 is used to construct a first training sample set by combining the logging-while-drilling sub-dataset in the drilling feature dataset and the formation pressure profile dataset.

[0182] The first training module 803 is used to train the first neural network using the first training sample set to obtain the first neural network model.

[0183] The first prediction module 804 is used to predict and update the formation pressure profile dataset in advance based on the first neural network model.

[0184] The second construction module 805 is used to construct a second training sample set from the drilling sub-dataset, logging sub-dataset, advanced updated formation pressure profile dataset and the overflow risk dataset in the drilling feature dataset.

[0185] The second training module 806 is used to train the second neural network using the second training sample set, and to use the trained second neural network model as a drilling overflow risk prediction model.

[0186] Based on the same inventive concept, and corresponding to the aforementioned drilling overflow risk prediction method, refer to... Figure 9 As shown, in some embodiments, the drilling overflow risk prediction device may include:

[0187] The second receiving module 901 is used to receive the drilling feature dataset of the well to be predicted;

[0188] The second prediction module 902 is used to input the drilling feature dataset of the well to be predicted into the drilling overflow risk prediction model trained using the method described in any of the preceding embodiments, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

[0189] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0190] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this specification are all information and data authorized and agreed upon by the user and fully authorized by all parties.

[0191] Embodiments of this specification also provide a computer device. For example... Figure 10As shown, in some embodiments of this specification, the computer device 1002 may include one or more processors 1004, such as one or more central processing units (CPUs) or graphics processing units (GPUs), each processing unit implementing one or more hardware threads. The computer device 1002 may also include any memory 1006 for storing information of any kind, such as code, settings, data, etc. In one specific embodiment, a computer program is stored on the memory 1006 and can run on the processor 1004. When the computer program is run by the processor 1004, it can execute instructions described in any of the above embodiments. Without limitation, for example, the memory 1006 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory may represent a fixed or removable component of the computer device 1002. In one scenario, when processor 1004 executes associated instructions stored in any memory or combination of memories, computer device 1002 can perform any operation of the associated instructions. Computer device 1002 also includes one or more drive mechanisms 1008 for interacting with any memory, such as hard disk drive mechanisms, optical disk drive mechanisms, etc.

[0192] Computer device 1002 may also include an input / output interface 1010 (I / O) for receiving various inputs (via input device 1012) and providing various outputs (via output device 1014). A specific output mechanism may include a presentation device 1016 and an associated graphical user interface 1018 (GUI). In other embodiments, the input / output interface 1010 (I / O), input device 1012, and output device 1014 may be omitted, and the device may function solely as a computer device within a network. Computer device 1002 may also include one or more network interfaces 1020 for exchanging data with other devices via one or more communication links 1022. One or more communication buses 1024 couple the components described above together.

[0193] The communication link 1022 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. The communication link 1022 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0194] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), computer-readable storage media, and computer program products according to some embodiments of this specification. 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 processor to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processor, create a mechanism 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.

[0195] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processor 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.

[0196] These computer program instructions may also be loaded onto a computer or other programmable data processor, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device 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.

[0197] In a typical configuration, a computer device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0198] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0199] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by computer equipment. As defined in this specification, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0200] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of computer program products implemented 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.

[0201] The embodiments described in this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. The embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processors connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0202] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0203] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0204] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments of this specification. 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. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0205] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A method for establishing a drilling overflow risk prediction model, characterized in that, The method includes: Receive the drilling characteristic dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well; The logging-while-drilling sub-dataset from the drilling feature dataset and the formation pressure profile dataset are used to construct the first training sample set; The first neural network is trained using the first training sample set to obtain the first neural network model; The formation pressure profile dataset is predicted and updated in advance based on the first neural network model; The drilling sub-dataset, logging sub-dataset, advanced updated formation pressure profile dataset, and overflow risk dataset in the drilling feature dataset are used to construct a second training sample set; The second neural network is trained using the second training sample set, and the trained second neural network model is used as a drilling overflow risk prediction model. The first neural network includes an encoder and a decoder: The encoder includes a first attention unit and a first sequential neural network unit. The first attention unit is used to obtain the input parameter attention between each logging data during drilling based on each logging data during drilling in the logging data during drilling sub-data set and the input relationship between each logging data during drilling. The first sequential neural network unit is used to update the hidden state of the encoder in real time based on the input parameter attention between each logging data during drilling and the hidden state of the encoder at the previous moment. The decoder includes a second attention unit, a second sequential neural network unit, and a nonlinear fitting unit. The second attention unit is used to determine the formation depth attention of the encoder's hidden state based on the real-time updated encoder's hidden state, and to determine the context vector based on the formation depth attention of the encoder's hidden state. The second sequential neural network unit is used to update the decoder's hidden state in real time based on the context vector and the decoder's hidden state at the previous time step. The nonlinear fitting unit is used to fit and predict the formation pressure based on the updated decoder's hidden state and the context vector at the corresponding time step. The first attention unit and / or the second attention unit are constructed using the following formula: in, This indicates the score for attention calculation. Represents the first state matrix. This represents the transpose of the first state matrix. Represents the learnable matrix parameters. Represents the second state matrix. Indicates attention weights, This represents the total number of states in the second state matrix. The ordinal number of the second state matrix. Indicates the first A second state matrix.

2. The method according to claim 1, characterized in that, After receiving the formation pressure profile dataset corresponding to the target well, the following further steps are included: The drilling feature dataset and the formation pressure profile dataset are used as third training samples to construct a third training sample set. The third neural network is trained using the third training sample set to obtain the third neural network model; The formation pressure profile dataset is updated in real time based on the third neural network model.

3. The method according to claim 2, characterized in that, The step of training the third neural network using the third training sample set to obtain the third neural network model further includes: The drilling feature dataset from the third training sample set is input into the first sub-neural network to obtain the drilling time series feature dataset corresponding to the drilling feature dataset. The drilling time-series feature dataset is input into the second sub-neural network to obtain the training results; The accuracy was calculated based on the training results and the formation pressure profile dataset in the third training sample set. If the accuracy is less than the first threshold, then the third neural network is retrained; If the accuracy is not less than the first threshold, then the training of the third neural network is completed, and the third neural network model is obtained.

4. The method according to claim 3, characterized in that, The first sub-neural network is a sequential neural network; the second sub-neural network is a feedforward neural network.

5. The method according to claim 1, characterized in that, After constructing the first training sample set, the following further steps are included: The correlation between each first sub-feature dataset of the logging-while-drilling sub-dataset in the first training sample set and the formation pressure profile dataset is calculated using the following formula: in, Let i be the correlation between the i-th type of first sub-feature dataset and the formation pressure profile dataset. For the i-th type of first sub-feature dataset, This is a dataset of formation pressure profiles. Let be the distance covariance between the i-th type of first sub-feature dataset and the formation pressure profile dataset. Let V be the distance variance of the first sub-feature dataset of the i-th type. This represents the distance variance of the formation pressure profile dataset; Filter the first sub-feature dataset whose relevance is greater than the second threshold; The logging-while-drilling sub-dataset is updated using the filtered first sub-feature dataset; After constructing the second training sample set, the following is further included: The correlation between each second sub-feature dataset of the input dataset in the second training sample set and the overflow risk dataset is calculated according to the following formula: in, Let the correlation between the j-th type of second sub-feature dataset and the overflow risk dataset be denoted as . For the j-th type of second sub-feature dataset, For overflow risk dataset, Let be the distance covariance between the j-th type of second sub-feature dataset and the overflow risk dataset. Let be the distance variance of the j-th type of second sub-feature dataset. The distance variance of the overflow risk dataset; wherein, the input dataset includes the drilling sub-dataset, the logging sub-dataset, and the advanced updated formation pressure profile dataset; Filter the second sub-feature dataset whose relevance is greater than the third threshold; The input dataset is updated using the filtered second sub-feature dataset.

6. A method for predicting drilling overflow risk, characterized in that, The method includes: Receive the drilling feature dataset of the well to be predicted; The drilling feature dataset of the well to be predicted is input into the drilling overflow risk prediction model trained using the method described in any one of claims 1-5, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

7. A device for establishing a drilling overflow risk prediction model, characterized in that, The device includes: The first receiving module is used to receive the drilling feature dataset, formation pressure profile dataset, and overflow risk dataset corresponding to the target well. The first construction module is used to construct the first training sample set by combining the logging-while-drilling sub-dataset in the drilling feature dataset and the formation pressure profile dataset. The first training module is used to train the first neural network using the first training sample set to obtain the first neural network model. The first prediction module is used to predict and update the formation pressure profile dataset in advance based on the first neural network model. The second construction module is used to construct a second training sample set from the drilling sub-dataset, logging sub-dataset, advanced updated formation pressure profile dataset and the overflow risk dataset in the drilling feature dataset. The second training module is used to train the second neural network using the second training sample set, and the trained second neural network model is used as a drilling overflow risk prediction model. The first neural network includes an encoder and a decoder: The encoder includes a first attention unit and a first sequential neural network unit. The first attention unit is used to obtain the input parameter attention between each logging data during drilling based on each logging data during drilling in the logging data during drilling sub-data set and the input relationship between each logging data during drilling. The first sequential neural network unit is used to update the hidden state of the encoder in real time based on the input parameter attention between each logging data during drilling and the hidden state of the encoder at the previous moment. The decoder includes a second attention unit, a second sequential neural network unit, and a nonlinear fitting unit. The second attention unit is used to determine the formation depth attention of the encoder's hidden state based on the real-time updated encoder's hidden state, and to determine the context vector based on the formation depth attention of the encoder's hidden state. The second sequential neural network unit is used to update the decoder's hidden state in real time based on the context vector and the decoder's hidden state at the previous time step. The nonlinear fitting unit is used to fit and predict the formation pressure based on the updated decoder's hidden state and the context vector at the corresponding time step. The first attention unit and / or the second attention unit are constructed using the following formula: in, This indicates the score for attention calculation. Represents the first state matrix. This represents the transpose of the first state matrix. Represents the learnable matrix parameters. Represents the second state matrix. Indicates attention weights, This represents the total number of states in the second state matrix. The ordinal number of the second state matrix. Indicates the first A second state matrix.

8. A drilling overflow risk prediction device, characterized in that, The device includes: The second receiving module is used to receive the drilling feature dataset of the well to be predicted; The second prediction module is used to input the drilling feature dataset of the well to be predicted into the drilling overflow risk prediction model trained by the method of any one of claims 1-5, so as to obtain the drilling overflow risk prediction result corresponding to the drilling feature dataset of the well to be predicted.

Images