Full-sample-based Detection Method, Device and Computer Equipment for Drilling Risks

Abstract

This specification provides a detection method, device, and computer equipment for drilling risks based on all samples, which can be used in the field of oil and gas drilling. Based on this method, before specific implementation, according to the preset training rules, in the model construction stage, by effectively using the sample logging data of all sample wells, an initial joint risk prediction model is constructed, which at least integrates an intermediate first-class detection model, an intermediate second-class detection model, and an intermediate third-class detection model; in the model application stage, by effectively using the logging data of the target well where the model prediction fails during the drilling process of the target well for negative sample learning, a target joint risk prediction model that meets the requirements is trained. During specific implementation, the target joint risk prediction model is used to detect whether there is a drilling risk in the target well by processing the logging data of the target well. Thus, it can efficiently and accurately detect and judge whether there is a drilling risk during the drilling process of the target well, and effectively protect the safety of drilling construction.

Description

Technical Field

[0001] This specification belongs to the technical field of oil and gas drilling, and particularly relates to a detection method, device, and computer equipment for drilling risks based on a full sample. Background Art

[0002] As shallow oil and gas resources are gradually depleted, oil and gas drilling is gradually turning to challenging formation areas such as deep and ultra-deep formations. However, in the above formation areas, due to the uncertainty of deep and deep-water geological conditions, the geological environment is relatively complex, and drilling accidents such as overflow are likely to occur during the drilling process, threatening the safety of drilling operations. Moreover, due to the complex and variable geological environment in the above formation areas, the generalization ability and accuracy of the risk detection model trained based on existing methods are relatively weak, resulting in easy false alarms during actual application and affecting normal drilling operations.

[0003] In response to the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] This specification provides a detection method, device, and computer equipment for drilling risks based on a full sample, which can efficiently and accurately detect and judge whether there are drilling risks during the drilling process of a target well, thereby effectively protecting the safety of drilling operations.

[0005] This specification provides a detection method for drilling risks based on a full sample, including:

[0006] Obtain the logging data of the target well for the current time period;

[0007] Extract target valid data that meets the requirements from the logging data of the current time period; wherein, the target valid data at least includes key features;

[0008] Process the target valid data using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule using the sample logging data of sample wells and the logging data of the target well where the model prediction fails during the drilling process of the target well.

[0009] Determine whether there is a drilling risk in the target well currently according to the target prediction result.

[0010] In one embodiment, the key features include at least one of the following: riser pressure, inlet flow rate, outlet flow rate, total pit volume, and total hydrocarbon content.

[0011] In one embodiment, the target joint risk prediction model is trained according to the following manner based on a preset training rule:

[0012] Obtain the sample logging data of the sample wells; and according to the first sample processing rule, use the sample logging data of the sample wells to construct a first training data set and a second training data set;

[0013] Construct an initial first - type detection model and an initial second - type detection model; wherein, the initial first - type detection model at least includes a parallel LSTM structure and GRU structure, and the initial second - type detection model at least includes an LSTM - AE structure;

[0014] Perform supervised learning on the initial first - type detection model using the first training data set to obtain an intermediate first - type detection model; perform unsupervised learning on the initial second - type detection model using the second training data set to obtain an intermediate second - type detection model;

[0015] Construct a corresponding intermediate third - type detection model according to the intermediate first - type detection model; wherein, the intermediate third - type detection model has the same model structure and model parameters as the intermediate first - type detection model;

[0016] Combine the intermediate first - type detection model, the intermediate second - type detection model, and the intermediate third - type detection model to obtain an initial joint risk prediction model;

[0017] Obtain the logging data of the target well collected during the drilling process of the target well; and according to the second sample processing rule, use the logging data of the target well to construct a third training data set and a fourth training data set; wherein, the fourth training data set is generated based on the logging data of the target wells for which the prediction using the initial joint risk prediction model fails;

[0018] Perform incremental learning on the intermediate first - type detection model and the intermediate second - type detection model in the initial joint risk prediction model by using the third training data set; at the same time, perform negative sample learning on the intermediate third - type detection model in the initial joint risk prediction model by using the fourth training data set to obtain a target joint risk prediction model that meets the requirements.

[0019] In one embodiment, according to the first sample processing rule, using the sample logging data of the sample wells to construct a first training data set and a second training data set includes:

[0020] According to the first sample processing rule, obtain and determine the key features associated with the drilling accident through correlation analysis based on the historical logging data of the drilling accident;

[0021] Extract sample valid data from the sample logging data of the sample wells according to the key features;

[0022] According to the valid sample data, filter out the valid sample data in the first time period when a drilling accident occurs and the valid sample data in the second time period before the drilling accident occurs, and construct a first training data set; and according to the valid sample data, filter out the valid sample data without a drilling accident and construct a second training data set.

[0023] In one embodiment, after constructing the first training data set and the second training data set, the method further includes:

[0024] According to the first sample processing rule, determine the sample wells involved in the valid sample data in the first training data set as training sample wells;

[0025] Filter out the valid sample data that does not involve the training sample wells from the valid sample data to construct a test data set and a validation data set for the first type of detection model and the second type of detection model; wherein, the test data set of the first type of detection model is the same as the test data set of the second type of detection model, and the validation data set of the first type of detection model is the same as the validation data set of the second type of detection model.

[0026] In one embodiment, according to the second sample processing rule, using the logging data of the target well, construct a third training data set and a fourth training data set, including:

[0027] Use the initial joint risk prediction model to process the logging data of the target well to obtain the corresponding prediction results;

[0028] According to the logging data of the target well, randomly extract valid data for combination to construct a third training data set;

[0029] According to the prediction results, filter out the logging data for which the model prediction fails from the logging data of the target well; and based on the logging data for which the model prediction fails, extract valid data as negative sample data to construct a fourth training data set.

[0030] In one embodiment, the method further includes:

[0031] Obtain a knowledge graph about the logging data during the drilling process;

[0032] According to the knowledge graph and the negative sample data, determine the failure reasons that cause the model prediction to fail;

[0033] According to the failure reasons, determine the data quality evaluation parameters of the negative sample data;

[0034] According to the data quality evaluation parameters, filter out the negative sample data that meet the requirements from the negative sample data to construct a fourth training data set.

[0035] This specification also provides a detection device for drilling risks based on all samples, including:

[0036] An acquisition module, configured to acquire the logging data of the target well in the current time period;

[0037] An extraction module, configured to extract target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features;

[0038] A processing module, configured to process the target valid data by using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of sample wells and the logging data of the target wells where model prediction fails during the drilling process of the target wells;

[0039] A determination module, configured to determine whether there is a drilling risk in the target well currently according to the target prediction result.

[0040] This specification also provides a computer device, including a processor and a memory for storing processor-executable instructions, and when the processor executes the instructions, the relevant steps of the detection method for drilling risks based on all samples are implemented.

[0041] This specification also provides a computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the relevant steps of the detection method for drilling risks based on all samples are implemented.

[0042] Based on the detection method, device and computer equipment for drilling risks based on all samples provided in this specification, before specific implementation, according to the preset training rules, in the model construction stage, first, based on the sample logging data of the sample wells obtained, construct a first training data set that simultaneously contains abnormal data and normal data where drilling accidents occur, and the samples of normal data and abnormal data are relatively balanced, as well as a second data set that only contains normal data where no drilling accidents occur; then, through supervised training of the initial first-class detection model using the first training data set to specifically learn the variation law of data when drilling accidents occur, and at the same time, perform unsupervised training of the initial second-class detection model using the second training data set to specifically learn the data law in normal data when no drilling accidents occur, so as to obtain an initial joint risk prediction model integrated with an intermediate first-class detection model, an intermediate second-class detection model, and an intermediate third-class detection model; then apply the above initial joint risk prediction model to the target well of concern, and in the model application stage, use this model to process the logging data of the target well; and screen out the logging data with certain ambiguity and complexity that fails in model prediction from the logging data of the target well as negative sample data for this model; furthermore, the logging data of the target well can be used to perform incremental learning on the intermediate first-class detection model and the intermediate second-class detection model to transfer and adapt the knowledge learned by the model based on the sample logging data of the sample wells to the target well; at the same time, use the negative sample data to perform negative sample learning on the intermediate third-class detection model to specifically learn the root causes and principles of failure that are likely to cause model prediction failure in the target well, so as to obtain a target joint risk prediction model that is adapted to the target well in a complex geological environment scenario, has strong generalization ability and good effect. During specific implementation, first extract the target valid data that meets the requirements and contains key features from the logging data of the target well in the current time period; then use the target joint risk prediction model to process this target valid data to detect whether there are drilling risks in the target well. Thus, by comprehensively and specifically using the sample logging data of all sample wells and the existing logging data of the target well, fully excavate and utilize the data information hidden in the relevant data, and construct and train a target joint risk prediction model with high accuracy and good effect for the target well at a relatively low data processing cost. Furthermore, by using this target joint risk prediction model, it is possible to efficiently and accurately detect and judge whether there are drilling risks during the drilling process of the target well, effectively protecting the safety of drilling construction. Description of the Drawings

[0043] To more clearly illustrate the embodiments of this specification, the following will briefly introduce the accompanying drawings required for the embodiments. The accompanying drawings in the following description are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0044] Figure 1 is a schematic flowchart of a detection method for drilling risks based on a full sample provided by an embodiment of this specification;

[0045] Figure 2 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0046] Figure 3 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0047] Figure 4 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0048] Figure 5 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0049] Figure 6 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0050] Figure 7 is a schematic diagram of the structural composition of a computer device provided by an embodiment of this specification;

[0051] Figure 8 is a schematic diagram of the structural composition of a detection device for drilling risks based on a full sample provided by an embodiment of this specification;

[0052] Figure 9 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example;

[0053] Figure 10 is a schematic diagram of an embodiment applying the detection method for drilling risks based on a full sample provided by an embodiment of this specification in a scenario example. Detailed implementation manners

[0054] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this specification.

[0055] It should be noted that in the embodiments of this specification, some existing industry solutions such as certain software, components, models, etc. may be mentioned. They should be considered exemplary. The purpose is only to illustrate the feasibility in the implementation of the technical solutions of this application, but it does not mean that the applicant has already or necessarily used this solution.

[0056] Refer to Figure 1 As shown, the embodiments of this specification provide a detection method for drilling risks based on full samples. Among them, when the method is specifically implemented, it may include the following content:

[0057] S101: Obtain the logging data of the target well in the current time period;

[0058] S102: Extract the target valid data that meets the requirements from the logging data in the current time period; among them, the target valid data at least includes key features;

[0059] S103: Process the target valid data using the target joint risk prediction model to obtain the corresponding target prediction result; among them, the target joint risk prediction model at least includes: the target first-class detection model, the target second-class detection model, and the target third-class detection model; the target joint risk prediction model is trained according to the preset training rules using the sample logging data of the sample well and the logging data of the target well where the model prediction fails during the drilling process of the target well.

[0060] S104: Determine whether there is a drilling risk in the target well currently according to the target prediction result.

[0061] Among them, the above-mentioned target well can be specifically understood as the oil and gas well that is currently under drilling construction and is of concern. Specifically, the above-mentioned target well can be an oil and gas well located in a formation area with a complex geological environment such as deep and ultra-deep layers.

[0062] When specifically implemented, while drilling the target well, the corresponding logging data can be collected synchronously.

[0063] The above-mentioned target joint risk prediction model can specifically be understood as an algorithm model applicable to the complex geological environment scenario of the target well, which is pre-trained according to preset training rules, using the sample logging data of all sample wells; and the logging data of the target well collected during the drilling process of the target well and failed in prediction when using the model for processing and prediction, through various learning methods such as supervised learning, unsupervised learning, and ensemble learning.

[0064] On the one hand, the above-mentioned target joint risk prediction model can comprehensively detect and identify drilling risks by learning the sample logging data of sample wells during the model construction stage; on the other hand, by learning the logging data of the target well and the negative sample data that failed in prediction when using the model alone during the model application stage, it can be well migrated and adapted to the target well in the complex geological environment scenario, and can also effectively avoid false alarms caused by detection and identification errors.

[0065] Specifically, the above-mentioned target joint risk prediction model at least includes a target first-class detection model, a target second-class detection model, and a target third-class detection model connected in parallel.

[0066] Among them, the above-mentioned target first-class detection model is a time series model mainly trained by supervised learning using relatively balanced abnormal data (logging data of drilling accidents) and normal data (logging data without drilling accidents) in the sample logging data of sample wells. The above-mentioned target second-class detection model is a time series model mainly trained by unsupervised learning using all normal data in the sample logging data of sample wells. The above-mentioned target third-class detection model is a time series model mainly trained by supervised learning using negative sample data (logging data that failed in prediction when using the model) in the logging data of the target well.

[0067] Based on the above-mentioned target joint risk prediction model, the input data can be separately processed by the above-mentioned target first-class detection model, target second-class detection model, and target third-class detection model to obtain three risk probability prediction values based on different dimensions; then the three risk probability prediction values are jointly used, and the final risk probability prediction value is determined by voting as the target prediction result output by the model.

[0068] Among them, the above-mentioned supervised learning can specifically refer to a learning method of training a model using labeled sample data.

[0069] The above-mentioned unsupervised learning can specifically refer to a learning method of training a model using unlabeled sample data.

[0070] The above-mentioned Ensemble Learning can specifically refer to a learning method that combines multiple models through a voting mechanism to improve the overall prediction or classification performance, which can effectively improve the generalization ability of the model and reduce overfitting in training.

[0071] The above-mentioned key features can specifically be understood as the features that have a relatively obvious impact and strong relevance on drilling risk prediction.

[0072] The above-mentioned drilling risk can specifically be understood as the risk of drilling accidents. Among them, the above-mentioned drilling accidents can specifically include overflow accidents, etc. Of course, it should be noted that the above-listed overflow accidents are only illustrative. In specific implementation, according to the specific application scenario and processing requirements, the above-mentioned drilling accidents can also include other types of accidents, such as well collapse, blowout, etc. This specification does not make any limitations in this regard.

[0073] In specific implementation, first, at least the target effective data containing key features can be extracted from the mud logging data in the current time period. In this way, on the one hand, the data processing volume of the subsequent model can be reduced, and on the other hand, the model can be more focused on analyzing the key features that are relatively more important for drilling risk prediction, reducing the interference of other factors in the mud logging data except for the key features, so that the model can subsequently predict the drilling risk relatively more efficiently and accurately.

[0074] Then, the pre-processed target joint risk prediction model can be used to process the target effective data while taking into account multiple dimensions such as the change law of the transformation from normal data to abnormal data when a drilling accident occurs, the data law of normal data when no drilling accident occurs, and the interference law that easily leads to model prediction failure, so as to accurately and efficiently detect and judge whether there is a drilling risk in the target well.

[0075] Based on the above embodiments, by training and using the target joint risk prediction model that is adapted to the target well in a complex geological environment scenario and has strong generalization and good effects, it is possible to efficiently and accurately detect and judge whether there is a drilling risk during the drilling process of the target well, effectively protecting the safety of drilling construction.

[0076] In some embodiments, the above-mentioned key features can specifically include at least one of the following: riser pressure, inlet flow rate, outlet flow rate, total pit volume, total hydrocarbon content, etc.

[0077] Among them, the above-mentioned riser pressure is identified as one of the key features because pressure changes are an important indicator for monitoring downhole dynamics during drilling operations. When a drilling accident such as an overflow occurs, the drilling fluid begins to flow freely and the balance of the static liquid column is broken. As the drilling fluid moves upward, the height of the static liquid column decreases, and the bottom hole pressure decreases accordingly. Inside the drill bit, the impact of fluid intrusion is small, and the static liquid column pressure and circulating pressure loss remain almost unchanged. Therefore, a decrease in bottom hole pressure will lead to a decrease in riser pressure. In turn, it is possible to consider using riser pressure as a key feature to analyze and predict drilling accidents such as overflows.

[0078] The changes in the inlet flow rate and outlet flow rate can also reflect the occurrence of drilling accidents such as overflow. Under normal circumstances, the inlet flow rate and outlet flow rate will remain relatively stable. However, when a drilling accident such as overflow occurs, the outlet flow rate will increase without increasing the inlet flow rate. Therefore, this feature can be used to analyze and predict drilling accidents such as overflow.

[0079] The above-mentioned total pool volume, as a key parameter to describe the drilling fluid reserve, also has an important impact on the prediction of drilling accidents such as overflows. Specifically, as the drilling fluid returns, the total pool volume will gradually increase. In the case of oil and water invasion, as the outlet flow rate increases, the total pool volume will also increase accordingly; in the case of gas invasion, due to the expansion of bubbles, the outlet flow rate increases, and the total pool volume will also increase. This change will appear soon after oil and water invasion and gas invasion occur, so this feature can be considered as an important clue to predict drilling accidents such as overflows.

[0080] The above changes in total hydrocarbon content usually involve the gas components in the drilling fluid. Specifically, when gas invasion occurs, as the gas circulates to the ground, the total hydrocarbon content will increase, which can effectively indicate drilling accidents such as overflows. Therefore, it is possible to consider introducing this feature to improve the model's ability to identify drilling accidents such as gas invasion and overflows.

[0081] In some embodiments, the above-mentioned extraction of target valid data that meets the requirements from the logging data of the current time period may include: based on key features, processing the logging data of the current time period to extract data that can characterize the key features and combining them as target valid data that meets the requirements corresponding to the current time period.

[0082] After obtaining the target valid data, the target valid data may be further preprocessed to eliminate data errors in the target valid data, so that the target valid data is relatively more suitable for processing by the target joint risk prediction model.

[0083] The above preprocessing may specifically include at least one of the following: data cleaning, outlier detection, missing value processing, data smoothing, etc.

[0084] In some embodiments, after determining whether there is a drilling risk in the target well according to the target prediction result, when the method is specifically implemented, the following contents may further be included:

[0085] S1: Determine the risk type and risk level of the drilling risk;

[0086] S2: Query the preset risk handling strategy library according to the risk type and risk level, and determine the matching target risk handling strategy;

[0087] S3: Eliminate the risk of the target well according to the target risk handling strategy.

[0088] Among them, the preset risk handling strategy library may specifically store multiple preset risk handling strategies, and each preset risk handling strategy corresponds to at least one combination of risk type and risk level.

[0089] Before specific implementation, the preset risk handling strategy library can be constructed in the following way: collect a large number of historical drilling risk handling records during historical drilling; screen out the records with successful handling from the historical drilling risk handling records as sample drilling risk handling records; perform clustering processing on the sample drilling risk handling records to obtain multiple data groups; among them, each data group corresponds to a combination of risk type and risk level, and contains common handling contents regarding drilling risk handling; according to the common handling contents regarding drilling risk handling included in each data group, combine with expert experience to construct multiple preset risk handling strategies; combine multiple preset risk handling strategies to obtain the corresponding preset risk handling strategy library.

[0090] During specific implementation, drilling risk prompt information about the target well can be generated; among them, the drilling risk prompt information carries at least the matching target risk handling strategy; send the drilling risk prompt information to the drilling construction end of the target well to prompt the drilling construction end of relevant drilling risks, and guide the drilling construction end to refer to the target risk handling strategy to eliminate the drilling risk in a timely and accurate manner.

[0091] In some embodiments, referring to Figure 2 as shown, the target joint risk prediction model can be specifically trained according to the preset training rules in the following way:

[0092] S1: Obtain the sample logging data of the sample well; and according to the first sample processing rule, use the sample logging data of the sample well to construct the first training data set and the second training data set;

[0093] S2: Construct an initial first - type detection model and an initial second - type detection model; wherein, the initial first - type detection model at least includes a parallel LSTM structure and GRU structure, and the initial second - type detection model at least includes an LSTM - AE structure;

[0094] S3: Perform supervised learning on the initial first - type detection model using the first training dataset to obtain an intermediate first - type detection model; perform unsupervised learning on the initial second - type detection model using the second training dataset to obtain an intermediate second - type detection model;

[0095] S4: Construct a corresponding intermediate third - type detection model according to the intermediate first - type detection model; wherein, the intermediate third - type detection model has the same model structure and model parameters as the intermediate first - type detection model;

[0096] S5: Combine the intermediate first - type detection model, the intermediate second - type detection model, and the intermediate third - type detection model to obtain an initial joint risk prediction model;

[0097] S6: Obtain the logging data of the target well collected during the drilling process of the target well; and according to the second sample processing rule, use the logging data of the target well to construct a third training dataset and a fourth training dataset; wherein, the fourth training dataset is generated based on the logging data of the target wells for which the prediction using the initial joint risk prediction model fails;

[0098] S7: Perform incremental learning on the intermediate first - type detection model and the intermediate second - type detection model in the initial joint risk prediction model by using the third training dataset; at the same time, perform negative - sample learning on the intermediate third - type detection model in the initial joint risk prediction model by using the fourth training dataset to obtain a target joint risk prediction model that meets the requirements.

[0099] Based on the above - mentioned embodiments, according to the preset training rules, by constructing and combining multiple detection models with different structures and different types, and at the same time introducing and adopting various different learning methods such as supervised learning, unsupervised learning, and ensemble learning, the existing full - volume sample data can be comprehensively and effectively utilized, the data information carried in the sample data can be fully mined and used, and with a relatively small data processing cost, a target joint risk prediction model with good effect and small error, which is adapted to the target well in a complex geological environment scenario, can be trained more efficiently.

[0100] Among them, the above - mentioned sample wells can be specifically understood as oil and gas wells that have been drilled and are different from the target well.

[0101] In some embodiments, the above-mentioned first training dataset may specifically be a sample dataset constructed by using the normal data and abnormal data with sample balance in the sample logging data of the sample well during the model construction phase. Here, the above-mentioned abnormal data may specifically be the logging data during drilling accidents such as blowouts. The above-mentioned normal data may specifically be the logging data during the normal drilling period before drilling accidents such as blowouts occur.

[0102] The above-mentioned second training dataset may specifically be a sample dataset constructed by using the normal data in the sample logging data of the sample well during the model construction phase. Here, the above-mentioned normal data may specifically be the logging data during the normal drilling period without drilling accidents such as blowouts.

[0103] The above-mentioned third training dataset may specifically be a sample dataset constructed by using the logging data of the target well collected during the drilling process of the target well during the model application phase.

[0104] The above-mentioned fourth training dataset may specifically be a sample dataset constructed by using the logging data that fails to be predicted or even gives false alarms (which can be recorded as negative sample data) when using the existing joint risk prediction model to process the logging data of the target well during the model application phase.

[0105] The above-mentioned LSTM (Long Short-Term Memory) structure can specifically be understood as a type of time series recurrent neural network.

[0106] The above-mentioned GRU (Gate Recurrent Unit) structure can specifically be understood as a variant structure of an RNN (Recurrent Neural Network). Among them, the reset gate in the GRU helps to extract short-term dependencies in the sequence, and the update gate helps to obtain long-term dependencies of the sequence data.

[0107] The above-mentioned LSTM-AE (Long Short Term Memory Autoencoder) structure can specifically be understood as a type of time series autoencoder.

[0108] It should be noted that by introducing and using the LSTM-AE structure here, the neural network can be used to learn the low-dimensional representation of the input data and try to restore the original data as much as possible during the decoding process. By minimizing the difference (such as mean square error) between the input data and the reconstructed data, useful feature representations of the data can be learned. Furthermore, by using the above-mentioned LSTM-AE structure, by fully learning the full amount of normal data in the sample logging data, the data laws of the normal data can be effectively mastered, so as to accurately detect and identify the normal data.

[0109] All of the above three model structures can capture and learn the temporal dependencies of the data.

[0110] In some embodiments, referring to Figure 3 as shown, according to the first sample processing rule, using the sample logging data of the sample well, the first training data set and the second training data set are constructed. Specifically, in implementation, it may include the following content:

[0111] S1: According to the first sample processing rule, obtain and based on the historical logging data of drilling accidents, through correlation analysis, determine the key features associated with drilling accidents;

[0112] S2: According to the key features, extract the sample valid data from the sample logging data of the sample well;

[0113] S3: According to the sample valid data, screen out the sample valid data in the first time period when a drilling accident occurs and the sample valid data in the second time period before a drilling accident occurs, and construct the first training data set; and according to the sample valid data, screen out the sample valid data without drilling accidents and construct the second training data set.

[0114] Based on the above embodiments, the sample logging data of the sample well can be fully and comprehensively utilized to construct the first training data set and the second training data set based on different dimensions and for different learning objectives.

[0115] In some embodiments, according to the first sample processing rule, obtain and based on the historical logging data of drilling accidents, through correlation analysis, determine the key features associated with drilling accidents. Specifically, in implementation, it may include the following content.

[0116] First, obtain and based on the historical logging data of drilling accidents, combine relevant background information and multi-physics field action models to determine the factors related to drilling accidents as candidate factors; based on the candidate factors, conduct test experiments and collect the corresponding experimental test data; according to the experimental test data, conduct data statistics to obtain the corresponding statistical results; according to the statistical results, construct a correlation matrix of candidate factors and drilling accidents, specifically, it can be referred to Figure 4 as shown.

[0117] When specifically constructing the correlation matrix, the candidate factors can be used as the vertical axis and the horizontal axis; at the same time, normalize the statistical results to determine the proportion of various combinations of candidate factors when a drilling accident occurs; and according to this proportion, map it into the corresponding gray value; use this gray value to fill the corresponding block to obtain the above correlation matrix.

[0118] Next, according to the correlation matrix, correlation analysis can be carried out by calculating and utilizing the contribution values of each candidate factor to the drilling accident, and the candidate factors with contribution values greater than the preset contribution threshold are screened out as the key features.

[0119] Specifically, during implementation, the appropriate first time period and second time period can be determined by counting and based on the historical logging data of drilling accidents. For the sample valid data, the occurrence time point of the drilling accident is used as the starting point; starting from the starting point, the valid data of the first time period is intercepted backward along the time axis to obtain abnormal data, that is, the sample valid data of the first time period when the drilling accident occurs. At the same time, starting from the starting point, the valid data of the second time period is intercepted forward along the time axis to obtain normal data, that is, the sample valid data of the second time period before the drilling accident occurs. Then, the sample valid data of the first time period when the drilling accident occurs and the sample valid data of the second time period before the drilling accident occurs are combined to obtain a relatively balanced first training data set.

[0120] At the same time, all the valid data without drilling accidents is screened out from the sample valid data for combination to obtain a second training data set.

[0121] When specifically using the first training data set for supervised learning of the initial first type of detection model, the first training data set can be labeled first; then the initial first type of detection model is trained using the labeled first training data set.

[0122] In addition, when specifically training the initial first type of detection model, the first training data set can be split into two subsets; and one subset is used to train the LSTM structure, and the other subset is used to train the GRU structure.

[0123] When specifically using the second training data set for unsupervised learning of the initial second type of detection model, the unlabeled second training data set can be directly used to train the initial second type of detection model.

[0124] After obtaining the intermediate first type of detection model through supervised learning, an intermediate first type of detection model can be copied to obtain the corresponding intermediate third type of detection model.

[0125] After obtaining the intermediate first type of detection model, intermediate second type of detection model, and intermediate third type of detection model, the Boosting integration method or Bagging integration method can also be used to combine the above three models; and corresponding voting rules are set to obtain the initial joint risk prediction model.

[0126] In some embodiments, considering that the proportion of abnormal data in the sample logging data is relatively small, in order to improve the training effect of the first type of detection model, after the sample valid data in the first time period when a drilling accident occurs and the sample valid data in the second time period before the drilling accident occur, the sample valid data in the first time period when a drilling accident occurs and the sample valid data in the second time period before the drilling accident occur can also be respectively subjected to data augmentation processing to obtain the augmented and extended sample valid data in the first time period and the augmented and extended sample valid data in the second time period; then, the augmented and extended sample valid data in the first time period and the augmented and extended sample valid data in the second time period are combined and used to construct a relatively richer and better-trained first training dataset.

[0127] In some embodiments, after the first training dataset and the second training dataset are constructed, when the method is specifically implemented, the following content may further be included:

[0128] S1: According to the first sample processing rule, determine the sample wells involved in the sample valid data in the first training dataset as the training sample wells;

[0129] S2: Screen out the sample valid data that does not involve the training sample wells from the sample valid data to construct a test dataset and a validation dataset for the first type of detection model and the second type of detection model; wherein, the test dataset of the first type of detection model is the same as the test dataset of the second type of detection model, and the validation dataset of the first type of detection model is the same as the validation dataset of the second type of detection model.

[0130] Specifically, during implementation, based on the corresponding cross-well data partitioning strategy, the test dataset and the validation dataset can be made to correspond to different sample wells from the first training dataset, so that the model can obtain better generalization ability during training.

[0131] Specifically, after using the first training dataset to perform supervised learning on the initial first type of detection model to obtain an intermediate first type of detection model, and using the second training dataset to perform unsupervised learning on the initial second type of detection model to obtain an intermediate second type of detection model, the test dataset and the validation dataset can also be used to test and verify the intermediate first type of detection model and the second type of detection model, and the test and verification results are collected.

[0132] Furthermore, according to the test and verification results, the intermediate first type of detection model and the intermediate second type of detection model can be respectively adjusted and optimized in a targeted manner to further improve the model accuracy of the intermediate first type of detection model and the intermediate second type of detection model.

[0133] Meanwhile, according to the test and verification results, the model performances of the first intermediate detection model and the second intermediate detection model can be evaluated; according to the evaluation results, the voting rules can be adjusted, for example, the weight coefficients of the voting can be adjusted.

[0134] In addition, according to the test and verification results, the valid sample data that are predicted to fail by the first detection model and / or the second detection model during the test and verification process can be screened out as negative sample data; and the negative sample data can be used to train the third intermediate detection model to improve the model accuracy of the third intermediate detection model.

[0135] By further introducing ensemble learning on the basis of supervised learning and unsupervised learning, on the one hand, the overall accuracy of the model prediction can be improved by combining the predictions of multiple models; on the other hand, by using a combination of multiple models, the possible overfitting problem based on a single model can be reduced, the generalization ability and robustness of the model can be improved, and false alarms can be reduced.

[0136] In some embodiments, as shown in Figure 5 According to the second sample processing rule, using the logging data of the target well, the third training data set and the fourth training data set are constructed. Specifically, the following contents may be included during implementation:

[0137] S1: Use the initial joint risk prediction model to process the logging data of the target well to obtain the corresponding prediction results;

[0138] S2: Randomly extract valid data for combination according to the logging data of the target well to construct the third training data set;

[0139] S3: According to the prediction results, screen out the logging data that fails to be predicted by the model from the logging data of the target well; and based on the logging data that fails to be predicted by the model, extract valid data as negative sample data to construct the fourth training data set.

[0140] Based on the above embodiments, the logging data of the target well collected during the drilling process of the target well can be fully and comprehensively utilized to construct the third training data set and the fourth training data set based on different dimensions and for different learning objectives.

[0141] In some embodiments, during specific implementation, the corresponding logging data can be randomly extracted from the logging data of the target well to construct the third training data set; and the third training data set can be used to perform incremental learning on the first intermediate detection model and the second intermediate detection model.

[0142] In specific implementation, incremental learning based on playback can be adopted, and the intermediate first-class detection model and the intermediate second-class detection model are incrementally learned by using the third training data set, so that the model can retain the old knowledge effective for the target well learned from the sample logging data of the previous sample well; at the same time, the model can further learn and master the new knowledge for the target well, so that the model can be better adapted to the target well in the complex geological environment scenario.

[0143] In specific implementation, the failure reasons for the model prediction failure caused by the negative sample data can be analyzed, and the negative sample data is labeled according to the failure reasons to obtain the corresponding fourth training data set; then the fourth training data set is used to perform negative sample learning on the intermediate third-class detection model, so that the model can avoid false alarms caused by prediction failure as much as possible in the future.

[0144] In some embodiments, after obtaining the sample valid data of the sample well and / or the valid data of the target well, when the method is specifically implemented, it may further include: preprocessing the sample valid data of the sample well and / or the valid data of the target well; wherein, the preprocessing includes at least one of the following: data cleaning, outlier detection, missing value processing, data smoothing processing, etc.

[0145] In some embodiments, when specifically implemented, the joint risk prediction model can be applied to the drilling process of the target well. At each specified time period, the logging data of the target well in the previous time period is obtained; and the third training data set and the fourth training data set in the previous time period are constructed by using the logging data of the target well in the previous time period; then the third training data set and the fourth training data set in the previous time period are used to train and update the target joint risk prediction model to continuously improve and optimize the joint risk prediction model.

[0146] In some embodiments, when specifically implemented, negative sample data with prediction failure can also be randomly extracted from the historical logging data of the target well at regular intervals, combined with positive sample data with successful prediction, to construct a test data set; then the test data set is used to test the current target joint risk prediction model; according to the test results, the target joint risk prediction model is adjusted and optimized.

[0147] In some embodiments, referring to Figure 6 as shown, when the method is specifically implemented, it may further include the following content:

[0148] S1: Obtain a knowledge graph about the logging data in the drilling process;

[0149] S2: Determine the failure reasons for the model prediction failure according to the knowledge graph and the negative sample data;

[0150] S3: Determine the data quality evaluation parameters of the negative sample data according to the failure reasons;

[0151] S4: According to the data quality evaluation parameters, filter out the qualified negative sample data from the negative sample data, and construct a fourth training data set.

[0152] Based on the above embodiments, by obtaining and using relevant knowledge graphs, effective in-depth analysis of negative sample data can be carried out, and then a fourth training data set with relatively better effects can be constructed.

[0153] In some embodiments, during specific implementation, typical case data regarding drilling accidents can be sorted out and constructed according to the historical mud logging data of drilling accidents; and according to the background information and the corresponding historical mud logging data, data elements for the typical case data (such as overflow accidents, riser parameters, inlet parameters, etc.) can be sorted out, as well as the mutual relationships between the data elements; then, based on the above data elements and the mutual relationships between the data elements, combined with expert experience, a knowledge graph can be constructed that can reflect the interaction relationships between different data factors during the drilling process based on drilling accidents.

[0154] Furthermore, based on the negative sample data, the knowledge graph can be queried, and combined with relevant background information, by deeply analyzing the negative sample data, the failure reasons that lead to the failure of model prediction can be determined, such as the characteristic manifestations of model errors and the root causes of abnormal input data, etc.

[0155] Further, according to the above failure reasons, the effects of the negative sample data during model training can be evaluated to obtain data quality evaluation parameters for the negative sample data; then, according to the data quality evaluation parameters, filter out the qualified negative sample data from the negative sample data, and construct a fourth training data set, so as to use the fourth training data set with higher data quality for negative sample learning, enabling the model to relatively more focus on overcoming the relatively wide range of problems that lead to model prediction failure, and improving the generalization and stability of the model.

[0156] In addition, according to the above failure reasons, the model structure and / or the preprocessing rules for valid data can be adjusted to overcome the problems of model prediction failure caused by internal reasons such as model errors or data processing errors.

[0157] In some embodiments, after obtaining the knowledge graph, the first type of detection model, the second type of detection model, and the knowledge graph can also be comprehensively used to process the negative sample data to obtain corresponding prediction classification results; according to the prediction classification results, label the negative sample data; then, use the labeled negative sample data to construct a corresponding fourth training data set for targeted negative sample learning of the third type of detection model.

[0158] As can be seen from the above, based on the detection method, device and computer equipment for drilling risks based on the full sample provided in the embodiments of this specification, before the specific implementation, according to the preset training rules, in the model construction stage, first, based on the sample logging data of the sample wells obtained, a first training data set is constructed, which contains both abnormal data and normal data of drilling accidents, and the normal data and abnormal data are relatively balanced, and a second data set that only contains normal data of wells without drilling accidents; then, the initial first-class detection model is supervised-trained using the first training data set to specifically learn the variation law of data during drilling accidents, and at the same time, the initial second-class detection model is unsupervised-trained using the second training data set to specifically learn the data law in normal data when no drilling accidents occur, so as to obtain an initial joint risk prediction model integrated with the intermediate first-class detection model, intermediate second-class detection model, and intermediate third-class detection model; then, the above initial joint risk prediction model is applied to the target well of concern. In the model application stage, the logging data of the target well is used; and the logging data that fails in model prediction and has a certain degree of ambiguity and complexity is screened out from the logging data of the target well as negative sample data; then, the incremental learning is performed on the intermediate first-class detection model and intermediate second-class detection model using the logging data of the target well to adapt the knowledge learned by the model based on the sample logging data of the sample wells to the target well; at the same time, the negative sample learning is performed on the intermediate third-class detection model using the negative sample data to specifically learn the root causes of failure in the target well that are likely to cause model prediction failure, so as to obtain a target joint risk prediction model that is adapted to the target well in a complex geological environment scenario, has strong generalization ability and good effect. When specifically implementing, the target valid data that meets the requirements can be extracted from the logging data of the target well in the current period; then, the target joint risk prediction model is used to process the target valid data to detect whether there is a drilling risk in the target well. Thus, by comprehensively and specifically using the full-volume sample logging data and the existing logging data of the target well, the relevant data is fully mined and utilized, and a target joint risk prediction model with high accuracy and good effect for the target well is constructed at a relatively low cost. Furthermore, the target joint risk prediction model can be used to efficiently and accurately detect and judge whether there is a drilling risk during the drilling process of the target well, effectively protecting the safety of drilling construction.

[0159] The embodiments of this specification also provide a data processing method. Specifically, when implementing this method, the following contents may be included:

[0160] S1: Obtain the sample logging data of the sample wells; and according to the first sample processing rule, use the sample logging data of the sample wells to construct a first training data set and a second training data set;

[0161] S2: Construct an initial first - type detection model and an initial second - type detection model; among them, the initial first - type detection model at least includes a parallel LSTM structure and GRU structure, and the initial second - type detection model at least includes an LSTM - AE structure;

[0162] S3: Perform supervised learning on the initial first - type detection model using the first training data set to obtain an intermediate first - type detection model; perform unsupervised learning on the initial second - type detection model using the second training data set to obtain an intermediate second - type detection model;

[0163] S4: Construct a corresponding intermediate third - type detection model according to the intermediate first - type detection model; among them, the intermediate third - type detection model has the same model structure and model parameters as the intermediate first - type detection model;

[0164] S5: Combine the intermediate first - type detection model, the intermediate second - type detection model, and the intermediate third - type detection model to obtain an initial joint risk prediction model;

[0165] S6: Obtain the logging data of the target well collected during the drilling process of the target well; and according to the second sample processing rule, use the logging data of the target well to construct a third training data set and a fourth training data set; among them, the fourth training data set is generated based on the logging data of the target wells for which the prediction of the initial joint risk prediction model fails;

[0166] S7: Perform incremental learning on the intermediate first - type detection model and the intermediate second - type detection model in the initial joint risk prediction model by using the third training data set; at the same time, perform negative - sample learning on the intermediate third - type detection model in the initial joint risk prediction model by using the fourth training data set to obtain a target joint risk prediction model that meets the requirements.

[0167] Based on the above - mentioned method, it is possible to train, with a relatively small data - processing cost, a target joint risk prediction model with good effects that is adapted to the target well in a complex geological environment scenario more efficiently.

[0168] The embodiments of this specification provide a computer device, as shown in Figure 7 Among them, the computer device includes a network communication port 701, a processor 702, and a memory 703. The above - mentioned structures are connected by internal cables so that each structure can perform specific data interactions.

[0169] Among them, the network communication port 701 can specifically be used to obtain the logging data of the target well in the current time period.

[0170] The processor 702 can specifically be used to extract target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features; the target joint risk prediction model is used to process the target valid data to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first type detection model, a target second type detection model, and a target third type detection model; the target joint risk prediction model is trained according to a preset training rule using the sample logging data of the sample well and the logging data of the target well where the model prediction fails during the drilling process of the target well; based on the target prediction result, it is determined whether there is a drilling risk in the target well currently.

[0171] The memory 703 can specifically be used to store corresponding instruction programs, as well as related data such as target valid data and the target joint risk prediction model.

[0172] Based on the above method, it is possible to effectively utilize the relevant structural performance of the computer device, improve the data processing speed of the electronic device, and efficiently implement the data processing for drilling risk detection.

[0173] In this embodiment, the network communication port 701 can be bound to different communication protocols, so as to send or receive different data virtual ports. For example, the network communication port can be a port responsible for web data communication, or a port responsible for FTP data communication, or a port responsible for mail data communication. In addition, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it can also be a Wifi chip; it can also be a Bluetooth chip.

[0174] In this embodiment, the processor 702 can be implemented in any suitable manner. For example, the processor can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, logic gates, switches, application specific integrated circuit (ASIC), programmable logic controller, and embedded microcontroller, etc. This specification does not make any limitations.

[0175] In this embodiment, the memory 703 can include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function without a physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0176] An embodiment of this specification also provides a computer-readable storage medium based on the above-mentioned detection method for drilling risks based on all samples. The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed, it realizes: obtaining the logging data of the target well in the current time period; extracting target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features; processing the target valid data by using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of the sample well and the logging data of the target well that fails in model prediction during the drilling process of the target well; determining whether there is a drilling risk for the target well currently according to the target prediction result.

[0177] In this embodiment, the above storage medium includes but is not limited to Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions. The network communication unit can be set according to the standards stipulated by the communication protocol and is used for the interface of network connection communication.

[0178] In this embodiment, the functions and effects specifically realized by the program instructions stored in the computer-readable storage medium can be explained by comparison with other embodiments and will not be elaborated here.

[0179] An embodiment of this specification also provides a computer program product, which at least includes a computer program. When the computer program is executed by a processor, the following method steps are realized: obtaining the logging data of the target well in the current time period; extracting target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features; processing the target valid data by using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of the sample well and the logging data of the target well that fails in model prediction during the drilling process of the target well; determining whether there is a drilling risk for the target well currently according to the target prediction result.

[0180] Refer to Figure 8As shown in the figure, the embodiment of the present specification also provides a detection device for drilling risks based on all samples. The device may specifically include the following structural modules:

[0181] An acquisition module 801, which may specifically be used to acquire the logging data of the target well in the current time period;

[0182] An extraction module 802, which may specifically be used to extract target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features;

[0183] A processing module 803, which may specifically be used to process the target valid data by using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of the sample well and the logging data of the target well where the model prediction fails during the drilling process of the target well;

[0184] A determination module 804, which may specifically be used to determine whether there is a drilling risk in the target well currently according to the target prediction result.

[0185] In some embodiments, the key features may specifically include at least one of the following: riser pressure, inlet flow rate, outlet flow rate, total pit volume, total hydrocarbon content, etc.

[0186] In some embodiments, the device may further include a training module.

[0187] Among them, the above training model can specifically obtain the target joint risk prediction model according to the following method based on the preset training rules: Obtain the sample logging data of the sample well; and according to the first sample processing rule, use the sample logging data of the sample well to construct the first training data set and the second training data set; Construct the initial first type of detection model and the initial second type of detection model; among them, the initial first type of detection model at least includes a parallel LSTM structure and a GRU structure, and the initial second type of detection model at least includes an LSTM-AE structure; Use the first training data set to perform supervised learning on the initial first type of detection model to obtain an intermediate first type of detection model; Use the second training data set to perform unsupervised learning on the initial second type of detection model to obtain an intermediate second type of detection model; According to the intermediate first type of detection model, construct a corresponding intermediate third type of detection model; among them, the intermediate third type of detection model has the same model structure and model parameters as the intermediate first type of detection model; Combine the intermediate first type of detection model, the intermediate second type of detection model, and the intermediate third type of detection model to obtain the initial joint risk prediction model; Obtain the logging data of the target well collected during the drilling process of the target well; and according to the second sample processing rule, use the logging data of the target well to construct the third training data set and the fourth training data set; among them, the fourth training data set is generated based on the logging data of the target well for which the prediction by the initial joint risk prediction model fails; By using the third training data set to perform incremental learning on the intermediate first type of detection model and the intermediate second type of detection model in the initial joint risk prediction model; At the same time, use the fourth training data set to perform negative sample learning on the intermediate third type of detection model in the initial joint risk prediction model to obtain the target joint risk prediction model that meets the requirements.

[0188] In some embodiments, when the above training module is specifically implemented, the first training data set and the second training data set can be constructed according to the following method based on the first sample processing rule, using the sample logging data of the sample well: According to the first sample processing rule, obtain and analyze the historical logging data of drilling accidents through correlation analysis to determine the key features associated with drilling accidents; According to the key features, extract the sample valid data from the sample logging data of the sample well; According to the sample valid data, screen out the sample valid data in the first time period when a drilling accident occurs and the sample valid data in the second time period before a drilling accident occurs to construct the first training data set; And according to the sample valid data, screen out the sample valid data without drilling accidents to construct the second training data set.

[0189] In some embodiments, after constructing the first training data set and the second training data set, when the training module is specifically implemented, it can also be used to: determine, according to the first sample processing rule, the sample wells involved in the sample valid data in the first training data set as the training sample wells; screen out the sample valid data that does not involve the training sample wells from the sample valid data, and construct a test data set and a validation data set for the first type of detection model and the second type of detection model; wherein, the test data set of the first type of detection model is the same as the test data set of the second type of detection model, and the validation data set of the first type of detection model is the same as the validation data set of the second type of detection model.

[0190] In some embodiments, when the above training module is specifically implemented, it can construct the third training data set and the fourth training data set according to the second sample processing rule by using the logging data of the target well in the following manner: process the logging data of the target well with the initial joint risk prediction model to obtain the corresponding prediction results; randomly extract and combine the valid data from the logging data of the target well to construct the third training data set; screen out the logging data for which the model prediction fails from the logging data of the target well according to the prediction results; and extract the valid data based on the logging data for which the model prediction fails as the negative sample data to construct the fourth training data set.

[0191] In some embodiments, when the device is specifically implemented, it can also be used to: obtain a knowledge graph about the logging data during the drilling process; determine the failure reasons that cause the model prediction to fail according to the knowledge graph and the negative sample data; determine the data quality evaluation parameters of the negative sample data according to the failure reasons; and screen out the negative sample data that meets the requirements from the negative sample data according to the data quality evaluation parameters to construct the fourth training data set.

[0192] It should be noted that the units, devices, or modules, etc., described in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described by dividing them into various modules according to functions. Of course, when implementing this specification, the functions of each module can be implemented in the same or multiple software and / or hardware, or the modules that implement the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0193] As can be seen from the above, based on the detection device for drilling risks based on all samples provided in the embodiments of this specification, by comprehensively and specifically utilizing the mud logging data of all samples and the existing mud logging data of the target well, relevant data is fully mined and utilized, and a target joint risk prediction model with relatively high accuracy and good effect for the target well is constructed at a relatively low cost. Furthermore, the target joint risk prediction model can be used to efficiently and accurately detect and determine whether there are drilling risks during the drilling process of the target well, effectively protecting the safety of drilling operations.

[0194] In a specific scenario example, the detection method for drilling risks based on all samples provided in this specification can be applied to implement the diagnosis module for drilling risks based on all samples and one-key update. The specific implementation process can refer to Figure 9 as shown, including the following content.

[0195] In this scenario example, considering that in complex geological conditions such as deep and ultra-deep layers (for example, complex geological environment scenarios), dangerous events such as overflow become risks that cannot be ignored during the drilling process. Due to the uncertainty of deep and deep-water geological conditions, the occurrence frequency and impact of risks such as overflow gradually increase with the increase in depth, bringing huge challenges to oil and gas exploration and development. Currently, artificial intelligence models have been preliminarily applied in the field of oil and gas drilling, especially in the field of risk detection. However, intelligent models are often too sensitive in actual applications, resulting in a large number of false alarms. Facing this challenge, how to more effectively deal with a large number of false alarms and effectively utilize this false alarm information has become an urgent problem to be solved. The model real-time update method provides a possibility to solve this problem.

[0196] Also considering negative samples (for example, negative sample data), that is, cases where the model prediction fails. After being screened by the failed model prediction, their value is proven. Using negative samples to retrain a new model and then integrating it into the system can ensure that the intelligent model is more suitable for the current well conditions and effectively reduce the false alarm rate. Through improvement, the accuracy of risk diagnosis can be effectively improved.

[0197] Based on the above situation, after creative thinking, the applicant proposed an idea for utilizing negative samples to improve the diagnosis accuracy of drilling risks, realize the real-time update of the intelligent model, and then provide an advanced and feasible solution for the industry, improve the safety and efficiency of the drilling process, and provide technical support for safe and efficient drilling.

[0198] Refer to Figure 9As shown, first, a conventional time-series supervised model (e.g., the first type of detection model) is trained using risk segment data (e.g., the first training dataset), and the model parameters are adjusted to perform well on the current dataset. Then, a time-series unsupervised model (e.g., the second type of detection model) is trained using data from the normal drilling phase (e.g., the second training dataset), which can identify the patterns of normal data and determine whether the data at a certain moment is normal. Once an anomaly is determined, the system will issue a risk alarm. Then, through a voting mechanism, three models (including: the first type of detection model, the second type of detection model, and the third type of detection model) are integrated to construct a total risk overflow intelligent model (e.g., the initial combined risk prediction model). Finally, for the samples with false alarms in the actual application of the model (e.g., the negative sample data with prediction failures obtained based on the mud logging data of the target well), these negative samples are selected and calibrated (e.g., constructing the fourth training dataset) and reused to train the supervised time-series model (e.g., the third type of detection model). Since these samples prove their confusing nature to the model, retraining can improve the accuracy of the model. At the same time, for the normal supervised-unsupervised time-series model, the steps of conventional incremental update are performed. Finally, through the voting mechanism, multiple models are integrated to further improve the reliability and prediction accuracy of the model.

[0199] Specifically, it may include the following multiple steps:

[0200] Step 1: Optimize the input features (e.g., key features), perform data preprocessing on the mud logging data, and form a data processing flow including outlier detection, interpolation, and smoothing filtering.

[0201] Before constructing the risk diagnosis model, the primary task is to optimize the input features and perform data preprocessing. In this embodiment, comprehensive mud logging time-series data (e.g., the sample mud logging data of the sample well) of more than 140 overflow events that occurred during the drilling operations of more than 130 wells in seven major oil regions of a certain area are widely collected and sorted out. These data cover detailed records of various fluid intrusions into the formation such as oil, gas, and water, providing valuable information resources for in-depth analysis of overflow events.

[0202] When initially analyzing the data, the applicant found that there was no single feature that had a significant correlation with whether an overflow occurred, as shown in the Figure 4 correlation calculation results shown. Therefore, on the premise of ensuring that the model can be quickly and effectively applied at the drilling site, the following key features are carefully selected as the inputs for model training: riser pressure, inlet flow rate, outlet flow rate, total pit volume, and total hydrocarbon content.

[0203] Among them, the riser pressure is taken into consideration because pressure changes are an important indicator for monitoring downhole dynamics during drilling operations. When overflow occurs, the drilling fluid begins to flow freely and the balance of the static column is broken. As the drilling fluid moves upward, the height of the static column decreases, and the bottomhole pressure decreases accordingly. Inside the drill bit, the impact of fluid intrusion is small, and the static column pressure and circulating pressure loss remain almost unchanged. Therefore, a decrease in bottomhole pressure will lead to a decrease in riser pressure, and this change is of great significance for predicting overflow.

[0204] Changes in inlet and outlet flow rates can also reflect the occurrence of overflow risks. Under normal circumstances, inlet and outlet flow rates should remain relatively stable. However, when overflow occurs, the outlet flow rate will increase without increasing the inlet flow rate, which is an important feature for identifying overflow.

[0205] As a key parameter describing the drilling fluid reserve, the total pool volume also has an important impact on the prediction of overflow. As the drilling fluid returns, the total pool volume will gradually increase. In the case of oil and water invasion, as the outlet flow rate increases, the total pool volume will also increase accordingly; in the case of gas invasion, due to the expansion of bubbles, the outlet flow rate increases, and the total pool volume will also increase. This change will appear soon after the oil and water invasion and gas invasion occur, providing important clues for predicting overflow.

[0206] The change in total hydrocarbon content is related to the gas composition in the drilling fluid. When gas invasion occurs, the total hydrocarbon content will increase as the gas circulates to the surface. This feature is of great significance for guiding field staff to detect overflow events in a timely manner. Therefore, incorporating total hydrocarbon content into the model input features will help improve the model's ability to identify gas invasion and overflow.

[0207] After selecting these key features, we performed a data cleaning step to reduce the impact of noise, outliers, and missing values. Data cleaning is a crucial part of machine learning pre-processing, which can improve data quality and make the data more suitable for subsequent analysis and modeling. In this example, we implemented outlier detection, missing value processing, data smoothing, and normalization by writing code, laying a solid foundation for subsequent model training.

[0208] Step 2: Sample balancing and normal data extraction.

[0209] Before model training, sample balance and normal data extraction are equally important. In this example, the data is divided into 60% training set, 20% validation set and 20% test set. In order to verify the accuracy of the model on different wells, a cross-well data partitioning strategy is adopted here to ensure that the training set and the test set use data from different wells. Doing so can improve the generalization ability of the model and enable it to better adapt to unseen well conditions.

[0210] In the anomaly detection task, the scarcity of anomaly data is a common problem. When the normal data is far greater than the anomaly data, supervised models may become inert and tend to predict that all data is normal. To avoid this problem, it is necessary to balance the quantity of normal data and anomaly data here. In this embodiment, for the supervised model dataset, the start time and duration of the overflow data occurrence are found, and normal data of the same duration is collected from the start time backwards (the first training dataset is constructed) to achieve the purpose of balancing the data.

[0211] Meanwhile, to ensure that the data during a large number of normal drilling operations can be fully utilized, all the data with the risk label of not occurred in the training dataset is also separately extracted as the training dataset of the unsupervised model (for example, the second training dataset is constructed). This can ensure that the unsupervised model can learn the patterns of normal drilling data and thus better identify anomaly data.

[0212] In the validation and testing phases, the unsupervised model uses the same validation set and test set as the supervised model for accurate comparison and validation of the model performance. This step provides strong data support for in-depth analysis of the overflow event and ensures the effectiveness of model training and the accuracy of evaluation.

[0213] Step 3: Establish an intelligent identification model for overflow risk.

[0214] After data preprocessing and sample balancing, it enters the model establishment phase. In this embodiment, LSTM (Long Short-Term Memory Network) and GRU (Gated Recurrent Unit) are selected as the supervised time series models for training, and LSTM-AE (Bidirectional Long Short-Term Memory Network) is selected as the unsupervised time series model for training. These models have significant advantages in processing time series data and can capture the time series dependencies in the data.

[0215] Taking the dataset composed of 59 wells in this example as an example, two random sampling operations are performed. Each time, the data of 35 wells is sampled to form two independent sub-datasets. These two sub-datasets are respectively used to train the two supervised time series models of GRU and LSTM. Due to the sampling with replacement feature during the sampling process, each subset may contain duplicate samples and may omit some original samples. This processing method increases the diversity of the model and helps to improve the generalization ability of the model.

[0216] During the model training process, the majority voting method (e.g., voting rules) is adopted to integrate the results of three models. This method votes on the prediction results of multiple models and takes the majority opinion as the final prediction result. In the Bagging integration method, no additional hyperparameters need to be introduced, mainly relying on the hyperparameters of each base learner itself. By optimizing these hyperparameters, the model performance is improved to achieve better prediction effects.

[0217] After training the unsupervised time series model with normal drilling data, the three models are integrated through a hard voting mechanism. This step significantly improves the prediction accuracy of the model, from 69% to 75%. This improvement indicates that by integrating the results of multiple models, the advantages of each model can be fully utilized to improve the overall prediction performance.

[0218] Step 4: Construct a knowledge graph based on expert experience to extract negative samples.

[0219] To improve the accuracy and reliability of the model, the extraction and selection of negative samples are crucial. In this embodiment, a method for selecting negative samples based on expert experience is also proposed, aiming to automatically intercept samples and ensure the quality of negative samples.

[0220] First, the cases where the intelligent model prediction fails can be used as the main source of negative samples. These negative samples usually have a certain degree of ambiguity and complexity, but they have the value of selection and analysis. By collecting these cases, a large amount of negative sample data for model optimization can be obtained.

[0221] Secondly, with the help of expert experience, a knowledge graph is established to conduct in-depth secondary analysis of these samples. A knowledge graph is a graph-based data structure that can intuitively display the relationships and rules between data. By constructing a knowledge graph, the specific reasons for the model to determine negative samples can be mined, including but not limited to the specific reasons for the model to determine negative samples and the root causes of data anomalies. This process helps to more deeply understand the performance and limitations of the model and provides strong support for subsequent optimization.

[0222] Finally, the negatively sampled data after secondary screening will be stored in a dedicated negative sample database. This database, as an important data resource for subsequent model optimization, can be used to train new models or update existing models to improve the accuracy and reliability of the models.

[0223] Step 5: Automatically update the intelligent model based on positive-negative sample pairs.

[0224] To make full use of negative samples and further improve the model performance, an automatic update method for the intelligent model based on positive-negative sample pairs is proposed. This method aims to continuously update the model to adapt to the new data environment and improve the accuracy of the model.

[0225] Specifically, for each of the two supervised models, select a model with exactly the same structure and hyperparameters as the backup model. During the actual application of the model, when new negative samples are generated, use the trained model and the constructed knowledge graph to label and classify these negative samples. Then, use these labeled negative samples to train the backup model. At the same time, for the existing three models, based on the incremental learning technology, use the samples to normally update the model. By continuously iterating and updating the backup model, the model performance can be gradually optimized and the accuracy of the model can be improved.

[0226] Finally, integrate multiple models (including the original two supervised models (LSTM and GRU), an unsupervised model (Bi-LSTM), and two backup models (LSTM and GRU)). By comparing and analyzing the prediction results of different models, the model performance can be further verified and optimized.

[0227] For specific details, please refer to Figure 10 As shown in the result comparison chart, the model after automatic update has a significant improvement in prediction accuracy. This result fully demonstrates the effectiveness and feasibility of the method proposed in the present invention.

[0228] Through the above scenario example, it is verified that the detection method of drilling risk based on the full sample provided in this specification can achieve a breakthrough in the intelligent monitoring technology of drilling risk under the full sample application. And it specifically has the following beneficial effects: 1) Improve the model prediction accuracy. This method uses the full sample intelligent model monitoring, which not only relies on a small amount of risk data, but also makes full use of a large amount of normal drilling data to train the unsupervised time series model. At the same time, by combining multiple supervised and unsupervised time series models and using the ensemble learning method, the accuracy and reliability of risk identification are significantly improved. 2) Ensure the real-time performance of the model. By introducing the negative sample update method and the incremental learning method, the model can be dynamically adjusted and optimized according to real-time data during the actual application process. Thus, in the face of changing environments and conditions, the model can respond quickly, improving the accuracy and immediacy of the prediction of downhole conditions.

[0229] Although this specification provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way among many possible execution orders and does not represent the only execution order. When an actual device or client product is executed, it may be executed in the order of the method shown in the embodiments or the drawings, or in parallel (e.g., in an environment with parallel processors or multi-threaded processing, or even in a distributed data processing environment). The terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, product or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, product or device. Without further limitation, there is no exclusion of additional identical or equivalent elements in the process, method, product or device comprising the said elements. The terms such as first, second, etc. are used to denote names and do not denote any particular order.

[0230] As is also known to those skilled in the art, in addition to implementing the controller in the form of pure computer-readable program code, the method steps can be logically programmed to enable the controller to implement the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Therefore, such a controller can be regarded as a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.

[0231] This specification can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. This specification can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer-readable storage media including storage devices.

[0232] As can be seen from the description of the above embodiments, those skilled in the art can clearly understand that this specification can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of this specification can essentially be embodied in the form of a software product, and this computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments of this specification.

[0233] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. This specification can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on.

[0234] Although this specification is depicted through embodiments, those of ordinary skill in the art know that this specification has many variations and changes without departing from the spirit of this specification. It is hoped that the appended claims will cover these variations and changes without departing from the spirit of this specification.

Claims

1. A detection method for drilling risks based on all samples, characterized in that, Including: Obtaining the logging data of the target well in the current time period; Extracting target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features; Processing the target valid data by using the target joint risk prediction model to obtain the corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first-class detection model, a target second-class detection model, and a target third-class detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of the sample well and the logging data of the target well where the model prediction fails during the drilling process of the target well; Determining whether there is a drilling risk in the target well currently according to the target prediction result; Wherein, the target joint risk prediction model is trained according to the preset training rule in the following manner: obtaining the sample logging data of the sample well; and constructing a first training data set and a second training data set by using the sample logging data of the sample well according to the first sample processing rule; constructing an initial first-class detection model and an initial second-class detection model; wherein, the initial first-class detection model at least includes a parallel LSTM structure and GRU structure, and the initial second-class detection model at least includes an LSTM-AE structure; performing supervised learning on the initial first-class detection model by using the first training data set to obtain an intermediate first-class detection model; performing unsupervised learning on the initial second-class detection model by using the second training data set to obtain an intermediate second-class detection model; constructing a corresponding intermediate third-class detection model according to the intermediate first-class detection model; wherein, the intermediate third-class detection model has the same model structure and model parameters as the intermediate first-class detection model; combining the intermediate first-class detection model, the intermediate second-class detection model, and the intermediate third-class detection model to obtain an initial joint risk prediction model; obtaining the logging data of the target well collected during the drilling process of the target well; and constructing a third training data set and a fourth training data set by using the logging data of the target well according to the second sample processing rule; wherein, the fourth training data set is generated based on the logging data of the target well where the initial joint risk prediction model fails in prediction; performing incremental learning on the intermediate first-class detection model and the intermediate second-class detection model in the initial joint risk prediction model by using the third training data set; at the same time, performing negative sample learning on the intermediate third-class detection model in the initial joint risk prediction model by using the fourth training data set to obtain a target joint risk prediction model that meets the requirements; According to the first sample processing rule, using the sample logging data of the sample well, the first training data set and the second training data set are constructed, including: according to the first sample processing rule, obtaining and based on the historical logging data of drilling accidents, through correlation analysis, determining the key features associated with drilling accidents; according to the key features, extracting sample valid data from the sample logging data of the sample well; according to the sample valid data, screening out the sample valid data in the first time period when a drilling accident occurs and the sample valid data in the second time period before a drilling accident occurs, and constructing the first training data set; and according to the sample valid data, screening out the sample valid data without drilling accidents and constructing the second training data set; According to the second sample processing rule, using the logging data of the target well, the third training data set and the fourth training data set are constructed, including: using the initial joint risk prediction model to process the logging data of the target well to obtain the corresponding prediction results; randomly extracting and combining valid data from the logging data of the target well to construct the third training data set; according to the prediction results, screening out the logging data for which the model prediction fails from the logging data of the target well; and based on the logging data for which the model prediction fails, extracting valid data as negative sample data to construct the fourth training data set.

2. The method according to claim 1, wherein The key features include at least one of the following: riser pressure, inlet flow rate, outlet flow rate, total pit volume, total hydrocarbon content.

3. The method according to claim 1, wherein After constructing the first training data set and the second training data set, the method further includes: According to the first sample processing rule, determining the sample well involved in the sample valid data in the first training data set as the training sample well; Screening out the sample valid data that does not involve the training sample well from the sample valid data to construct the test data set and the validation data set for the first type of detection model and the second type of detection model; wherein, the test data set of the first type of detection model is the same as the test data set of the second type of detection model, and the validation data set of the first type of detection model is the same as the validation data set of the second type of detection model.

4. The method according to claim 1, wherein The method further includes: Obtaining a knowledge graph of the logging data during the drilling process; According to the knowledge graph and the negative sample data, determining the failure reasons for the model prediction failure; According to the failure reasons, determining the data quality evaluation parameters of the negative sample data; According to the data quality evaluation parameters, screening out the qualified negative sample data from the negative sample data to construct the fourth training data set.

5. A detection device for drilling risks based on all samples, characterized in that, Including: An acquisition module for acquiring the logging data of the target well in the current time period; An extraction module for extracting the target valid data that meets the requirements from the logging data in the current time period; wherein, the target valid data at least includes key features; A processing module, configured to process target valid data by using a target joint risk prediction model to obtain a corresponding target prediction result; wherein, the target joint risk prediction model at least includes: a target first type detection model, a target second type detection model, and a target third type detection model; the target joint risk prediction model is trained according to a preset training rule by using the sample logging data of a sample well and the logging data of a target well where model prediction fails during the drilling process of the target well; A determination module, configured to determine whether there is a drilling risk in the target well currently according to the target prediction result; Wherein, the target joint risk prediction model is trained according to the preset training rule in the following manner: Obtain the sample logging data of the sample well; and according to the first sample processing rule, use the sample logging data of the sample well to construct a first training data set and a second training data set; Construct an initial first type detection model and an initial second type detection model; wherein, the initial first type detection model at least includes a parallel LSTM structure and GRU structure, and the initial second type detection model at least includes an LSTM-AE structure; Perform supervised learning on the initial first type detection model by using the first training data set to obtain an intermediate first type detection model; Perform unsupervised learning on the initial second type detection model by using the second training data set to obtain an intermediate second type detection model; Construct a corresponding intermediate third type detection model according to the intermediate first type detection model; wherein, the intermediate third type detection model has the same model structure and model parameters as the intermediate first type detection model; Combine the intermediate first type detection model, the intermediate second type detection model, and the intermediate third type detection model to obtain an initial joint risk prediction model; Obtain the logging data of the target well collected during the drilling process of the target well; and according to the second sample processing rule, use the logging data of the target well to construct a third training data set and a fourth training data set; wherein, the fourth training data set is generated based on the logging data of the target well where the initial joint risk prediction model fails to predict; Perform incremental learning on the intermediate first type detection model and the intermediate second type detection model in the initial joint risk prediction model by using the third training data set; At the same time, perform negative sample learning on the intermediate third type detection model in the initial joint risk prediction model by using the fourth training data set to obtain a target joint risk prediction model that meets the requirements; According to the first sample processing rule, using the sample logging data of the sample well, the first training data set and the second training data set are constructed, including: according to the first sample processing rule, obtaining and based on the historical logging data of drilling accidents, determining the key features associated with drilling accidents through correlation analysis; extracting sample valid data from the sample logging data of the sample well according to the key features; screening out the sample valid data in the first time period when drilling accidents occur and the sample valid data in the second time period before drilling accidents occur based on the sample valid data to construct the first training data set; and screening out the sample valid data without drilling accidents based on the sample valid data to construct the second training data set. According to the second sample processing rule, using the logging data of the target well, the third training data set and the fourth training data set are constructed, including: processing the logging data of the target well using the initial joint risk prediction model to obtain the corresponding prediction results; randomly extracting and combining valid data from the logging data of the target well to construct the third training data set; screening out the logging data for which the model prediction fails from the logging data of the target well according to the prediction results; and extracting valid data based on the logging data for which the model prediction fails as negative sample data to construct the fourth training data set.

6. A computer device, characterized in that, It includes a processor and a memory for storing processor-executable instructions, and when the processor executes the instructions, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium, characterized in that, Computer instructions are stored thereon, and when the instructions are executed by the processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images