Wellhead multimodal edge detection and closed loop control method, system, device and medium

Abstract

The application discloses a wellhead multi-modal edge detection and closed-loop control method, system, device and medium, relates to the technical field of safety monitoring and automatic control of oil drilling well control, and comprises the following steps: acquiring real-time data of visual, vibration and pressure sensors and performing time synchronization and preprocessing; extracting modal characteristic parameters by using a sliding time window and calculating a cross-modal consistency score by fusion; automatically updating a threshold baseline according to historical data and setting early warning threshold and accident threshold; entering an early warning state when the score exceeds the early warning threshold, and determining that well kick or overflow occurs if the score further exceeds the accident threshold, generating a control instruction and sending the control instruction to at least one of a blowout preventer, a choke valve and a mud circulating pump; and continuously acquiring target real-time data and generating a control log. The application can realize accurate early identification of well kick and automatic closed-loop control of well control equipment, and significantly improves response speed and reliability.

Description

Technical Field

[0001] This invention relates to the field of well control safety monitoring and automatic control technology for oil drilling, and particularly to wellhead multimodal edge detection and closed-loop control methods, systems, equipment and media. Background Technology

[0002] In oil and gas drilling, well kick (the uncontrolled entry of formation fluid into the wellbore) is a key precursor to blowout accidents. Preventing blowouts hinges on timely detection and suppression when a well kick is just beginning and still relatively weak. However, existing monitoring methods primarily rely on single signals such as mud tank level, inlet and outlet flow rates, and wellhead pressure. These methods are relatively slow and uncertain, often only being detected after the well kick has developed and intensified, leading to missed opportunities for optimal intervention. Furthermore, existing well control equipment (such as blowout preventers (BOPs) and choke valves) is typically triggered manually or through simple threshold controls, lacking automated collaborative control. In emergencies, human response may be untimely, and traditional monitoring systems upload data to remote centers for processing, which, limited by network bandwidth and latency, cannot meet the stringent real-time requirements of transient events like well kicks.

[0003] To address the aforementioned issues, existing technologies have proposed various improvement schemes. For example, Chinese patent CN120337077A discloses an intelligent management system for oil well data, which adopts a layered architecture of edge layer and platform layer. Through edge computing nodes combined with a lightweight LSTM model, it autonomously completes data cleaning, feature extraction, and initial anomaly screening at the well site, reducing the latency of centralized cloud processing. Chinese patent CN120579482B proposes a mine gas control method based on multimodal data. By fusing real-time sensor data, roadway point cloud data, and CFD airflow simulation grid data, it constructs a real-time roadway data cube and uses a multimodal depth causal model to predict the gas concentration field distribution, achieving a shift from passive monitoring to proactive prediction. Furthermore, Chinese patent CN120139792A discloses a monitoring method based on geothermal well water level gauges, employing a main and auxiliary monitoring component combining pressure sensors and fiber optic sensors, combined with an edge computing module to achieve localized data preprocessing and real-time monitoring.

[0004] However, the aforementioned existing technologies all have significant limitations and cannot be directly applied to the real-time detection and closed-loop control of well kicks in oil drilling. Specifically, CN120337077A mainly targets equipment failure prediction during oil well production. Its edge layer is only used for initial anomaly screening, and the final decision and model training still rely on the cloud platform layer. For transient events like well kicks that require millisecond-level response, there is still a delay, and the system does not involve direct linkage control of well control equipment such as blowout preventers. Although CN120579482B achieves multimodal data fusion and edge-side decision-making, it is designed for mine gas ventilation control scenarios, which are fundamentally different from the process requirements and safety requirements of oil drilling well control. Its ventilation network control method cannot be directly applied to the emergency control of wellhead blowout preventers and throttle valves. CN120139792A mainly targets static water level monitoring in geothermal wells. Although it uses a combination of multiple sensors, it lacks adaptive threshold judgment and graded early warning mechanisms for the dynamic characteristics of well kicks, and its control objects are limited to water pumps, failing to achieve emergency closed-loop control of key well control equipment at the wellhead. Therefore, existing technologies have failed to solve the core problems in well kick detection, such as response lag, insufficient reliability of single sensors, and lack of automated linkage of well control equipment.

[0005] In summary, there is an urgent need for a real-time monitoring system and method capable of fusing multimodal sensor data at the well site edge. This system should be able to directly perform unified time-based fusion and real-time analysis of multi-source data such as visual, vibration, and pressure data at the wellhead. It should achieve early and accurate identification of well kicks / overflows through an adaptive dual-threshold mechanism and automatically link well control equipment such as blowout preventers and choke valves to execute emergency responses, forming a closed-loop control link from detection to control. This would significantly improve the well control response speed and reliability, and make up for the shortcomings of existing technologies. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies. Specifically, it provides a method, system, equipment, and medium for wellhead multimodal edge detection and closed-loop control, as detailed below: 1) In a first aspect, the present invention provides a wellhead multimodal edge detection and closed-loop control method, the specific technical solution of which is as follows: S1, acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; S2, a sliding time window is used to extract feature parameters from each of the preprocessed modal data, and the extracted feature parameters are fused in a multimodal manner to calculate the cross-modal consistency score; S3 automatically updates the threshold baseline based on historical data and sets the warning threshold and accident threshold, where the accident threshold is greater than the warning threshold; S4, when the cross-modal consistency score exceeds the warning threshold, enter the warning state and continue to monitor. In the warning state, determine whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result. S5, when the judgment result is yes, it is determined that a well kick event or a overflow event has occurred, and at the same time, a control command is generated and sent to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. S6. Continuously acquire the target real-time data of the multimodal sensor module within a preset time, and generate a control log based on the target real-time data.

[0007] The beneficial effects of the wellhead multimodal edge detection and closed-loop control method provided by this invention are as follows: This invention solves the problems of existing well kick monitoring methods relying on a single sensor, having a delayed response, and lacking automatic control by fusing multimodal sensor data such as vision, vibration, and pressure at the wellhead edge and performing unified time base alignment, combined with sliding window feature extraction and adaptive dual threshold judgment. It enables early and accurate identification of well kick / overflow events and automatic triggering of well control equipment such as blowout preventers and choke valves, thereby improving the safety and reliability of drilling operations.

[0008] Based on the above solution, the present invention can be further improved as follows.

[0009] Furthermore, the preprocessing includes: noise filtering, outlier removal, and unit conversion.

[0010] Furthermore, the characteristic parameters include: the average value, growth rate, and fluctuation amplitude of the pressure signal; the effective value and main frequency band energy of the vibration signal; and the frame difference characteristics of the visual image sequence.

[0011] Furthermore, the automatic updating of the threshold baseline based on historical data, and the setting of warning thresholds and accident thresholds, specifically includes: The warning threshold and the accident threshold are automatically adjusted based on the dynamic changes in wellhead conditions, including changes in drilling depth and mud properties.

[0012] Furthermore, after sending the control command to at least one of the blowout preventer, throttle valve, and mud circulation pump, the method further includes: The control effect is evaluated based on the continuously acquired real-time data of the target; if the evaluation result indicates that the wellhead condition is unstable, predetermined secondary control measures are implemented, including increasing mud density or activating the standby pump.

[0013] 2) Secondly, the present invention also provides a wellhead multimodal edge detection and closed-loop control system, the specific technical solution of which is as follows: The acquisition module is used to: acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; The calculation module is used to: extract feature parameters from each preprocessed modal data using a sliding time window, perform multimodal fusion on the extracted feature parameters, and calculate a cross-modal consistency score; The update module is used to: automatically update the threshold baseline based on historical data, and set the warning threshold and the accident threshold, wherein the accident threshold is greater than the warning threshold; The judgment module is used to: enter a warning state and continuously monitor when the cross-modal consistency score exceeds the warning threshold; in the warning state, judge whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result; The adjustment module is used to: determine whether a well kick or overflow event has occurred when the judgment result is yes, and generate a control command to send the control command to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. The closed-loop control module is used to: continuously acquire the target real-time data of the multimodal sensor module within a preset time, and generate a control log based on the target real-time data.

[0014] Based on the above solution, the present invention can be further improved as follows.

[0015] Furthermore, the preprocessing includes: noise filtering, outlier removal, and unit conversion.

[0016] Furthermore, the characteristic parameters include: the average value, growth rate, and fluctuation amplitude of the pressure signal; the effective value and main frequency band energy of the vibration signal; and the frame difference characteristics of the visual image sequence.

[0017] Furthermore, the automatic updating of the threshold baseline based on historical data, and the setting of warning thresholds and accident thresholds, specifically includes: The warning threshold and the accident threshold are automatically adjusted based on the dynamic changes in wellhead conditions, including changes in drilling depth and mud properties.

[0018] Furthermore, after sending the control command to at least one of the blowout preventer, throttle valve, and mud circulation pump, the method further includes: The evaluation module is used to: evaluate the control effect based on the continuously acquired real-time data of the target; if the evaluation result is that the wellhead state is unstable, then a predetermined secondary control measure is executed, which includes increasing the mud density or turning on the standby pump.

[0019] 3) In a third aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the electronic device to perform any of the above methods.

[0020] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to enable a computer to implement any of the above methods.

[0021] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0022] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a wellhead multimodal edge detection and closed-loop control method according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the wellhead multimodal edge detection and closed-loop control system structure of a wellhead multimodal edge detection and closed-loop control method according to an embodiment of the present invention; Figure 3 This is a flowchart of a wellhead multimodal edge detection and closed-loop control method according to an embodiment of the present invention; Figure 4 This is a structural framework diagram of an electronic device according to the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0024] like Figures 1 to 3 As shown in the figure, a wellhead multimodal edge detection and closed-loop control method according to an embodiment of the present invention includes the following steps: S1, acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; S2, a sliding time window is used to extract feature parameters from each of the preprocessed modal data, and the extracted feature parameters are fused in a multimodal manner to calculate the cross-modal consistency score; S3 automatically updates the threshold baseline based on historical data and sets the warning threshold and accident threshold, where the accident threshold is greater than the warning threshold; S4, when the cross-modal consistency score exceeds the warning threshold, enter the warning state and continue to monitor. In the warning state, determine whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result. S5, when the judgment result is yes, it is determined that a well kick event or a overflow event has occurred, and at the same time, a control command is generated and sent to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. S6. Continuously acquire the target real-time data of the multimodal sensor module within a preset time, and generate a control log based on the target real-time data.

[0025] The beneficial effects of the wellhead multimodal edge detection and closed-loop control method provided by this invention are as follows: This invention solves the problems of existing well kick monitoring methods relying on a single sensor, having a delayed response, and lacking automatic control by fusing multimodal sensor data such as vision, vibration, and pressure at the wellhead edge and performing unified time base alignment, combined with sliding window feature extraction and adaptive dual threshold judgment. It enables early and accurate identification of well kick / overflow events and automatic triggering of well control equipment such as blowout preventers and choke valves, thereby improving the safety and reliability of drilling operations.

[0026] It should be noted that, for ease of understanding, the terms involved in this invention are explained as follows: Cross-modal consistency score: This represents a dimensionless value obtained by fusing feature parameters extracted from visual sensors, vibration sensors, and pressure sensors within the same time window. It is used to quantify the degree of coordinated change among different modal data. When a well kick or overflow occurs at the wellhead, the feature parameters of multiple sensors will simultaneously show abnormal changes, leading to a significant increase in the cross-modal consistency score. The specific calculation formula can be determined using the weighted Euclidean distance or cosine similarity after normalization of each modal feature.

[0027] Sliding time window: refers to the process of extracting a segment of data from a continuous data stream for analysis at fixed time lengths (e.g., 1 second, 5 seconds, or 10 seconds), and continuously moving the window forward as new data arrives, thereby achieving real-time feature extraction.

[0028] Warning threshold (T1) and incident threshold (T2): The warning threshold is a scoring limit used to trigger an early warning state; the incident threshold is a higher scoring limit used to confirm the occurrence of a well kick or overflow event, and the incident threshold value is greater than the warning threshold.

[0029] Blowout preventer: A large hydraulic control valve assembly installed at the wellhead to seal the wellhead in an emergency and prevent uncontrolled ejection of formation fluids.

[0030] Choke Valve: A regulating valve used to control back pressure at the wellhead, adjusting the pressure inside the wellbore by changing the opening degree.

[0031] Mud circulation pump: A high-pressure pump used to inject drilling fluid (mud) into the wellbore, which can pump in high-density mud during well control operations.

[0032] Control log: An electronic record of the time of an abnormal event, sensor data characteristics, changes in cross-modal consistency scores, issued control commands, and feedback from equipment actions, used for post-event traceability and analysis.

[0033] In another embodiment of this solution, S1 is specifically implemented as follows: A multimodal sensor module is deployed at the wellhead. This module includes a vision sensor, a vibration sensor, and a pressure sensor. The vision sensor, using an industrial camera, is mounted directly above or to the side of the wellhead to capture image sequences of visible signs such as wellhead overflows and ejecta. The vibration sensor, using an accelerometer, is mounted on the wellhead assembly or derrick to collect derrick vibration and impact signals. The pressure sensor is installed at the wellhead casing pressure gauge, riser, or choke line to detect pressure changes within the wellbore. All sensors are connected to an edge computing device via a wired or wireless industrial bus.

[0034] The edge computing device continuously acquires real-time data from the aforementioned sensors. To ensure time alignment of data from different modalities, the real-time data undergoes time synchronization processing: the edge computing device has a built-in unified clock source (e.g., based on the IEEE 1588 precise time protocol or the global positioning second pulse) to add a synchronization timestamp to each sensor data, so that visual data frames, vibration sampling points, and pressure sampling values ​​have a unified time base reference.

[0035] The time-synchronized data undergoes preprocessing. Preprocessing includes the following steps: (1) Noise removal: Low-pass filters (e.g., Butterworth filters with a cutoff frequency of 100Hz) are used to remove high-frequency interference from pressure and vibration signals; median filtering is used to remove salt-and-pepper noise from visual images.

[0036] (2) Outlier removal: Based on the Laida criterion (3σ criterion), outlier data points caused by sensor momentary failure or communication interference are identified and removed.

[0037] (3) Unit conversion: Convert the original values ​​of each sensor into standard engineering units. For example, pressure is converted into megapascals (MPa) and vibration acceleration is converted into meters per square second (m / s²).

[0038] It should be noted that a visual sensor is an electronic device capable of acquiring visible light images or videos of the wellhead area, used to monitor visually identifiable phenomena such as liquid spills and solid particle ejection.

[0039] Vibration sensor: A sensing element that can detect the mechanical vibration acceleration, velocity or displacement of the derrick or wellhead equipment. It is usually a piezoelectric or microelectromechanical accelerometer.

[0040] Pressure sensor: A transmitter that converts fluid pressure into an electrical signal, used to measure wellhead casing pressure, riser pressure, or choke line pressure.

[0041] By acquiring real-time data from three different modalities—visual, vibration, and pressure—and performing time-synchronized processing, the limitations of incomplete information from a single sensor are overcome, providing a reliable data foundation for subsequent multimodal fusion. The preprocessing steps, including noise filtering, outlier removal, and unit conversion, effectively improve data quality, reduce the risk of false alarms, and enhance robustness in harsh wellfield environments.

[0042] In another embodiment of this solution, S2 is specifically implemented as follows: The edge computing device extracts features from each modal data after preprocessing using a sliding time window. Specifically, a fixed-length time window is set (e.g., a window length of 5 seconds and a sliding step of 1 second). For the pressure data stream within each time window, the following feature parameters are extracted: the average pressure signal (arithmetic mean of all pressure sampling points within the window), the pressure growth rate (the difference between the pressure values ​​at the end and beginning of the window divided by the window duration), and the pressure fluctuation amplitude (the difference between the maximum and minimum pressure values ​​within the window). For the vibration data stream, the effective value of the vibration signal (i.e., the root mean square value, reflecting vibration energy) and the dominant frequency band energy are extracted (the energy integral is calculated within the frequency band corresponding to the dominant frequency component after performing a fast Fourier transform on the vibration data within the window). For the visual data stream, the frame difference features of the visual image sequence are extracted: the sum of the absolute values ​​of the pixel intensity differences between two adjacent frames is calculated and normalized as a measure of the intensity of motion; or the velocity field amplitude of the liquid flow in the wellhead area is calculated using the optical flow method.

[0043] The extracted modal feature parameters are then fused using multimodal methods to calculate a cross-modal consistency score. The specific calculation method is as follows: each feature parameter is normalized (subtracting the historical mean and dividing by the standard deviation). Weights are assigned to the normalized pressure, vibration, and visual features (e.g., pressure weight 0.4, vibration weight 0.3, visual weight 0.3). Weighted Euclidean distance or weighted cosine similarity is then calculated to obtain a cross-modal consistency score between 0 and 1. A higher score indicates a greater likelihood of simultaneous anomalous changes across different sensors. The normalization parameters and weighting coefficients can be obtained through offline calibration or adaptive learning.

[0044] Among them, the sliding time window is a real-time data processing mechanism that extracts continuous data segments with a fixed time length and moves forward over time, with each window performing feature calculations independently.

[0045] Feature parameters: Key quantitative indicators extracted from raw sensor data that reflect the wellhead operating conditions.

[0046] Average pressure signal: the arithmetic mean of all pressure sampling points within a time window.

[0047] Pressure growth rate: The change in pressure per unit time. The formula is (pressure value at the end of the window minus pressure value at the beginning of the window) divided by the window duration.

[0048] Pressure fluctuation amplitude: The difference between the maximum and minimum pressure values ​​within a time window.

[0049] The effective value of a vibration signal is the root mean square of the vibration amplitude, which reflects the average energy level of the vibration.

[0050] Main frequency band energy: The energy integral value within the frequency range where the energy is most concentrated in the vibration signal spectrum.

[0051] Frame difference features of visual image sequences: a measure of the difference between two consecutive frames of images, used to detect the intensity of fluid spillage at the wellhead or the movement of an object.

[0052] Multimodal fusion: The process of combining feature parameters from different types of sensors into a comprehensive index according to certain rules.

[0053] Cross-modal consistency score: a dimensionless numerical value used to quantify the degree of synchronous change in time among visual, vibration, and pressure sensor data. The higher the score, the greater the likelihood that the data from multiple sources jointly indicate an anomaly.

[0054] The sliding time window enables continuous tracking of changes in operating conditions, avoiding misjudgments caused by single-point instantaneous noise. Extracted feature parameters such as pressure growth rate, vibration dominant frequency band energy, and visual frame differences capture precursory information of well kicks / overflows from the perspectives of fluid dynamics, structural mechanics, and visual morphology, respectively. Multimodal fusion calculation of cross-modal consistency scores fully utilizes the redundancy and complementarity between sensors, significantly improving the reliability of anomaly detection and exhibiting stronger anti-interference capabilities compared to single-sensor methods.

[0055] In another embodiment of this solution, S3 is specifically implemented as follows: The edge computing device automatically updates the threshold baseline based on historical data and sets warning thresholds and accident thresholds. Specifically, during normal drilling phases (no well kicks, no overflows), sensor data is continuously collected, cross-modal consistency scores are calculated at each moment, and scores for the most recent period (e.g., the past 24 hours) are stored. The mean μ and standard deviation σ of these historical scores are calculated. The warning threshold T1 is set as μ + k1·σ, and the accident threshold T2 is set as μ + k2·σ, where k2 > k1 (e.g., k1 is 2.5, k2 is 4.0), thus ensuring that the accident threshold is greater than the warning threshold.

[0056] To adapt to dynamic changes in drilling conditions (such as gradual increases in drilling depth, adjustments to mud properties, and changes in formation pressure), the threshold baseline is updated in real time. Every update cycle (e.g., every 10 minutes), the mean and standard deviation of scores within the most recent time window are recalculated, and the warning and accident thresholds are automatically adjusted accordingly. This adaptive mechanism allows the thresholds to adjust slowly with the normal operating conditions, avoiding missed or frequent false alarms caused by fixed thresholds.

[0057] Historical data refers to the cross-modal consistency score sequence collected and stored during periods deemed to be under normal operating conditions (without abnormal events).

[0058] Threshold baseline: A reference benchmark calculated based on the statistical characteristics of historical data (such as mean and standard deviation), used to dynamically set warning thresholds and accident thresholds.

[0059] Warning threshold: A numerical limit. When the cross-modal consistency score exceeds this limit, a warning state is entered, indicating that there may be early signs of well kick or overflow.

[0060] Accident threshold: A numerical limit. When the value exceeds the warning threshold, the cross-modal consistency score is confirmed to have caused a well kick or overflow event.

[0061] Dynamic changes in wellhead conditions: refers to the changes in wellhead environment, formation conditions, drilling parameters, and other factors over time during drilling operations, including but not limited to increases in drilling depth, adjustments in mud density, changes in discharge rate, and fluctuations in formation pressure.

[0062] Drilling depth variation: refers to the continuous process of the drill bit gradually descending from the surface into deeper formations. As the depth increases, the hydrostatic pressure in the wellbore and the formation pressure will change.

[0063] Mud performance changes: refers to the adjustment or natural changes in the physicochemical properties of drilling fluid (mud) such as density, viscosity, and water loss during circulation.

[0064] An automatic threshold update mechanism based on historical data enables adaptive tracking of slow changes in normal operating conditions, preventing fixed thresholds from becoming invalid due to environmental drift. Two levels of thresholds (early warning threshold and incident threshold) are set, achieving tiered early warning: a more sensitive early warning threshold is used to detect potential problems in advance, while a stricter incident threshold is used to confirm events, balancing sensitivity and specificity. Thresholds are dynamically adjusted based on changes in drilling depth and mud properties, further improving consistency across different operational stages.

[0065] In another embodiment of this solution, S4 is specifically implemented as follows: The edge computing device monitors the calculated cross-modal consistency score S(t) in real time. When S(t) exceeds the warning threshold T1, a warning state is entered. In the warning state, strong actions such as well shut-in are not immediately performed. Instead, the monitoring frequency is increased (e.g., the sliding step size of the sliding time window is shortened from 1 second to 0.5 seconds), and S(t) for subsequent time windows is continuously calculated. At the same time, a yellow warning is issued through the local human-machine interface to remind operators of the possibility of wellhead anomalies.

[0066] In the early warning state, the system continues to assess whether subsequent cross-modal consistency scores exceed the accident threshold T2. If S(t) exceeds T2 within several consecutive time windows (e.g., three consecutive windows), a "confirmed well kick or overflow event" result is generated; if S(t) falls below the early warning threshold, the system exits the early warning state and resumes normal monitoring. To reduce state fluctuations caused by noise, a hysteresis comparison mechanism can be used: the system only exits the early warning state after S(t) has been continuously below T1 for a certain period of time (e.g., 2 seconds).

[0067] The warning status is an intermediate state between normal monitoring and accident confirmation, indicating that abnormal signs have been detected but the event has not yet been finally confirmed.

[0068] Continuous monitoring: This refers to analyzing sensor data at a higher sampling or calculation frequency during an alert state in order to promptly detect whether the score has risen further.

[0069] Judgment result: A Boolean value, which takes the values ​​"yes" or "no", indicating whether, after continuous observation in the warning state, it is confirmed that the cross-modal consistency score has exceeded the accident threshold.

[0070] By setting an early warning state as a buffer phase, accident actions are avoided from being directly triggered by instantaneous noise or brief disturbances, reducing the risk of erroneous actions. Increasing the monitoring frequency during the early warning state ensures close tracking of anomaly developments. Combined with continuous multi-window judgment or hysteresis mechanisms, the reliability of event confirmation is further improved, ensuring that subsequent control commands are only triggered when the anomaly persists and intensifies.

[0071] In another embodiment of this solution, S5 is specifically implemented as follows: When the judgment result is "yes" (i.e., a kick or overflow event is confirmed), the edge computing device immediately determines that a dangerous condition has occurred at the wellhead. It then generates control commands and sends them via the well control execution module to at least one of the following devices: blowout preventer, choke valve, and mud circulation pump. The specific type of control command is automatically selected based on a preset well control strategy. If the event type is a well kick and the wellhead pressure rises rapidly, a command to shut down the blowout preventer is generated, which drives the blowout preventer hydraulic control unit to quickly close the gate and seal the wellhead.

[0072] If the event type is overflow but the pressure rises slowly, a command to adjust the throttle valve opening is generated to reduce the throttle valve opening from the current value (e.g., close it by 20%), in order to increase the wellhead back pressure and inhibit the formation fluid from continuing to enter the wellbore.

[0073] If well control operations are required, a command to start the mud circulation pump is generated, and the pump speed is set (e.g., starting at 50% of the rated displacement) to pump high-density mud into the wellbore.

[0074] The aforementioned control commands are transmitted to the corresponding actuators via industrial fieldbuses (such as Modbus control protocol / Internet protocol, or Controller Area Network bus). To ensure reliability, each command is accompanied by a checksum, and a retransmission mechanism is employed to ensure successful delivery at least once.

[0075] Control commands are commands in digital or analog signal format used to instruct well control equipment to perform specific actions.

[0076] Blowout preventer (BOP): A hydraulically driven valve assembly used to close and seal the wellbore within seconds, preventing formation fluids from gushing out.

[0077] Throttling valve: An adjustable valve installed on a throttling line to control wellhead back pressure.

[0078] Mud circulation pump: A high-pressure pump used to inject drilling fluid (mud) into the wellbore. In well control operations, it is used to pump in high-density mud to balance formation pressure.

[0079] Close Blowout Preventer Command: A command that instructs the blowout preventer control unit to close the gate.

[0080] Throttle valve opening adjustment command: A command that instructs the throttle valve actuator to increase or decrease the valve opening.

[0081] Start mud circulation pump command: A command that instructs the mud pump frequency converter or starter to start running and reach the set speed.

[0082] Upon confirmation of a well kick or overflow event, control commands are automatically generated and sent without waiting for manual intervention, achieving a closed-loop linkage from detection to execution, reducing response time to within a few seconds. Targeted control commands (closing the blowout preventer, adjusting the throttle valve, or starting the mud pump) are selected based on different anomaly characteristics, reflecting a tiered control strategy and avoiding over- or under-response. This step directly solves the problems of delayed manual operation and lack of automated collaborative control in the background technology.

[0083] In another embodiment of this solution, S6 is specifically implemented as follows: After sending the control command, the edge computing device continuously acquires real-time target data from the multimodal sensor module within a preset time period (e.g., within 30 seconds after sending the command). Here, "real-time target data" refers to the visual, vibration, and pressure data newly acquired after sending the control command, used to evaluate the effectiveness of the control action.

[0084] Control logs are generated based on real-time target data. The control logs include: event trigger time, peak cross-modal consistency score at event confirmation, type and specific parameters of sent control commands (e.g., the issuance time of the blowout preventer (BOP) closing command, target throttle valve opening), characteristic change curves of each sensor before and after the control action, and equipment action feedback signals (e.g., BOP position switch status, actual throttle valve opening, mud pump operating current). The control logs are stored in encrypted format on the local solid-state drive of the edge computing device and can optionally be uploaded to the central control room or cloud server via a remote communication interface for post-event analysis and auditing.

[0085] Furthermore, the control effectiveness is evaluated based on continuously acquired real-time target data. The evaluation method involves comparing the differences between the wellhead pressure, vibration characteristics, and visual characteristics after the control command is issued and the expected target. For example, if the wellhead pressure should quickly stabilize within a safe range after the blowout preventer is shut off, and the pressure continues to rise or fluctuates drastically, the wellhead condition is considered unstable. When the evaluation result indicates that the wellhead condition is unstable, predetermined secondary control measures are implemented. Secondary control measures include increasing the mud density (e.g., increasing the mud density by 0.2 g / cm³ by mixing in weighting materials) or activating the standby mud circulation pump to increase the circulation rate. After implementing the secondary measures, monitoring continues until the wellhead condition stabilizes.

[0086] Once the wellhead condition stabilizes (e.g., pressure stabilizes, vibration energy drops to normal levels, and visual images show no signs of overflow), release the control command according to safety procedures: slowly open the blowout preventer (if previously closed), gradually adjust the throttle valve opening to normal drilling conditions, and stop the mud circulation pump or restore it to normal flow rate. Simultaneously, automatically generate a complete report of this event, including all data from the control log and records of secondary action execution, and restore normal monitoring mode.

[0087] Preset time: A preset time length, such as 30 seconds or 1 minute, used to limit the duration of continuous monitoring after control.

[0088] Target real-time data: refers to the real-time data acquired from the multimodal sensor module within a time period after the control command is sent, which is used to compare with the data before the command to evaluate the control effect.

[0089] Control Log: A structured event log file containing timestamps, sensor characteristics, cross-modal consistency scores, control command content, device feedback, and subsequent evaluation results.

[0090] Evaluating control effectiveness: This is the process of determining whether the control action has achieved the expected safety goals by analyzing the changing trends of sensor data after the control command is issued.

[0091] Secondary control measures: When the primary control instructions fail to completely suppress the anomaly, stronger or different remedial actions are automatically taken.

[0092] Increase mud density: Increase mud density by adding weighting materials (such as barite powder) to the drilling fluid, thereby increasing the fluid column pressure in the wellbore and balancing the formation pressure.

[0093] Start the backup pump: Start another backup mud circulation pump to increase the circulation discharge and improve well control efficiency.

[0094] Continuously acquiring real-time target data after control and generating control logs achieves closed-loop feedback and full-process traceability, providing data support for subsequent well control strategy optimization. Evaluating control effectiveness and implementing secondary control measures further enhances the ability to respond to complex or severe events, preventing escalation of accidents due to single control failures. The generation and storage of control logs meet safety audit requirements, improving engineering practicality. Compared with existing technologies, the closed-loop control method of this invention independently completes detection, decision-making, execution, evaluation, and secondary response at the edge, significantly improving the timeliness, reliability, and adaptability of well control.

[0095] In another embodiment of the invention, a security and traceability module is also provided. This module includes a hardware watchdog timer and a software watchdog process for monitoring the operating status of the edge computing device. When a computation timeout or communication anomaly is detected, the system is automatically switched to a safe state (e.g., maintaining the current well control equipment status, prohibiting the issuance of new control commands, and issuing a fault alarm). Simultaneously, the module employs encrypted storage and hash chain log technology to tamper-proofly preserve every warning, event confirmation, control command issuance, and equipment action feedback, ensuring the authenticity and integrity of the data during post-event auditing. Furthermore, the module has version management functionality. During algorithm or control strategy upgrades, if an anomaly occurs (such as upgrade package verification failure or system instability after upgrade), it can automatically roll back to the previous stable version, preventing on-site failure due to software malfunctions.

[0096] In another embodiment of the invention, the communication and remote interaction module also supports remote policy distribution. The monitoring platform located in the central control room or in the cloud can distribute configuration updates to the edge computing device via a secure communication channel (e.g., an MQTT connection based on the Transport Layer Security (TLS) protocol). These updates include adjusting parameters such as the warning threshold coefficient k1, the accident threshold coefficient k2, the sliding time window length, the feature extraction weight, and the control policy priority (e.g., whether to shut down the blowout preventer first or adjust the throttle valve first). Upon receiving the new policy, the edge computing device dynamically updates its operating parameters after verifying its legality, without requiring downtime or manual on-site intervention.

[0097] This invention also provides a wellhead multimodal edge detection and closed-loop control system, the specific technical solution of which is as follows: The acquisition module is used to: acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; The calculation module is used to: extract feature parameters from each preprocessed modal data using a sliding time window, perform multimodal fusion on the extracted feature parameters, and calculate a cross-modal consistency score; The update module is used to: automatically update the threshold baseline based on historical data, and set the warning threshold and the accident threshold, wherein the accident threshold is greater than the warning threshold; The judgment module is used to: enter a warning state and continuously monitor when the cross-modal consistency score exceeds the warning threshold; in the warning state, judge whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result; The adjustment module is used to: determine whether a well kick or overflow event has occurred when the judgment result is yes, and generate a control command to send the control command to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. The closed-loop control module is used to: continuously acquire the target real-time data of the multimodal sensor module within a preset time, and generate a control log based on the target real-time data.

[0098] It should be noted that the beneficial effects of the wellhead multimodal edge detection and closed-loop control system provided in the above embodiments are the same as those of the wellhead multimodal edge detection and closed-loop control method described above, and will not be repeated here. Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, and will not be repeated here.

[0099] like Figure 4 As shown, an electronic device 300 according to an embodiment of the present invention includes a processor 320 coupled to a memory 310. The memory 310 stores at least one computer program 330, which is loaded and executed by the processor 320 to enable the electronic device 300 to implement any of the above-mentioned methods. Specifically: The electronic device 300 can vary considerably due to differences in configuration or performance. It may include one or more processors 320 (Central Processing Units, CPUs) and one or more memories 310. The memories 310 store at least one computer program 330, which is loaded and executed by the processors 320 to enable the electronic device 300 to implement the wellhead multimodal edge detection and closed-loop control method provided in the above embodiments. Of course, the electronic device 300 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. It may also include other components for implementing device functions, which will not be elaborated upon here.

[0100] An embodiment of the present invention provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to enable a computer to implement any of the above-described methods.

[0101] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0102] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the methods described above.

[0103] It should be noted that the terms "first" and "second" in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0104] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0105] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0106] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A wellhead multimodal edge detection and closed-loop control method, characterized in that, include: S1, acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; S2, a sliding time window is used to extract feature parameters from each of the preprocessed modal data, and the extracted feature parameters are fused in a multimodal manner to calculate the cross-modal consistency score; S3 automatically updates the threshold baseline based on historical data and sets the warning threshold and accident threshold, where the accident threshold is greater than the warning threshold; S4, when the cross-modal consistency score exceeds the warning threshold, enter the warning state and continue to monitor. In the warning state, determine whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result. S5, when the judgment result is yes, it is determined that a well kick event or a overflow event has occurred, and at the same time, a control command is generated and sent to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. S6. Continuously acquire the target real-time data of the multimodal sensor module within a preset time period, and generate a control log based on the target real-time data.

2. The wellhead multimodal edge detection and closed-loop control method according to claim 1, characterized in that, The preprocessing includes: noise filtering, outlier removal, and unit conversion.

3. The wellhead multimodal edge detection and closed-loop control method according to claim 1, characterized in that, The characteristic parameters include: the average value, growth rate, and fluctuation amplitude of the pressure signal; the effective value and main frequency band energy of the vibration signal; and the frame difference characteristics of the visual image sequence.

4. The wellhead multimodal edge detection and closed-loop control method according to claim 1, characterized in that, The automatic updating of the threshold baseline based on historical data, and the setting of early warning thresholds and accident thresholds, specifically include: The warning threshold and the accident threshold are automatically adjusted based on the dynamic changes in wellhead conditions, including changes in drilling depth and mud properties.

5. The wellhead multimodal edge detection and closed-loop control method according to claim 1, characterized in that, After sending the control command to at least one of the blowout preventer, throttle valve, and mud circulation pump, the method further includes: The control effect is evaluated based on the continuously acquired real-time data of the target; if the evaluation result indicates that the wellhead condition is unstable, predetermined secondary control measures are implemented, including increasing mud density or activating the standby pump.

6. A wellhead multimodal edge detection and closed-loop control system, characterized in that, include: The acquisition module is used to: acquire real-time data collected by the multimodal sensor module at the wellhead, the multimodal sensor module including a vision sensor, a vibration sensor and a pressure sensor, and perform time synchronization processing and preprocessing on the real-time data; The calculation module is used to: extract feature parameters from each preprocessed modal data using a sliding time window, perform multimodal fusion on the extracted feature parameters, and calculate a cross-modal consistency score; The update module is used to: automatically update the threshold baseline based on historical data, and set the warning threshold and the accident threshold, wherein the accident threshold is greater than the warning threshold; The judgment module is used to: enter a warning state and continuously monitor when the cross-modal consistency score exceeds the warning threshold; in the warning state, judge whether the cross-modal consistency score exceeds the accident threshold and generate a judgment result; The adjustment module is used to: determine whether a well kick or overflow event has occurred when the judgment result is yes, and generate a control command to send the control command to at least one of the blowout preventer, throttle valve and mud circulation pump. The control command includes: a blowout preventer shutdown command, a throttle valve opening adjustment command or a mud circulation pump start command. The closed-loop control module is used to: continuously acquire the target real-time data of the multimodal sensor module within a preset time, and generate a control log based on the target real-time data.

7. A wellhead multimodal edge detection and closed-loop control system according to claim 6, characterized in that, The preprocessing includes: noise filtering, outlier removal, and unit conversion.

8. A wellhead multimodal edge detection and closed-loop control system according to claim 6, characterized in that, The characteristic parameters include: the average value, growth rate, and fluctuation amplitude of the pressure signal; the effective value and main frequency band energy of the vibration signal; and the frame difference characteristics of the visual image sequence.

9. An electronic device, characterized in that, The electronic device includes a processor coupled to a memory storing at least one computer program, which is loaded and executed by the processor to enable the electronic device to perform the method as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to enable the computer to perform the method as described in any one of claims 1 to 5.

Images