A device and method for real-time monitoring of deep water overflow

Abstract

The present invention discloses a device and method for real-time monitoring of deepwater overflow, comprising: a data acquisition module: responsible for real-time collection of various data in the drilling process; a data transmission module: responsible for real-time transmission of the data collected by the probe sensor to a data processing center; a data processing center: receiving and processing the data transmitted by the probe sensor, storing, analyzing and displaying it; an early warning and decision-making module: according to the analysis results of the data processing center, when overflow or abnormal water quality is monitored, a warning message is issued; in the present invention, the intelligent recognition module utilizes transfer learning technology to transfer the learned knowledge to new working condition recognition tasks, thereby improving the model's adaptability to new working conditions and recognition accuracy. By constructing an intelligent recognition model, the system can automatically determine whether the drilling data indicates the existence of overflow risk, thereby reducing the reliance on manual judgment and improving the intelligence level of decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of deepwater overflow identification, and in particular to a device and method for real-time monitoring of deepwater overflow. Background Art

[0002] Intelligent identification of downhole overflow is to use deepwater drilling history and real-time data, combined with machine learning algorithms, to carry out research on data-driven working condition identification models. Drilling is a key link in deepwater oil and gas exploration and development. Compared with land and shallow water drilling, deepwater drilling is characterized by high technology, high investment, and high risk, and faces many difficulties and challenges. Among them, deepwater overflow occurs frequently and takes a long time to handle, which seriously limits the drilling timeliness and is particularly harmful. Once timely warning and control are not given, the downhole overflow condition will further deteriorate and eventually develop into a major accident of uncontrolled blowout, causing huge losses. Timely and accurate identification of downhole overflow will help reduce well control risks and prevent major accidents such as blowout, which is of great significance to promoting the development of deepwater oil and gas resources in my country.

[0003] In the prior art, automatic recognition of six common drilling conditions, including pulling out of the hole, drilling down, circulation, drilling, reaming and single-root connection, is achieved. However, in the actual drilling process, more complex and changeable conditions may be encountered. These conditions may be beyond the current recognition range of the model, resulting in deviation or inaccuracy in the recognition results. Therefore, a device and method for real-time monitoring of deepwater overflow are proposed. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the prior art and realize the automatic identification of six common drilling conditions, such as pulling out of the hole, drilling down, circulation, drilling, reaming and single root connection. However, in the actual drilling process, more complex and changeable conditions may be encountered. These conditions may exceed the scope that the model can currently identify, thereby causing deviations or inaccuracies in the identification results. A device and method for real-time monitoring of deepwater overflow are proposed.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A device for real-time monitoring of deep water overflow, comprising:

[0007] Data acquisition module: responsible for real-time collection of various data during the drilling process, including but not limited to key parameters such as pressure, temperature, flow rate, etc.;

[0008] Data transmission module: responsible for transmitting the data collected by the probe sensor to the data processing center in real time;

[0009] Data processing center: receives and processes data transmitted by probe sensors, and performs storage, analysis and display;

[0010] Early warning and decision-making module: According to the analysis results of the data processing center, when overflow or abnormal water quality is detected, early warning information will be issued and sent to relevant personnel via SMS, email, phone, etc.;

[0011] The data collected by the probe sensor module is transmitted to the data processing center in real time through the data transmission module. After the data transmission module transmits the collected data to the data processing center, the data processing center will store, analyze and process the data. The data processing center includes a data processing module and an intelligent recognition module. The data processing module is responsible for preprocessing and analyzing the collected data. The intelligent recognition module constructs an intelligent recognition model to determine whether the drilling data indicates the existence of overflow risk. The intelligent recognition module uses transfer learning to transfer the learned knowledge to the new working condition identification task. After completing the data analysis, the data processing center will pass the analysis results to the early warning and decision-making module.

[0012] The above technical solution further includes:

[0013] Furthermore, the data acquisition module includes a probe sensor unit, a data receiving unit and a calibration and verification unit. The probe sensor unit is used to measure various parameters downhole, such as drilling pressure, bottom hole mud flow, injection pressure of the drill nozzle, injection speed, drill torque, etc. The data receiving unit is responsible for receiving the signal output by the probe sensor and converting the signal output by the probe sensor into a recognizable data format. The calibration and verification unit regularly calibrates the probe sensor and verifies the data accuracy of the data receiving unit. The probe sensor unit transmits the collected real-time data (such as pressure, temperature, flow, etc.) to the data receiving unit.

[0014] Furthermore, the data transmission module converts the data collected by the probe sensor into an electrical signal and generates a pulse signal in a coded manner. The pulse signal changes the extension and retraction of the control pulse generator, thereby changing the mud flow area between the mushroom head and the current limiting ring. When the mushroom head retracts, the mud in the drill string can pass smoothly through the current limiting ring; when the mushroom head rises, the mud flow area is reduced, thereby generating a positive mud pressure pulse in the drill string. The mud pressure pulse signal carries the information of the downhole parameters and is transmitted to the data processing center.

[0015] Furthermore, the data processing module includes a data processing unit, a data storage unit, a data monitoring unit and a data output unit. The data processing unit contains a data cleaning subunit, a data interpolation subunit and a filtering subunit. The data cleaning subunit first processes the data to remove outliers and fill in missing values. The data interpolation subunit interpolates missing data points. The filtering subunit filters the data to remove noise. The processed data is transmitted to the data storage unit. The data storage unit is responsible for storing original data and processed data and providing a data access interface for query and use by other units. The data monitoring unit monitors the performance and errors in the data processing process and reports data quality issues, such as the number of outliers or missing values. The data output unit outputs the processed data to the intelligent recognition module in an appropriate format.

[0016] Furthermore, the intelligent recognition module includes a model training unit, a model testing and optimization unit, a working condition data collection and processing unit, a transfer learning unit and a recognition accuracy improvement unit. The model training unit is responsible for designing and training the intelligent recognition model. The model training unit uses the processed data to perform model training and outputs the trained model. The model testing and optimization unit is responsible for testing the trained model, including performance testing under different working conditions, tuning the model according to the test results, and outputting the optimized model. The working condition data collection and processing unit is responsible for collecting new drilling working condition data, processing the new data, and adding it to the model training set. The transfer learning unit is responsible for using transfer learning technology to transfer the learned knowledge to the new working condition recognition task, and output the model after transfer learning to improve the model's adaptability to the new working condition. The recognition accuracy improvement unit is responsible for more refined tuning and training of the model to improve recognition accuracy. The recognition accuracy improvement unit uses multi-source heterogeneous data fusion to improve the input data quality and feature expression capabilities.

[0017] Furthermore, the data processing module transmits the processed data to the model training unit, and the model training unit uses these data to perform model training. The model training unit outputs the trained model to the model testing and optimization unit, and the model testing and optimization unit tests the model and optimizes it according to the test results. The operating condition data collection and processing unit collects and processes new data, and then adds it to the model training set. The model training unit uses the updated training set to perform model training. The transfer learning unit uses transfer learning technology to output the model after transfer learning. The model after transfer learning is used as the initial model of the model training unit, or is used for further testing and optimization by the model testing and optimization unit. The recognition accuracy improvement unit performs more refined tuning and training on the model, and the optimized model is returned to the model training unit for training, or is directly tested by the model testing and optimization unit.

[0018] Furthermore, when the intelligent identification module detects a potential overflow risk by analyzing the drilling operation data, the early warning and decision-making module responds quickly, and the early warning and decision-making module displays detailed early warning information, including the overflow risk level, possible impact range, and recommended response measures. The early warning and decision-making module records detailed information of each early warning, including the warning time, risk level, response measures, etc.

[0019] A method for using a device for real-time monitoring of deep water overflow comprises the following steps:

[0020] S1: Data collection: The device will automatically collect drilling data in real time from the probe sensor installed on the drilling platform;

[0021] S2: Data transmission: The data transmission module transmits the collected data to the data processing center;

[0022] S3: Feature extraction and selection: The data processing center extracts features related to overflow risk from the received data;

[0023] S4: Intelligent recognition model construction: construct an intelligent recognition model, which takes the input features as input, outputs the overflow risk level, and determines whether there is an overflow phenomenon or abnormal water quality;

[0024] S5: Decision recommendation generation: Based on the output of the intelligent recognition model, decision recommendations are generated in combination with preset rules or strategies, which may be based on expert experience, industry standards or regulatory requirements.

[0025] Furthermore, in S4, the specific steps of constructing the intelligent recognition model and performing data analysis are:

[0026] Model construction and training: According to the characteristics and rules of drilling data, design an intelligent recognition model based on deep learning, use the processed data to train the model, and adjust the model parameters to improve the recognition accuracy;

[0027] Model testing and optimization: Input the test data set into the model, evaluate the recognition accuracy and performance of the model, and tune the model according to the test results, such as adjusting the model structure, adding training data, etc.

[0028] Expansion of working condition types and improvement of recognition accuracy: Continuously collect new drilling condition data, enrich the training data set, use transfer learning technology to transfer the learned knowledge to new working condition recognition tasks, and use technical means such as multi-source heterogeneous data fusion to improve the input data quality and feature expression ability of the model;

[0029] Real-time data recognition and risk warning: The real-time drilling data is input into the trained intelligent recognition model, which recognizes the real-time data and outputs the recognition results.

[0030] Furthermore, in the process of expanding the types of working conditions and improving the recognition accuracy, the transfer learning technology is used to transfer the learned knowledge to the new working condition recognition task. The specific steps are as follows:

[0031] Select source domain and target domain: Source domain: the trained drilling condition recognition model and its data; Target domain: new drilling condition and its data;

[0032] Data preprocessing: cleaning and normalizing new drilling condition data;

[0033] Model selection and fine-tuning: Select a deep learning model as the base model for transfer learning, and fine-tune the last layer or layers of the model according to the characteristics of the new drilling condition data;

[0034] Training and evaluation: Use new drilling condition data to train the model, and use indicators such as accuracy to evaluate the model's recognition performance under new conditions.

[0035] The present invention has the following beneficial effects:

[0036] In the present invention, the intelligent recognition module uses transfer learning technology to transfer the learned knowledge to new working condition recognition tasks, thereby improving the model's adaptability to new working conditions and recognition accuracy. By constructing an intelligent recognition model, the system can automatically determine whether the drilling data indicates the existence of overflow risk, reducing reliance on manual judgment and improving the level of intelligent decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A system block diagram of a device for real-time monitoring of deep water overflow proposed by the present invention;

[0038] Figure 2 A flow chart of a method for using a real-time monitoring deepwater overflow device proposed by the present invention;

[0039] Figure 3 This is a flow chart for expanding the types of working conditions and improving recognition accuracy in the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] The present invention provides Figure 1-Figure 3 A device for real-time monitoring of deep water overflow is shown, comprising:

[0042] Data acquisition module: responsible for real-time collection of various data during the drilling process, including but not limited to key parameters such as pressure, temperature, flow rate, etc.;

[0043] Data transmission module: responsible for transmitting the data collected by the probe sensor to the data processing center in real time;

[0044] Data processing center: receives and processes data transmitted by probe sensors, and performs storage, analysis and display;

[0045] Early warning and decision-making module: According to the analysis results of the data processing center, when overflow or abnormal water quality is detected, early warning information will be issued and sent to relevant personnel via SMS, email, phone, etc.;

[0046] The data collected by the probe sensor module is transmitted to the data processing center in real time through the data transmission module. After the data transmission module transmits the collected data to the data processing center, the data processing center will store, analyze and process the data. The data processing center includes a data processing module and an intelligent recognition module. The data processing module is responsible for preprocessing and analyzing the collected data. The intelligent recognition module constructs an intelligent recognition model to determine whether the drilling data indicates the existence of overflow risk. The intelligent recognition module uses transfer learning to transfer the learned knowledge to the new working condition identification task. After completing the data analysis, the data processing center will pass the analysis results to the early warning and decision-making module.

[0047] In one embodiment, for the above-mentioned data acquisition module, the data acquisition module includes a probe sensor unit, a data receiving unit and a calibration and verification unit. The probe sensor unit is used to monitor various parameters in the drilling process in real time. The data receiving unit is responsible for receiving the signal output by the probe sensor and converting the signal output by the probe sensor into a recognizable data format. The calibration and verification unit regularly calibrates the probe sensor and verifies the data accuracy of the data receiving unit. The probe sensor unit transmits the collected real-time data (such as pressure, temperature, flow, etc.) to the data receiving unit.

[0048] In one embodiment, for the above-mentioned data transmission module, the data transmission module converts the data collected by the probe sensor into an electrical signal, and generates a pulse signal in a coded manner. The pulse signal changes with the extension and retraction of the control pulse generator, thereby changing the mud flow area between the mushroom head and the current limiting ring. When the mushroom head retracts, the mud in the drill string can pass smoothly through the current limiting ring; and when the mushroom head rises, the mud flow area is reduced, thereby generating a positive mud pressure pulse in the drill string. The mud pressure pulse signal carries the information of the downhole parameters and is transmitted to the data processing center.

[0049] In one embodiment, for the above-mentioned data processing module, the data processing module includes a data processing unit, a data storage unit, a data monitoring unit and a data output unit. The data processing unit includes a data cleaning subunit, a data interpolation subunit and a filtering subunit. The data cleaning subunit first processes the data to remove outliers and fill in missing values. The data interpolation subunit interpolates missing data points. The filtering subunit filters the data to remove noise. The processed data is transmitted to the data storage unit. The data storage unit is responsible for storing the original data and processed data, and provides a data access interface for other units to query and use. The data monitoring unit monitors the performance and errors in the data processing process and reports data quality issues, such as the number of outliers or missing values. The data output unit outputs the processed data to the intelligent recognition module in an appropriate format.

[0050] In one embodiment, for the above-mentioned intelligent recognition module, the intelligent recognition module includes a model training unit, a model testing and optimization unit, a working condition data collection and processing unit, a transfer learning unit and a recognition accuracy improvement unit. The model training unit is responsible for designing and training the intelligent recognition model. The model training unit uses the processed data to perform model training and outputs the trained model. The model testing and optimization unit is responsible for testing the trained model, including performance testing under different working conditions, tuning the model according to the test results, and outputting the optimized model. The working condition data collection and processing unit is responsible for collecting new drilling working condition data, processing the new data, and adding it to the model training set. The transfer learning unit is responsible for using transfer learning technology to transfer the learned knowledge to the new working condition recognition task, and output the model after transfer learning to improve the model's adaptability to the new working condition. The recognition accuracy improvement unit is responsible for more refined tuning and training of the model to improve recognition accuracy. The recognition accuracy improvement unit uses multi-source heterogeneous data fusion to improve the input data quality and feature expression capabilities.

[0051] In one embodiment, for the above-mentioned data processing module, the data processing module transmits the processed data to the model training unit, the model training unit uses these data to perform model training, the model training unit outputs the trained model to the model testing and optimization unit, the model testing and optimization unit tests the model and optimizes it according to the test results, the operating condition data collection and processing unit collects and processes new data, and then adds it to the model training set, the model training unit uses the updated training set to perform model training, the transfer learning unit uses the transfer learning technology to output the model after transfer learning, the model after transfer learning is used as the initial model of the model training unit, or is used for further testing and optimization by the model testing and optimization unit, the recognition accuracy improvement unit performs more refined tuning and training on the model, and the optimized model is returned to the model training unit for training, or is directly tested to the model testing and optimization unit.

[0052] In one embodiment, for the above-mentioned intelligent identification module, when the intelligent identification module detects a potential overflow risk by analyzing the drilling operation data, the early warning and decision module responds quickly, and the module early warning and decision module displays detailed early warning information, including the overflow risk level, possible impact range, and recommended response measures. The early warning and decision module records detailed information of each early warning, including the warning time, risk level, response measures, etc.

[0053] A method for using a device for real-time monitoring of deep water overflow comprises the following steps:

[0054] S1: Data collection: The device will automatically collect drilling data in real time from the probe sensor installed on the drilling platform;

[0055] S2: Data transmission: The data transmission module transmits the collected data to the data processing center;

[0056] S3: Feature extraction and selection: The data processing center extracts features related to overflow risk from the received data;

[0057] S4: Intelligent recognition model construction: Build an intelligent recognition model, which takes the input features as input, outputs the overflow risk level, and determines whether there is overflow phenomenon or water quality abnormality;

[0058] S5: Decision recommendation generation: Based on the output of the intelligent recognition model, decision recommendations are generated in combination with preset rules or strategies, which may be based on expert experience, industry standards or regulatory requirements.

[0059] In one embodiment, for the above S4, in S4, the specific steps of constructing the intelligent recognition model and performing data analysis are:

[0060] Model construction and training: According to the characteristics and rules of drilling data, design an intelligent recognition model based on deep learning, use the processed data to train the model, and adjust the model parameters to improve the recognition accuracy;

[0061] Design a drilling data recognition model based on deep learning, including input layer, hidden layer and output layer. Use preprocessed drilling data to train the model, and adjust parameters such as learning rate and number of iterations to enable the model to accurately identify the features in the drilling data;

[0062] Model testing and optimization: Input the test data set into the model, evaluate the recognition accuracy and performance of the model, and tune the model according to the test results, such as adjusting the model structure, adding training data, etc.

[0063] Input a set of test data sets into the trained model to evaluate the recognition accuracy of the model. If it is found that the recognition accuracy of the model is low under certain working conditions, the model can be optimized by adding training data for related working conditions, adjusting the model structure, etc.

[0064] Expansion of working condition types and improvement of recognition accuracy: Continuously collect new drilling condition data, enrich the training data set, use transfer learning technology to transfer the learned knowledge to new working condition recognition tasks, and use technical means such as multi-source heterogeneous data fusion to improve the input data quality and feature expression ability of the model;

[0065] With the continuous development of drilling technology, new drilling conditions have emerged. In order to identify these new conditions, relevant new condition data can be collected, and the learned knowledge can be transferred to the new condition identification task using transfer learning technology. At the same time, by integrating multi-source heterogeneous data (such as geological data, drilling equipment data, etc.), the input data quality and feature expression ability of the model are improved, and the recognition accuracy of the model is further improved;

[0066] Real-time data recognition and risk warning: Real-time drilling data is input into the trained intelligent recognition model, which recognizes the real-time data and outputs the recognition results.

[0067] In one embodiment, for the above-mentioned expansion of the types of working conditions and improvement of the recognition accuracy, in the expansion of the types of working conditions and improvement of the recognition accuracy, the specific steps of using the transfer learning technology to transfer the learned knowledge to the new working condition recognition task are as follows:

[0068] Select source domain and target domain: Source domain: the trained drilling condition recognition model and its data; Target domain: new drilling condition and its data;

[0069] Data preprocessing: cleaning and normalizing new drilling condition data;

[0070] Model selection and fine-tuning: Select a deep learning model as the base model for transfer learning, and fine-tune the last layer or layers of the model according to the characteristics of the new drilling condition data;

[0071] Training and evaluation: Use new drilling condition data to train the model, and use indicators such as accuracy to evaluate the model's recognition performance under new conditions.

[0072] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A device for real-time monitoring of deep water overflow, characterized in that: include: Data acquisition module: responsible for real-time collection of various data during the drilling process; Data transmission module: responsible for transmitting the data collected by the probe sensor to the data processing center in real time; Data processing center: receives and processes data transmitted by probe sensors, and performs storage, analysis and display; Early warning and decision-making module: Based on the analysis results of the data processing center, when overflow or abnormal water quality is detected, an early warning message is issued; The data collected by the probe sensor is transmitted to the data processing center in real time through the data transmission module. After the data transmission module transmits the collected data to the data processing center, the data processing center will store, analyze and process the data. The data processing center includes a data processing module and an intelligent recognition module. The data processing module is responsible for preprocessing and analyzing the collected data. The intelligent recognition module constructs an intelligent recognition model to determine whether the drilling data indicates the existence of overflow risk. The intelligent recognition module uses transfer learning to transfer the learned knowledge to the new working condition recognition task. After completing the data analysis, the data processing center will pass the analysis results to the early warning and decision-making module; The intelligent recognition module includes a feature extraction unit, a model training unit, a model testing and optimization unit, a working condition data collection and processing unit, a transfer learning unit and a recognition accuracy improvement unit. The feature extraction unit extracts features related to overflow risk from the received data and transmits the features related to overflow risk to the model training unit. The model training unit is responsible for designing and training the intelligent recognition model. The model training unit uses the processed data to train the model and outputs the trained model. The model testing and optimization unit is responsible for testing the trained model, including performance testing under different working conditions, tuning the model according to the test results, and outputting the optimized model. The working condition data collection and processing unit is responsible for collecting new drilling working condition data, processing the new data, and adding it to the model training set. The transfer learning unit is responsible for using the transfer learning technology to transfer the learned knowledge to the new working condition recognition task and output the model after transfer learning. The recognition accuracy improvement unit is responsible for more refined tuning and training of the model. The recognition accuracy improvement unit uses multi-source heterogeneous data fusion to improve the input data quality and feature expression ability. The data acquisition module includes a probe sensor unit, a data receiving unit and a calibration and verification unit. The probe sensor unit is used to monitor various parameters in the drilling process in real time. The data receiving unit is responsible for receiving the signal output by the probe sensor and converting the signal output by the probe sensor into a recognizable data format. The calibration and verification unit calibrates the probe sensor regularly and verifies the data accuracy of the data receiving unit. The probe sensor unit transmits the collected real-time data to the data receiving unit. The data transmission module converts the data collected by the probe sensor into an electrical signal and generates a pulse signal in a coded manner. The pulse signal changes the mud flow area between the mushroom head and the current limiting ring as the control pulse generator expands and contracts, thereby generating a positive mud pressure pulse in the drill string. The mud pressure pulse signal carries the information of the downhole parameters and is transmitted to the data processing center. The data processing module includes a data processing unit, a data storage unit, a data monitoring unit and a data output unit. The data processing unit includes a data cleaning subunit, a data interpolation subunit and a filtering subunit. The data cleaning subunit first processes the data to remove abnormal values ​​and fill missing values. The data interpolation subunit interpolates the missing data points. The filtering subunit filters the data to remove noise. The processed data is transmitted to the data storage unit. The data storage unit is responsible for storing the original data and the processed data. The data monitoring unit monitors the performance and errors in the data processing process. The data output unit outputs the processed data to the intelligent recognition module in an appropriate format. The data processing module transmits the processed data to the model training unit, and the model training unit uses these data to perform model training. The model training unit outputs the trained model to the model testing and optimization unit, and the model testing and optimization unit tests the model and optimizes it according to the test results. The operating condition data collection and processing unit collects and processes new data, and then adds it to the model training set. The model training unit uses the updated training set to perform model training. The transfer learning unit uses transfer learning technology to output the model after transfer learning. The model after transfer learning is used as the initial model of the model training unit, or is used for further testing and optimization by the model testing and optimization unit. The recognition accuracy improvement unit performs more refined tuning and training on the model, and the optimized model is returned to the model training unit for training, or is directly tested by the model testing and optimization unit.

2. A device for real-time monitoring of deep water overflow according to claim 1, characterized in that: When the intelligent identification module detects a potential overflow risk by analyzing the drilling operation data, the early warning and decision module responds quickly. The early warning and decision module displays detailed early warning information, including the overflow risk level, possible impact range, and recommended response measures. The early warning and decision module records the detailed information of each early warning.

3. A method for using a device for real-time monitoring of deepwater overflow according to any one of claims 1-2, characterized in that: The following steps are involved: S1: Data collection: The device will automatically collect drilling data in real time from the probe sensor installed on the drilling platform; S2: Data transmission: The data transmission module transmits the collected data to the data processing center; S3: Feature extraction and selection: The data processing center extracts features related to overflow risk from the received data; S4: Intelligent recognition model construction: construct an intelligent recognition model, which takes the input features as input, outputs the overflow risk level, and determines whether there is an overflow phenomenon or abnormal water quality; S5: Decision recommendation generation: Based on the output of the intelligent recognition model and combined with preset rules or strategies, decision recommendations are generated.

4. The method for using the device for real-time monitoring of deep water overflow according to claim 3, characterized in that: In S4, the specific steps of constructing the intelligent recognition model and performing data analysis are: Model construction and training: According to the characteristics and rules of drilling data, design an intelligent recognition model based on deep learning, use the processed data to train the model, and adjust the model parameters to improve the recognition accuracy; Model testing and optimization: Input the test data set into the model, evaluate the recognition accuracy and performance of the model, and tune the model according to the test results; Expansion of working condition types and improvement of recognition accuracy: Continuously collect new drilling condition data, enrich the training data set, use transfer learning technology to transfer the learned knowledge to new working condition recognition tasks, and use multi-source heterogeneous data fusion to improve the input data quality and feature expression ability of the model; Real-time data recognition and risk warning: The real-time drilling data is input into the trained intelligent recognition model, which recognizes the real-time data and outputs the recognition results.

5. The method for using the device for real-time monitoring of deep water overflow according to claim 4, characterized in that: In the process of expanding the types of working conditions and improving the recognition accuracy, the transfer learning technology is used to transfer the learned knowledge to the new working condition recognition task. The specific steps are as follows: Select source domain and target domain: Source domain: the trained drilling condition recognition model and its data; Target domain: new drilling condition and its data; Data preprocessing: cleaning and normalizing new drilling condition data; Model selection and fine-tuning: Select a deep learning model as the base model for transfer learning, and fine-tune the last layer or layers of the model according to the characteristics of the new drilling condition data; Training and evaluation: Use new drilling condition data to train the model and evaluate the model's recognition performance on the new conditions.

Images