Drilling accident early warning method and device

Abstract

This invention discloses a drilling accident early warning method and apparatus. The method includes: acquiring multi-source drilling data; aligning the multi-source drilling data using a precise time synchronization protocol; extracting multi-modal temporal features to obtain temporal features of the multi-source drilling data; inputting the temporal features of the multi-source drilling data into a drilling accident occurrence probability generation model to output the drilling accident occurrence probability; the TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder; the Transformer encoder employs a probabilistic sparse self-attention mechanism; and triggering a drilling accident early warning based on the multi-source drilling data, the drilling accident occurrence probability, and a pre-configured physical rule base; the physical rule base includes drilling accident judgment logic, which can reduce the drilling accident false negative rate, shorten the drilling accident early warning inference delay, and improve the cross-well migration capability and accuracy of the drilling accident early warning.

Description

Technical Field

[0001] This invention relates to the field of oil drilling engineering technology, and in particular to a drilling accident early warning method and device. Background Technology

[0002] This section is intended to provide background or context for embodiments of the present invention. The description herein is not intended to imply that it is prior art simply because it is included in this section.

[0003] In the field of oil drilling engineering technology, in order to improve the safety and efficiency of drilling operations, existing technologies typically employ methods such as expert systems and shallow machine learning to achieve early warning of drilling accidents.

[0004] However, the aforementioned existing technologies have one or more drawbacks in practical applications: First, a high false negative rate is a prominent problem. Traditional methods have a false negative rate of up to 40% for drilling accidents, resulting in a large number of accidents going undetected and posing serious safety hazards to drilling operations. Second, inference latency is also a significant bottleneck, especially when processing long-sequence data with more than 1,000 time steps. The inference latency of existing technology models exceeds 20 seconds, which cannot meet the requirements of real-time control and makes it difficult to respond quickly when accidents occur. In addition, existing technologies perform poorly in terms of cross-well migration capability. In new wells within the same block, the model accuracy drops by more than 30%, which greatly limits the effective application of early warning technology under different well conditions.

[0005] Therefore, reducing the rate of missed drilling accidents, shortening the reasoning delay for drilling accident early warning, and improving the cross-well migration capability and accuracy of drilling accident early warning are important technical challenges currently faced by those skilled in the art. Summary of the Invention

[0006] This invention provides a drilling accident early warning method to reduce the drilling accident false alarm rate, shorten the drilling accident early warning inference delay, and improve the cross-well migration capability and accuracy of drilling accident early warning. The method includes:

[0007] Acquire multi-source drilling data; multi-source drilling data includes real-time data of dynamic parameters of drilling engineering, static parameters of drilling fluid, and real-time data of dynamic geological parameters.

[0008] A precise time synchronization protocol is used to align drilling data from multiple sources;

[0009] Multimodal time-series features are extracted from the aligned drilling multi-source data to fuse static and dynamic drilling multi-source data and obtain the time-series features of the drilling multi-source data.

[0010] The time-series features of multi-source drilling data are input into a pre-trained drilling accident probability generation model, which outputs the probability of drilling accidents. The drilling accident probability generation model is generated by training a TLM (Transaction-Level Modeling) model using the time-series features of historical multi-source drilling data. The TLM model includes a Transformer (self-attention mechanism) encoder and a bidirectional long short-term memory (LSTM) decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism.

[0011] Based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base, a drilling accident warning is triggered; the physical rule base includes the judgment logic for drilling accidents.

[0012] This invention also provides a drilling accident early warning device to reduce the drilling accident false alarm rate, shorten the drilling accident early warning inference delay, and improve the cross-well migration capability and accuracy of drilling accident early warning. The device includes:

[0013] The acquisition module is used to acquire multi-source drilling data, which includes real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, and geological dynamic parameters.

[0014] The alignment module is used to align multi-source drilling data using a precise time synchronization protocol;

[0015] The feature extraction module is used to extract multimodal time-series features from the aligned drilling multi-source data, so as to fuse static and dynamic drilling multi-source data to obtain the time-series features of drilling multi-source data.

[0016] The output module is used to input the time-series features of drilling multi-source data into a pre-trained drilling accident probability generation model and output the drilling accident probability. The drilling accident probability generation model is generated by training the TLM model using the time-series features of historical drilling multi-source data. The TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism.

[0017] The early warning triggering module is used to trigger drilling accident early warnings based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base; the physical rule base includes the judgment logic for drilling accidents.

[0018] Compared with existing drilling accident early warning technologies, this invention acquires multi-source drilling data, including real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, and geological dynamic parameters. A precise time synchronization protocol is used to align the multi-source drilling data. Multi-modal time-series features are extracted from the aligned data to fuse static and dynamic data, resulting in multi-source drilling data time-series features. These features are then input into a pre-trained drilling accident probability generation model, which outputs the drilling accident probability. Therefore, the probability generation model is generated by training the TLM model using the time-series features of historical drilling multi-source data. The TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism. Based on the drilling multi-source data, the probability of drilling accidents, and the pre-configured physical rule base, drilling accident early warning is triggered. The physical rule base includes the judgment logic of drilling accidents, which can reduce the false negative rate of drilling accidents, shorten the inference delay of drilling accident early warning, and improve the cross-well migration capability and accuracy of drilling accident early warning. Attached Figure Description

[0019] 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:

[0020] Figure 1 This is a flowchart of a drilling accident early warning method provided in an embodiment of the present invention;

[0021] Figure 2 A flowchart illustrating a specific example of a drilling accident early warning method provided in this embodiment of the invention;

[0022] Figure 3 This is a schematic diagram of a drilling accident early warning device provided in an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of a specific example of a drilling accident early warning device provided in an embodiment of the present invention;

[0024] Figure 5 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0025] 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 of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0026] The acquisition, storage, use, and processing of data in this application all comply with relevant regulations.

[0027] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0028] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended terms, meaning that they include but are not limited to. The terms "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example that are included in at least one embodiment or example of this application. 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, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0029] This invention relates to the fields of intelligent oil drilling engineering and artificial intelligence, and in particular to an intelligent early warning method for complex drilling conditions and accidents based on a time-series large-scale model. This method enables real-time detection of abnormal downhole conditions during drilling, prediction of risk probabilities, and linkage with automated control, thus ensuring the safety and efficiency of drilling operations.

[0030] This invention aims to provide an intelligent early warning system for complex drilling conditions and accidents based on a large time-series model, improving the ability to capture long-range dependencies. The pre-trained model based on Transformer demonstrates the ability to capture long-range dependencies in industrial time-series prediction, and its attention mechanism can cover more than 5,000 time steps, providing a new approach to solving the problem of long-series data processing. It also enhances the generalization ability; the transfer learning framework can improve the generalization ability of the pre-trained model to more than 70% of that of new well data, effectively enhancing the model's adaptability under different well conditions.

[0031] Figure 1 This is a flowchart of a drilling accident early warning method provided in an embodiment of the present invention, such as... Figure 1 As shown, the method may include:

[0032] Step 101: Obtain drilling multi-source data; drilling multi-source data includes real-time data of drilling engineering dynamic parameters, drilling fluid static parameter data, and real-time data of geological dynamic parameters;

[0033] Step 102: Align the drilling multi-source data using a precise time synchronization protocol;

[0034] Step 103: Extract multimodal time series features from the aligned drilling multi-source data to fuse static and dynamic drilling multi-source data and obtain drilling multi-source data time series features;

[0035] Step 104: Input the time series features of drilling multi-source data into the pre-trained drilling accident probability generation model and output the drilling accident probability. The drilling accident probability generation model is generated by training the TLM model using the time series features of historical drilling multi-source data. The TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism.

[0036] Step 105: Trigger a drilling accident warning based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base; the physical rule base includes the judgment logic for drilling accidents.

[0037] In one embodiment, the real-time data of drilling dynamic parameters includes: real-time data of drilling pressure, real-time data of torque and / or real-time data of rotational speed; the data of drilling fluid static parameters includes: drilling fluid density data, drilling fluid gas content data and / or hydrogen sulfide concentration data; and the data of geological dynamic parameters includes: gamma ray data, resistivity data and / or pore pressure data.

[0038] In this embodiment, the acquired drilling multi-source data can be shown in Table 1 below:

[0039] Table 1

[0040] category parameter Engineering parameters Drilling pressure, torque, and rotational speed Drilling fluid <![CDATA[Density, gas content, H2S concentration <!-- 3 -->]]> Geological parameters Gamma rays, resistivity, pore pressure

[0041] This step ensures the collection of comprehensive and multi-dimensional drilling data, providing a reliable foundation for subsequent analysis and early warning.

[0042] In one embodiment, in step 102, the data synchronization mechanism can support sampling frequencies above 10Hz and employ a Precise Time Protocol (PTP), such as the IEEE 1588 Precise Time Protocol, to align multi-source drilling data with a time error of <1ms, ensuring data synchronization and accuracy. This step eliminates errors caused by data asynchrony, improving the reliability of subsequent feature extraction and model prediction.

[0043] In one embodiment, in step 103, multimodal temporal feature extraction is performed on the aligned drilling multi-source data to fuse static and dynamic drilling multi-source data, thereby obtaining the drilling multi-source data temporal features. Since the drilling multi-source data includes static parameter data (drilling fluid static parameter data) and dynamic parameter data (real-time data of drilling engineering dynamic parameters and real-time data of geological dynamic parameters), step 103 employs a multimodal temporal feature extraction technique capable of fusing static and real-time dynamic data. This step extracts comprehensive features reflecting complex drilling conditions, improving the model's feature representation capability.

[0044] In one embodiment, the structural details of the TLM model are as follows:

[0045] Encoder: 4-layer Transformer, 8 heads per layer, 512 hidden layer dimensions, activation function GELU (Gaussian Error Linear Unit).

[0046] Decoder: 2-layer bidirectional LSTM, number of hidden units = 256, Dropout rate = 0.2.

[0047] The TLM model can generate high-precision drilling accident probability, providing a scientific basis for early warning.

[0048] Figure 2 A flowchart illustrating a specific example of a drilling accident early warning method provided in this embodiment of the invention is shown below. Figure 2 As shown, in one embodiment, the drilling accident early warning method may further include:

[0049] Step 201: Obtain historical drilling multi-source data; historical drilling multi-source data includes historical data of drilling engineering dynamic parameters, drilling fluid static parameters, and geological dynamic parameters.

[0050] Step 202: Align historical drilling multi-source data using a precise time synchronization protocol;

[0051] Step 203: Extract multimodal time series features from the aligned historical drilling multi-source data to fuse static and dynamic historical drilling multi-source data and obtain the time series features of historical drilling multi-source data.

[0052] Step 204: Using the time-series characteristics of historical drilling multi-source data, the TLM model is trained by weighting the mask reconstruction loss and contrastive learning loss to generate a drilling accident probability generation model.

[0053] By following the steps above, historical data can be fully utilized to improve the training quality and generalization ability of the model.

[0054] In one embodiment, the pre-training task of the TLM model is to jointly optimize masked reconstruction and contrastive learning to improve the model's feature representation capabilities.

[0055] In one embodiment, the loss function of the TLM model is:

[0056]

[0057] Where λ1 = 0.7, λ2 = 0.3, the reconstruction loss is used to enhance the robustness of the feature representation; L is the total loss value; The focus loss function; MAE(X) recon ,X true ) represents the average absolute error loss.

[0058] In one embodiment, the Transformer encoder employs a probabilistic sparse self-attention mechanism, the calculation formula for which is as follows:

[0059]

[0060] Where Q is the query matrix; K is the key matrix; and V is the value matrix; For the sparsed Query matrix, K T d is the transpose of the key matrix; k is the dimension of the key vector.

[0061] In one embodiment, the drilling accident determination logic includes a dynamic threshold for determining the drilling accident; the adjustment algorithm for the dynamic threshold is shown in the following formula:

[0062] τ t =μ t-1 +k×σ t-1 +β×P(y t =1|X 0:t );

[0063] Where, τ t For dynamic threshold; μ t-1 σ is the mean within the t-1 sliding window; t-1 y is the standard deviation within the t-1 sliding window; k and β are Bayesian weighting factors; t X represents the probability of a drilling accident occurring at time t. 0:t This represents all data sequences up to time t. For example, with a window size of 60 seconds, k = 3, and β = 0.5. A dynamic threshold adjustment algorithm can adaptively adjust the warning threshold, improving the sensitivity and accuracy of the warning.

[0064] In this embodiment, the adaptive threshold verification experiment is as follows:

[0065] Test conditions: The dataset contains historical data from an oilfield with 30 false alarm events. The comparison methods include a fixed threshold (±3σ), a sliding window threshold, and a dynamic threshold in this embodiment of the invention. The comparison results are shown in Table 2 below:

[0066] Table 2

[0067]

[0068]

[0069] In one embodiment, the drilling multi-source data further includes: vibration spectrum data; aligning the drilling multi-source data using a precise time synchronization protocol, including: aligning real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, geological dynamic parameters, and vibration spectrum data using a precise time synchronization protocol; inputting the time-series features of the drilling multi-source data into a pre-trained drilling accident probability generation model and outputting the drilling accident probability, including: inputting the time-series features of the drilling multi-source data obtained after extracting multi-modal time-series features from the aligned real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, geological dynamic parameters, and vibration spectrum data into the pre-trained drilling accident probability generation model and outputting the drill string fracture accident probability.

[0070] In this embodiment, in the drill string failure prediction scenario, vibration spectrum data is added to the multi-source drilling data. After aligning and extracting multimodal temporal features from the multi-source drilling data including vibration spectrum data, the data is input into a pre-trained drilling accident probability generation model. The result obtained is the drill string fracture accident probability. The drill string fracture accident probability can display the probability of a drill string fracture accident occurring.

[0071] In this embodiment, predicting drill string failure may further include the following steps:

[0072] Data preprocessing: Wavelet packet decomposition was performed on the vibration spectrum data (0-10kHz) to extract the energy of 6 sub-bands as features, using the following formula: Among them, W j，k E represents the wavelet coefficients of the j-th layer; j denoted as the energy of the j-th subband; k is the index of the wavelet coefficient; and N is the total number of wavelet coefficients in the j-th layer.

[0073] Model training: Pre-training data was introduced during the training of the drilling accident probability generation model. The pre-training data included data from 10 wells with 2,000 hours of drill string failure records. The model fine-tuning strategy adopted the Adapter module, training only 0.5% of the parameters (approximately 50,000 parameters).

[0074] Deployment results: In a deep water well, fatigue cracks in the drill collar were detected 45 minutes in advance (risk score > 0.85), with an accuracy rate of 95%.

[0075] By incorporating vibration spectrum data, the condition of the drill string can be assessed more comprehensively, improving the accuracy of early warning for drill collar fatigue crack accidents. This embodiment allows for the identification of vibration instability risk trends tens of minutes before visible cracks appear in the drill string, through deep fusion of multi-source data and time-series modeling. This provides operators with valuable intervention time, avoiding costly drill string breakage and non-productive time losses.

[0076] In one embodiment, the physical rule base may include logic for determining well kick accidents, lost circulation accidents, and / or stuck pipe accidents; the well kick accident determination logic includes: triggering a well kick accident warning when the following formula is true:

[0077] (ΔQ=Q in -Q out >0.2) AND (P 立管 <0.8P 预测 );

[0078] Where ΔQ is the difference between the inflow and outflow rates; Q in For import flow; Q out P represents the export flow rate. 立管 For riser pressure; P 预测 This is the predicted value for riser pressure.

[0079] In this embodiment, the physical rule base can be implemented using encoding, for example: IF(ΔQ=Q in -Q out >0.2) AND (P 立管 <0.8P 预测THEN triggers a well kick warning (priority = emergency). The physical rule base contains the judgment logic for well kicks, lost circulation, and stuck pipe, and the priorities are graded according to the (API RP 59) standard. Through this step, the accuracy and reliability of the warning can be improved by combining the physical rule base and model prediction results.

[0080] This invention also proposes a drilling accident early warning device, the principle of which is similar to the drilling accident early warning method, and will not be described in detail here.

[0081] Figure 3 This is a schematic diagram of a drilling accident early warning device provided in an embodiment of the present invention, such as... Figure 3 As shown, a drilling accident early warning device may include:

[0082] The acquisition module 301 is used to acquire drilling multi-source data; the drilling multi-source data includes real-time data of drilling engineering dynamic parameters, drilling fluid static parameter data, and real-time data of geological dynamic parameters.

[0083] Alignment module 302 is used to align drilling multi-source data using a precise time synchronization protocol;

[0084] The feature extraction module 303 is used to extract multimodal time-series features from the aligned drilling multi-source data in order to fuse static and dynamic drilling multi-source data and obtain the time-series features of the drilling multi-source data.

[0085] Output module 304 is used to input the time series features of drilling multi-source data into a pre-trained drilling accident probability generation model and output the drilling accident probability. The drilling accident probability generation model is generated by training the TLM model using the time series features of historical drilling multi-source data. The TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism.

[0086] The early warning triggering module 305 is used to trigger a drilling accident early warning based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base; the physical rule base includes the judgment logic for drilling accidents.

[0087] In one embodiment, the real-time data of drilling dynamic parameters includes: real-time data of drilling pressure, real-time data of torque and / or real-time data of rotational speed; the data of drilling fluid static parameters includes: drilling fluid density data, drilling fluid gas content data and / or hydrogen sulfide concentration data; and the data of geological dynamic parameters includes: gamma ray data, resistivity data and / or pore pressure data.

[0088] Figure 4 This is a schematic diagram of a drilling accident early warning device provided in an embodiment of the present invention, such as... Figure 4As shown, the drilling accident early warning device may further include: a training module 401, used for:

[0089] Acquire historical drilling multi-source data; historical drilling multi-source data includes historical data of drilling engineering dynamic parameters, drilling fluid static parameter data, and geological dynamic parameter data.

[0090] A precise time synchronization protocol is used to align historical drilling data from multiple sources.

[0091] Multimodal time-series features are extracted from the aligned historical drilling multi-source data to fuse static and dynamic historical drilling multi-source data and obtain the time-series features of historical drilling multi-source data.

[0092] By utilizing the time-series characteristics of historical drilling multi-source data, the TLM model is trained using the weighted result of mask reconstruction loss and contrastive learning loss to generate a drilling accident probability generation model.

[0093] In one embodiment, the drilling accident determination logic includes a dynamic threshold for determining drilling accidents.

[0094] The algorithm for adjusting the dynamic threshold is shown in the following formula:

[0095] τ t =μ t-1 +k×σ t-1 +β×P(y t =1|X 0:t );

[0096] Where, τ t For dynamic threshold; μ t-1 σ is the mean within the t-1 sliding window; t-1 y is the standard deviation within the t-1 sliding window; k and β are Bayesian weighting factors; t X represents the probability of a drilling accident occurring at time t. 0:t This represents all data sequences up to time t.

[0097] In one embodiment, drilling multi-source data may further include: vibration spectrum data;

[0098] Alignment module 302 is specifically used for:

[0099] A precise time synchronization protocol is used to align real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, geological dynamic parameters, and vibration spectrum data.

[0100] Output module 304 is specifically used for:

[0101] The drilling multi-source data time series features obtained by extracting multimodal time series features from the aligned drilling engineering dynamic parameter real-time data, drilling fluid static parameter data, geological dynamic parameter real-time data and vibration spectrum data are input into a pre-trained drilling accident probability generation model, and the output is the probability of drill string fracture accident.

[0102] In one embodiment, the physical rule base includes the logic for determining well kick accidents, lost circulation accidents, and / or stuck drill accidents;

[0103] The logic for determining a well kick accident may include:

[0104] A well kick accident warning is triggered when the following formula is true:

[0105] (ΔQ=Q in -Q out >0.2) AND (P 立管 <0.8P 预测 );

[0106] Where ΔQ is the difference between the inflow and outflow rates; Q in For import flow; Q out P represents the export flow rate. 立管 For riser pressure; P 预测 This is the predicted value for riser pressure.

[0107] Compared with existing drilling accident early warning technologies, this invention acquires multi-source drilling data, including real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, and geological dynamic parameters. A precise time synchronization protocol is used to align the multi-source drilling data. Multi-modal time-series features are extracted from the aligned data to fuse static and dynamic data, resulting in multi-source drilling data time-series features. These features are then input into a pre-trained drilling accident probability generation model, which outputs the drilling accident probability. Therefore, the probability generation model is generated by training the TLM model using the time-series features of historical drilling multi-source data. The TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder. The Transformer encoder adopts a probabilistic sparse self-attention mechanism. Based on the drilling multi-source data, the probability of drilling accidents, and the pre-configured physical rule base, drilling accident early warning is triggered. The physical rule base includes the judgment logic of drilling accidents, which can reduce the false negative rate of drilling accidents, shorten the inference delay of drilling accident early warning, and improve the cross-well migration capability and accuracy of drilling accident early warning.

[0108] The embodiments of the present invention can achieve the following technical effects compared with the prior art:

[0109] (a) The comparison of model inference speed is shown in Table 3 below:

[0110] Table 3

[0111] Model Parameters Inference delay (1000 time steps) Existing technology 1.2M 320ms TLM (This invention) 4.8M 85ms

[0112] (II) The economic benefits are compared as shown in Table 4 below:

[0113] Table 4

[0114] index Existing technology This invention Losses from a single well accident (in ten thousand yuan) 280 160 Non-productive time (days / well) 5.2 2.1

[0115] This invention also provides a computer device. Figure 5 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device 500 includes a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 530, it implements the above-mentioned drilling accident early warning method.

[0116] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described drilling accident early warning method.

[0117] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described drilling accident early warning method.

[0118] 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.

[0119] 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 1A device that provides the functions specified in one or more boxes.

[0120] 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.

[0121] 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.

[0122] 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 drilling accident early warning method, characterized in that, include: Acquire multi-source drilling data; the multi-source drilling data includes real-time data of dynamic parameters of drilling engineering, static parameters of drilling fluid, and real-time data of dynamic geological parameters. A precise time synchronization protocol is used to align drilling data from multiple sources; Multimodal time-series features are extracted from the aligned drilling multi-source data to fuse static and dynamic drilling multi-source data and obtain the time-series features of the drilling multi-source data. The time-series features of the drilling multi-source data are input into a pre-trained drilling accident probability generation model, which outputs the drilling accident probability. The drilling accident probability generation model is generated by training a TLM model using the time-series features of historical drilling multi-source data; the TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder; the Transformer encoder adopts a probabilistic sparse self-attention mechanism; Based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base, a drilling accident early warning is triggered; the physical rule base includes the judgment logic for drilling accidents.

2. The method as described in claim 1, characterized in that, The real-time dynamic parameters of the drilling project include: real-time drilling pressure data, real-time torque data, and / or real-time rotational speed data; the static parameters of the drilling fluid include: drilling fluid density data, drilling fluid gas content data, and / or hydrogen sulfide concentration data; the real-time geological dynamic parameters include: gamma ray data, resistivity data, and / or pore pressure data.

3. The method as described in claim 1, characterized in that, Also includes: Acquire historical drilling multi-source data; the historical drilling multi-source data includes historical data of drilling engineering dynamic parameters, drilling fluid static parameter data, and geological dynamic parameter data. A precise time synchronization protocol is used to align historical drilling data from multiple sources. Multimodal time-series features are extracted from the aligned historical drilling multi-source data to fuse static and dynamic historical drilling multi-source data and obtain the time-series features of historical drilling multi-source data. By utilizing the time-series characteristics of historical drilling multi-source data, the TLM model is trained using the weighted result of mask reconstruction loss and contrastive learning loss to generate a drilling accident probability generation model.

4. The method as described in claim 1, characterized in that, The drilling accident determination logic includes a dynamic threshold for determining drilling accidents. The algorithm for adjusting the dynamic threshold is shown in the following formula: t t =μ t-1 +k×σ t-1 +β×P(y t =1|X 0:t ); Where, τ t For dynamic threshold; μ t-1 σ is the mean within the t-1 sliding window; t-1 y is the standard deviation within the t-1 sliding window; k and β are Bayesian weighting factors; t X represents the probability of a drilling accident occurring at time t. 0:t This represents all data sequences up to time t.

5. The method as described in claim 1, characterized in that, The drilling multi-source data also includes: vibration spectrum data; Alignment of multi-source drilling data using a precise time synchronization protocol, including: A precise time synchronization protocol is used to align real-time data of drilling engineering dynamic parameters, drilling fluid static parameters, geological dynamic parameters, and vibration spectrum data. The time-series features of these multi-source drilling data are input into a pre-trained drilling accident probability generation model, which outputs the drilling accident probability, including: The drilling multi-source data time series features obtained by extracting multimodal time series features from the aligned drilling engineering dynamic parameter real-time data, drilling fluid static parameter data, geological dynamic parameter real-time data and vibration spectrum data are input into a pre-trained drilling accident probability generation model, and the output is the probability of drill string fracture accident.

6. The method as described in claim 1, characterized in that, The physical rule base includes the judgment logic for well kick accidents, well leakage accidents, and / or stuck drill accidents; The logic for determining the well kick accident includes: A well kick accident warning is triggered when the following formula is true: (ΔQ=Q in -Q out >0.2)AND(P 立管 <0.8P 预测 ); Where ΔQ is the difference between the inflow and outflow rates; Q in For import flow; Q out P represents the export flow rate. 立管 For riser pressure; P 预测 This is the predicted value for riser pressure.

7. A drilling accident early warning device, characterized in that, include: The acquisition module is used to acquire multi-source drilling data, which includes real-time data of dynamic drilling parameters, static drilling fluid parameters, and real-time geological dynamic parameters. The alignment module is used to align multi-source drilling data using a precise time synchronization protocol; The feature extraction module is used to extract multimodal time-series features from the aligned drilling multi-source data, so as to fuse static and dynamic drilling multi-source data to obtain the time-series features of drilling multi-source data. The output module is used to input the time-series features of the drilling multi-source data into a pre-trained drilling accident probability generation model and output the drilling accident probability. The drilling accident probability generation model is generated by training a TLM model using the time-series features of historical drilling multi-source data; the TLM model includes a Transformer encoder and a bidirectional long short-term memory network decoder; the Transformer encoder adopts a probabilistic sparse self-attention mechanism; The early warning triggering module is used to trigger a drilling accident early warning based on multi-source drilling data, the probability of drilling accidents, and a pre-configured physical rule base; the physical rule base includes the judgment logic for drilling accidents.

8. 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 6.

9. 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 6.

10. 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 6.

Images