Underground multi-mode overflow monitoring system and method

Abstract

The invention relates to an underground multi-mode overflow monitoring system and method, and belongs to the technical field of well drilling monitoring, the system is arranged in an axial through hole of a drill collar body, and the system comprises a central fluid channel assembly comprising a Venturi tube section communicated with the through hole; the annular sensor array assembly comprises a barrel which is arranged on the outer side of the Venturi tube section in a sleeving mode and attached to the inner wall of the drill collar body, an annular flow channel is formed in the position, corresponding to a high-pressure area of the Venturi tube section, in the barrel, one end of the annular flow channel is closed, and the other end of the annular flow channel is a Venturi drainage ejector communicated to the high-pressure area of the Venturi tube section; the annular flow channel is provided with a plurality of drainage channels, and the annular flow channel is provided with a multi-mode sensor along the circumferential inner wall of the annular flow channel; a plurality of sampling holes communicated with the drainage channel are formed in the drill collar body; the system further comprises a processor, all the sensors are connected with the processor, and the processor communicates with the measurement-while-drilling system. The method and the device have the effect of improving the stability of monitoring work in a complex drilling environment.

Description

Technical Field

[0001] This application relates to the technical field of drilling monitoring, and in particular to a downhole multimodal overflow monitoring system and method. Background Technology

[0002] During drilling operations, the phenomenon of formation fluids accidentally entering the wellbore is called an overflow. If left uncontrolled, the overflow may develop into a blowout, leading to a catastrophic accident.

[0003] Currently, overflow monitoring in the field mainly relies on surface parameters (such as mud pit level, outlet flow rate, and changes in standpipe and casing pressure). However, these methods have significant detection lag, especially for complex conditions such as deepwater drilling and narrow-density window drilling. By the time formation fluid intrudes into the wellbore and is detected by the surface system, the optimal control opportunity has often been missed. Some downhole monitoring solutions have emerged in the existing technology, but the sensors are directly exposed to the high-speed, multiphase, and cuttings-containing complex flow field of the wellbore enclosure. The measured values ​​are easily interfered with, and the sensor body is easily eroded and damaged.

[0004] Therefore, there is an urgent need for an overflow monitoring system that has strong anti-interference capabilities and can work stably for a long time in harsh underground environments. Summary of the Invention

[0005] To improve the stability of monitoring operations in complex drilling environments, this application provides a downhole multimodal overflow monitoring system and method.

[0006] Firstly, this application provides a downhole multimodal overflow monitoring system, which adopts the following technical solution: A downhole multimodal overflow monitoring system, installed within the axial through-hole of the drill collar body, includes: A central fluid channel assembly, including a Venturi tube segment communicating with the through-hole; The annular sensor array assembly includes a cylinder sleeved on the outside of the Venturi tube section and attached to the inner wall of the drill collar body. An annular flow channel is provided inside the cylinder at the position corresponding to the high-pressure zone of the Venturi tube section. One end of the annular flow channel is closed, and the other end is connected to a Venturi flow injector in the high-pressure zone of the Venturi tube section. Multiple flow channels are provided on the annular flow channel, and multi-mode sensors are provided along the circumferential inner wall of the annular flow channel. The drill collar body is provided with multiple sampling holes that communicate with the drainage channel; The system also includes a processor, and each of the sensors is connected to the processor, which communicates with the measurement while drilling system.

[0007] By adopting the above technical solution, during the drilling process, when the drilling fluid passes through the Venturi section, a low-pressure zone is formed at the throat of the Venturi section, creating a suction force from the high-pressure zone to the low-pressure zone. Under this suction force, the liquid in the annulus enters the annular flow channel through the sampling port. While flowing in the annular flow channel, multimodal sensors monitor the liquid to obtain multimodal monitoring data. The liquid in the annulus flows into the main drilling fluid stream through the Venturi jet. At the same time, the processor collects the monitoring data from the sensors and uploads the analyzed and processed results, along with the monitoring data, to the surface through the measurement while drilling system.

[0008] Therefore, this system, through a coaxial integrated structural design, organically combines multimodal sensors with a Venturi drainage structure to achieve synchronous and accurate monitoring of drilling fluid parameters. The Venturi tube section and the drainage ejector construct a stable sampling channel, preventing sensor erosion by cuttings and adapting to harsh downhole conditions. Multimodal sensors synchronously collect key fluid parameters, which are analyzed in real time by a processor and linked to the measurement-while-drilling system, solving the problems of single monitoring parameters and asynchronous data in traditional methods. This significantly improves the accuracy of early overflow identification, reduces false alarms and missed alarms, shortens emergency response time, effectively prevents blowout accidents, and meets the safety monitoring needs of complex drilling scenarios.

[0009] Furthermore, the multimodal sensor includes one or more of the following: impedance sensor, temperature sensor, acoustic sensor, density monitoring sensor, and solid phase monitoring sensor. A flow guide grid is installed upstream of the inlet of the Venturi tube section, and a filter screen is installed inside the sampling orifice.

[0010] By adopting the above technical solutions, multi-modal sensors can be flexibly combined to simultaneously collect multi-dimensional parameters such as impedance, temperature, and density, making up for the limitations of single sensor monitoring, accurately identifying the overflow type and intrusion degree, and significantly reducing the false alarm rate; the flow guide can sort the fluid and eliminate eddies, ensuring the stability of the flow field; the sampling hole filter screen intercepts rock cuttings, avoiding flow channel blockage and sensor damage, and improving the stability of the system and the accuracy of measurement data in harsh downhole environments.

[0011] Secondly, this application provides a downhole multimodal overflow monitoring method based on the system described in the first aspect, employing the following technical solution: Simultaneously acquire real-time monitoring data from multimodal sensors; The real-time monitoring data is preprocessed to obtain preprocessed real-time monitoring data; Based on the preprocessed real-time monitoring data and combined with drilling engineering design parameters, a three-dimensional downhole working condition virtual model is constructed. Feature extraction is performed on the preprocessed real-time monitoring data to obtain multi-dimensional features. Principal component analysis algorithm is used to reduce the dimensionality of the multi-dimensional features to obtain the core feature vector. The core feature vector is input into a weighted fusion model to obtain the fusion result. The weighted fusion model has different weights for each mode of sensor. Based on the fusion results, an overflow risk index diagnostic result is generated.

[0012] By adopting the above technical solution, multimodal real-time data is collected synchronously. After preprocessing to remove noise and outliers to ensure data quality, a three-dimensional virtual model is constructed based on engineering parameters to accurately recreate the downhole working conditions. Core information is extracted and refined through feature extraction and dimensionality reduction. The importance of each sensor is highlighted by a weighted fusion model, which efficiently integrates multi-dimensional data. The resulting overflow risk index enables quantitative diagnosis of overflow risk, solves the problems of fragmented monitoring data and subjective judgment in traditional monitoring, and significantly improves the accuracy and timeliness of overflow monitoring, providing a reliable decision-making basis for safe downhole drilling.

[0013] Furthermore, the method also includes: Measured annular pressure data were obtained using a measurement-while-drilling system. Based on the three-dimensional downhole working condition virtual model, simulated and predicted drilling fluid pressure distribution data are obtained; By comparing the drilling fluid pressure distribution data with the measured annular pressure data, the deviation ratio is obtained; If the drilling fluid pressure distribution data is normal, but the measured annular pressure data is abnormally reduced, a prompt message will be generated to troubleshoot sensor or drilling tool malfunctions. If the deviation ratio is less than the preset value, the overflow risk index diagnosis result will not be modified. If the deviation ratio is not less than the preset value, the model parameters are corrected, and the overflow risk index diagnosis result is recalculated based on the corrected three-dimensional downhole working condition virtual model.

[0014] By adopting the above technical solution, the pressure data predicted by the actual measurement and the virtual model are compared to accurately control the data deviation. When an anomaly occurs, the fault can be promptly identified and investigated. If the deviation exceeds the standard, the model parameters are corrected and the risk index is recalculated, thus avoiding model distortion and misjudgment. This significantly improves the reliability and accuracy of overflow risk diagnosis results and provides dual protection for downhole monitoring.

[0015] Furthermore, the method also includes: The core feature vector is matched with the features of historical cases in the historical overflow case feature library to obtain the similarity score. If the similarity is greater than the similarity threshold, the overflow risk index is increased to obtain the updated overflow risk index diagnosis result.

[0016] By adopting the above technical solution and using the similarity matching between the core feature vector and historical overflow cases, and by fully drawing on past experience, the overflow risk index can be increased in a timely manner when the similarity exceeds the threshold. This can accurately capture early overflow signals, effectively make up for the limitations of pure model calculation, reduce the risk of overflow omission, and further improve the rationality and reliability of risk diagnosis results, providing experience support for downhole safety early warning.

[0017] Furthermore, if there are two or more historical case features in the historical overflow case feature library, the method for calculating the similarity includes: Define the core feature vector as X, and the i-th historical case feature in the historical overflow case feature library as... The cosine similarity algorithm is applied to calculate the similarity between X and each... Single case similarity : ; in, Let X be the Euclidean norm. for The Euclidean norm; Set minimum similarity threshold Remove < The characteristics of invalid historical cases are retained, and the set of characteristics of valid historical cases is preserved. } and the corresponding effective similarity set { }; the aforementioned Less than the similarity threshold; Similarity of features for each valid historical case We calculate a weighted average of all valid similarities, using the given weights: Where n is the number of valid cases.

[0018] By adopting the above technical solution, the cosine similarity algorithm is used to accurately calculate the similarity of a single case. After removing invalid cases, the similarity is used as the weight for weighted averaging. This not only avoids the interference of randomness in a single case, but also highlights the reference value of high similarity cases, greatly improving the accuracy and reliability of similarity calculation, and providing a scientific and robust basis for subsequent adjustment of the overflow risk index.

[0019] Furthermore, the improvement of the overflow risk index to obtain the updated overflow risk index diagnostic result includes: Define the original overflow risk index as The updated overflow risk index is The basic adjustment range is ∆, the aforementioned The value range is [0, 10], according to The corresponding risk level classification adjustment upper limit: like ∈[0,3], then ≤5, ∆=min (basic adjustment range, 5- ); like ∈[3,6], then ≤7.5, ∆=min (basic adjustment range, 7.5- ); like If ∈[6,8], then ≤9, ∆=min (basic adjustment range, 9- ); like If ∈[8,10], then ∆=0, no adjustment is needed. ; If there exist k ≥ 2 similarity features of historical overflow cases If the similarity exceeds the threshold, the weighted basic adjustment range will be calculated. ,in, This represents the base adjustment range corresponding to the characteristics of the i-th historical overflow case. = + .

[0020] By adopting the above technical solution, an adjustment upper limit is set according to the original risk level, and the magnitude is controlled by a minimum function to avoid misjudgment across levels. When multiple cases are involved, the adjustment magnitude is calculated using weighted averages to highlight the weight of highly similar cases. This approach balances the targetedness and rationality of the adjustment, making the update of the spillover risk index accurate and controllable, and further improving the rigor and reliability of risk diagnosis.

[0021] Furthermore, the method also includes: The overflow risk index results are as follows: [0,3] indicates no risk, [3,6] indicates low risk, [6,8] indicates medium risk, and [8,10] indicates high risk. If the overflow risk index diagnosis result is low risk, the original data sampling frequency is maintained, and a risk status report is output once every preset time. If the overflow risk index diagnosis result is medium risk, an early warning message will be generated, drilling fluid density adjustment suggestions will be automatically pushed, and the data collection frequency will be increased to twice the original frequency. If the overflow risk index diagnosis result is high risk, an emergency linkage command is sent to the drilling measurement system, an audible and visual alarm is sent to the ground monitoring center, and key abnormal parameters are uploaded.

[0022] By adopting the above technical solution, the risk index is divided into four levels, corresponding to differentiated response operations: low-risk cases are reported regularly, medium-risk cases receive early warnings and are monitored more frequently, and high-risk cases trigger emergency linkage and alarms. This achieves precise risk classification and handling, avoiding over-operation while responding quickly to emergencies, providing efficient end-to-end protection for downhole drilling safety.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. Through coaxial integrated structural design, multimodal sensors are organically combined with Venturi drainage structures to achieve synchronous and accurate monitoring of drilling fluid parameters; 2. The Venturi tube section and the flow jet form a stable sampling channel to avoid sensor erosion by rock cuttings and adapt to harsh downhole conditions; 3. Multimodal sensors synchronously collect key parameters of the fluid, which are then analyzed in real time by a processor and linked to the drilling measurement system, solving the problems of single monitoring parameters and asynchronous data in traditional monitoring. Attached Figure Description

[0024] Figure 1 This is a cross-sectional schematic diagram of the downhole multimodal overflow monitoring system in the embodiments of this application.

[0025] Figure 2 This is a cross-sectional schematic diagram illustrating the structure of the ring sensor array assembly in the embodiments of this application.

[0026] Figure 3 This is a structural block diagram of the electrical control connection in the embodiments of this application.

[0027] Figure 4 This is a flowchart of the downhole multimodal overflow monitoring method in the embodiments of this application.

[0028] Reference numerals: 1. Drill collar body; 11. Sampling hole; 12. Filter screen; 2. Central fluid channel assembly; 21. Venturi tube section; 22. Flow guide grid; 3. Annular sensor array assembly; 31. Cylinder; 32. Annular flow channel; 33. Venturi flow injector; 34. Flow channel; 35. Sensor. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the character " / " in this text, unless otherwise specified, generally indicates that the objects before and after it are in an "or" relationship.

[0031] This application provides a downhole multimodal overflow monitoring system, referring to... Figure 1 and Figure 2 It is installed on the drill collar body 1. The drill collar body 1 has a through hole in the axial direction. The central fluid channel assembly 2 and the annular sensor array assembly 3 are arranged sequentially from the inside to the outside in the through hole.

[0032] The annular sensor array assembly 3 and the central fluid channel assembly 2 are nested together from the outside in and coaxially arranged. Both use a honeycomb-shaped integral bracket supported by titanium alloy as a base to fix the electronic components included in the assembly. The titanium alloy bracket provides excellent vibration and shock resistance, and its excellent thermal conductivity quickly conducts the heat generated by the electronic components to the drill collar body 1 for dissipation.

[0033] The central fluid channel assembly 2 includes a Venturi tube section 21 communicating with the through-hole. The Venturi tube section 21 is arranged along the central axis of the through-hole, and its throat diameter is half the diameter of the through-hole. When drilling fluid flows through the Venturi tube section 21, the pressure of the drilling fluid at the throat of the Venturi tube section 21 decreases. Therefore, the upstream of the Venturi tube section 21 is a high-pressure zone, and the throat is a low-pressure zone. A cross-shaped flow guide grid 22 is installed upstream of the inlet of the Venturi tube section 21. The flow guide grid 22 can organize the fluid, eliminate eddies, and ensure the stability of the flow field.

[0034] The annular sensor array assembly 3 includes a cylindrical body 31. The inner diameter of the cylindrical body 31 is interference-fitted with the outer diameter of the Venturi tube section 21, and the outer diameter of the cylindrical body 31 is fitted to the inner wall of the drill collar body 1. An annular flow channel 32 is provided inside the cylindrical body 31 at a position corresponding to the high-pressure zone of the Venturi tube section 21. The annular flow channel 32 is arranged around the circumference of the cylindrical body 31, with one end closed and the other end penetrating through the cylindrical body 31 and the Venturi tube section 21 to connect to the high-pressure zone, thereby forming a Venturi flow injector 33. Multiple flow channels 34 are provided on the annular flow channel 32 in the direction close to the drill collar body 1.

[0035] Multiple sampling holes 11 are arranged along the circumference on the drill collar body 1, and each sampling hole 11 is connected to a corresponding drainage channel 34. A filter screen 12 is installed in the sampling hole 11 to intercept rock cuttings, avoid blockage of the flow channel and damage to the sensor, and improve the stability of the system and the accuracy of measurement data in the harsh downhole environment.

[0036] Therefore, when the drilling fluid flows in the through hole of the drill collar, the pressure difference in the Venturi section 21 generates a suction force from the high-pressure area to the low-pressure area. Under the action of the suction force, the drilling fluid in the annulus enters the drainage channel 34 through the sampling hole 11, and then enters the Venturi section 21 through the Venturi drainage jet 33, and finally mixes with the drilling fluid.

[0037] In order to detect the composition of the annular drilling fluid, multimodal sensors 35 are evenly distributed on the inner circumference of the annular flow channel 32, which are in direct contact with the fluid.

[0038] Reference Figure 3 The multimodal sensor 35 includes: Impedance sensor: Sensitively detects phase changes in oil, gas and water by measuring changes in impedance and capacitance; Distributed temperature sensors: used to acquire continuous temperature profiles; Acoustic sensors: used to acquire vibration signals; Density monitoring sensor: used to obtain the density of drilling fluid in the annulus; Solid phase monitoring sensor: used to acquire solid phase information of annular drilling fluid.

[0039] A sealed electronic chamber is installed inside the cylinder 31, integrating multiple processors, such as a dual-core architecture of FPGA and ARM. The processors are connected to individual sensors 35 via high-temperature shielded cables. The FPGA is responsible for high-speed synchronous acquisition and preprocessing of the massive amounts of raw data generated by the multiple sensors 35; the ARM processor is responsible for running complex data fusion algorithms, pre-stored with fusion algorithms, and performing weighted decision-making on data from multiple modes such as flow rate, acoustic signals, impedance, temperature, and pressure to obtain comprehensive diagnostic results. A standard API interface is provided in the sealed electronic chamber to connect to the pulse generator of the drilling measurement-while-drilling system, directly encoding and uploading the diagnostic results and key data to the surface.

[0040] This application discloses a method for monitoring multimodal overflows in wells. (Refer to...) Figure 4 This is executed by the processor. It includes (steps S101 to S105): Step S101: Synchronously acquire real-time monitoring data from multimodal sensors.

[0041] Step S102: Preprocess the real-time monitoring data to obtain preprocessed real-time monitoring data.

[0042] Specifically, the processor controls the synchronous activation of each modal sensor to acquire real-time monitoring data such as impedance, temperature, acoustic signal, density, and solid phase information. First, the FPGA preprocesses the raw data, using a Kalman filter algorithm to remove noise generated by fluid turbulence and vibration, and an isolated forest algorithm to identify abnormal data. When the abnormal data score exceeds 0.8, it is judged as invalid data and removed. Finally, the min-max normalization method is used to map the data to the [0-1] interval to eliminate dimensional differences.

[0043] Step S103: Based on the preprocessed real-time monitoring data and combined with drilling engineering design parameters, construct a three-dimensional downhole working condition virtual model.

[0044] Specifically, the engineering design parameters include wellbore diameter, drill collar specifications, drilling fluid design density, drilling pump displacement, and predicted formation pressure. Based on the preprocessed data and these drilling engineering design parameters, the processor constructs a three-dimensional virtual model of the downhole working conditions, containing three core parameter layers. In the model, the drilling fluid physical properties layer updates parameters such as density and viscosity in real time; the wellbore flow state layer is based on real-time monitoring data; and the layer interaction layer dynamically adjusts the invasion intensity parameter in conjunction with formation pressure. Thus, the three-dimensional virtual model of downhole working conditions can intuitively display the downhole conditions.

[0045] Step S103: Extract features from the preprocessed real-time monitoring data to obtain multi-dimensional features. Reduce the dimensionality of the multi-dimensional features using principal component analysis to obtain the core feature vector.

[0046] Specifically, modal feature extraction is performed on the preprocessed real-time monitoring data. For example, features such as the rate of change of dielectric constant are extracted from impedance data, features such as temperature gradient are extracted from temperature data, features such as frequency spectrum peaks are extracted from acoustic information through FFT transformation, and features such as deviation values ​​and rates of change are extracted from density and solid phase information. The extracted multi-dimensional features are then reduced to 5-8 dimensional core feature vectors using principal component analysis. These core feature vectors are a condensation of all sensor monitoring information and contain the key features for judging overflow.

[0047] Step S104: Input the core feature vector into the weighted fusion model to obtain the fusion result. The weighted fusion model has different weights for each mode of sensor.

[0048] Specifically, the weighted fusion model is constructed based on a fuzzy logic algorithm. The initial weight allocation for each modality is as follows: density 0.28, acoustic 0.22, impedance 0.18, temperature 0.12, solid phase 0.1, with the remainder of 0.1 as an adaptive adjustment margin. The processor dynamically adjusts the weights according to the downhole conditions. When the well depth exceeds 3000 meters, the weight corresponding to density is increased by 0.35; when the solid phase content exceeds 8%, the weight corresponding to solid phase is increased to 0.15. Therefore, the fusion result within the range of 0-10 is obtained by multiplying the core feature vector corresponding to each dimension with its corresponding weight and then summing the results.

[0049] Step S105: Generate overflow risk index diagnostic results based on the fusion results.

[0050] Specifically, the processor obtains the overflow risk index diagnosis result based on the fusion result using preset mapping rules. For example, a fusion result of 0-2 corresponds to a risk index of 0-3, a fusion result of 7-10 corresponds to a risk index of 8-10, and so on.

[0051] Furthermore, in order to correct model errors and ensure the accuracy of diagnostic results, the method also includes steps S11 to S16: Step S11: Obtain measured annular pressure data based on the measurement while drilling system.

[0052] Specifically, the pressure sensor of the measurement while drilling system is installed on the outside of the drill collar, and the measurement position is at the same horizontal height as the annular sensor array assembly. The processor communicates with the measurement while drilling system through a standard API interface and acquires the measured annular pressure data once every preset time.

[0053] Step S12: Obtain simulated and predicted drilling fluid pressure distribution data based on the three-dimensional downhole working condition virtual model.

[0054] Specifically, the processor extracts predicted annular pressure values ​​corresponding to the measured locations from the wellbore flow state layer of the 3D downhole virtual model. The model uses the pressure loss formula... The pressure distribution is calculated, where λ is the friction coefficient, L is the pipe section length, D is the equivalent diameter of the annulus, ρ is the drilling fluid density, and v is the flow velocity, ensuring that the predicted values ​​correspond accurately to the measured locations.

[0055] Step S13: Compare the drilling fluid pressure distribution data with the measured annular pressure data to obtain the deviation ratio.

[0056] Specifically, the processor uses the relative deviation calculation formula: Deviation ratio = |Predicted value - Measured value| / Measured value × 100%. For example, if the predicted pressure is 75 MPa and the measured pressure is 72 MPa, then the deviation ratio is |75 - 72| / 72 × 100% ≈ 4.2%. The processor calculates the deviation ratio in real time and compares it with the preset value.

[0057] Step S14: If the drilling fluid pressure distribution data is normal, but the measured annular pressure data is abnormally low, a prompt message is generated to troubleshoot sensor or drilling tool malfunctions.

[0058] Specifically, if the normal range of drilling pressure distribution is ±5% of the design pressure value, then if the model predicts the pressure within the normal range, but the actual measured pressure is 10% lower than the lower limit of the normal range, it is judged as an abnormal decrease. The processor then generates a fault message and uploads it to the surface monitoring center via the measurement-while-drilling system.

[0059] Step S15: If the deviation ratio is less than the preset value, the overflow risk index diagnosis result will not be modified.

[0060] Specifically, if the preset deviation ratio is 5%, when the calculated deviation ratio is <5%, it indicates that the virtual model prediction value is highly consistent with the actual downhole working conditions and no adjustment is needed. The original overflow risk index diagnosis result remains unchanged, and the deviation data is recorded in the working condition log.

[0061] Step S16: If the deviation ratio is not less than the preset value, then correct the model parameters and recalculate the overflow risk index diagnosis result based on the corrected three-dimensional downhole working condition virtual model.

[0062] Specifically, when the deviation ratio is ≥5%, the drilling fluid density parameter and friction coefficient λ in the model are corrected first. The actual drilling fluid density is derived from the measured pressure value using the following formula: The original density parameters in the model are replaced with the actual density obtained from the derivation; at the same time, the friction coefficient λ is adjusted by ±0.02.

[0063] After the correction is completed, the model is rerun to calculate the pressure distribution data, and then the feature extraction and fusion calculation steps are re-executed to obtain the corrected overflow risk index diagnosis results.

[0064] On the other hand, the process of optimizing diagnostic results by referencing historical data includes steps S21 to S22: Step S21: Perform similarity matching between the core feature vector and the historical case features in the historical overflow case feature library to obtain the similarity.

[0065] Specifically, the processor's memory stores a feature library of historical overflow cases. Each case includes information such as a multi-dimensional core feature vector, overflow type, and processing scheme. The processor uses cosine similarity calculation to match the current core feature vector with the features of each historical case, obtaining the matching degree of a single case, i.e., the similarity score.

[0066] Furthermore, if there are more than two historical case features, the similarity calculation method includes steps S211 to S213: Step S211: Define the core feature vector as X, and the feature of the i-th historical case in the historical overflow case feature library is... The cosine similarity algorithm is applied to calculate the similarity between X and each... Single case similarity : ; in, Let X be the Euclidean norm. for The Euclidean norm.

[0067] For example, X = (0.2, 0.5, 0.3, 0.6, 0.4). = (0.3, 0.4, 0.2, 0.7, 0.5), then the calculated value is... ≈0.96.

[0068] Step S212: Set the minimum similarity threshold Remove < The characteristics of invalid historical cases are retained, and the set of characteristics of valid historical cases is preserved. } and the corresponding effective similarity set { }; Less than the similarity threshold.

[0069] Specifically, after elimination, valid cases with high similarity are retained to ensure the reliability of subsequent calculations.

[0070] Step S213: Calculate the similarity of features for each valid historical case. We calculate a weighted average of all valid similarities, using the given weights: Where n is the number of valid cases.

[0071] Step S22: If the similarity is greater than the similarity threshold, the overflow risk index is increased to obtain the updated overflow risk index diagnosis result, including steps S221 to S222.

[0072] Step S221: Define the original overflow risk index as The updated overflow risk index is The basic adjustment range is ∆. The value range is [0, 10], according to The corresponding risk level classification adjustment upper limit: like ∈[0,3], then ≤5, ∆=min (basic adjustment range, 5- ); like ∈[3,6], then ≤7.5, ∆=min (basic adjustment range, 7.5- ); like If ∈[6,8], then ≤9, ∆=min (basic adjustment range, 9- ); like If ∈[8,10], then ∆=0, no adjustment is needed. .

[0073] Specifically, when ∈[0,3], i.e., no risk. ≤5 to avoid directly jumping to medium risk; when ∈[3,6], that is, when the risk is low, then If ≤7.5, the updated overflow risk index will be limited to the medium risk range; if When the risk is ∈[6,8], that is, medium risk, then ≤9, not exceeding the lower limit of high risk; if Since ∈[8,10] is already high-risk, no adjustment will be made. .

[0074] Furthermore, in order to link the diagnostic results with actual operations based on the risk index-based response steps, the method also includes steps S31 to S34: Step S31: Overflow risk index results: [0,3] is no risk, [3,6] is low risk, [6,8] is medium risk, and [8,10] is high risk.

[0075] Step S32: If the overflow risk index diagnosis result is low risk, maintain the original data sampling frequency and output a risk status report once every preset time.

[0076] Specifically, when the diagnosis result is low risk, the original data sampling frequency is maintained, and the processor generates a risk status report every preset time. The report includes the current risk index, key data of each sensor, and model operation status, and is uploaded to the drilling measurement system on a regular basis.

[0077] Step S33: If the overflow risk index diagnosis result is medium risk, an early warning message is generated, drilling fluid density adjustment suggestions are automatically pushed, and the data acquisition frequency is increased to twice the original frequency.

[0078] Specifically, the processor immediately generates early warning information, while automatically increasing the data acquisition frequency, strengthening real-time monitoring of downhole conditions, and updating the risk index at shorter intervals.

[0079] Step S34: If the overflow risk index diagnosis result is high risk, send an emergency linkage command to the drilling measurement system, send an audible and visual alarm message to the ground monitoring center, and upload key abnormal parameters.

[0080] Specifically, the processor sends an emergency linkage command to the measurement-while-drilling system, triggering the drilling control system to perform operations such as reducing the discharge rate and activating the choke manifold; at the same time, it sends audible and visual alarm information to the ground monitoring center through the pulse generator; and it uploads key abnormal parameters such as density deviation, pressure change, and acoustic signal peak value in real time, increasing the upload frequency to provide data support for ground emergency decision-making.

[0081] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

[0082] Additionally, it should be understood that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

Claims

1. A downhole multi-modal overflow monitoring system, characterized in that, The application relates to a system for monitoring overflow risk of drilling fluid, which comprises a drill collar body (1) and a sensor array assembly (3) arranged in the axial through hole of the drill collar body (1). The system comprises a central fluid passage assembly (2) and a sensor array assembly (3). The sensor array assembly (3) comprises a barrel (31) arranged outside the Venturi tube segment (21) and matched with the inner wall of the drill collar body (1), and a plurality of drainage channels (34) are arranged on the barrel (31). The drill collar body (1) is provided with a plurality of sampling holes (11) communicated with the drainage channels (34). The system further comprises a processor, each sensor (35) is connected with the processor, and the processor is communicated with a measurement-while-drilling system.

2. The system of claim 1, wherein, The multi-modal sensor (35) comprises one or more of an impedance sensor, a temperature sensor, an acoustic sensor, a density monitoring sensor and a solid phase monitoring sensor.

3. The system of claim 1, wherein, A flow guide grid (22) is arranged upstream of the inlet of the Venturi tube segment (21), and a filter screen (12) is arranged in the sampling hole (11).

4. A downhole multi-modal overflow monitoring method based on the system of any one of claims 1-3, characterized in that, The method comprises the following steps: Synchronously acquiring real-time monitoring data of the multi-modal sensor; Pretreating the real-time monitoring data to obtain pretreated real-time monitoring data; Based on the pretreated real-time monitoring data, a three-dimensional downhole working condition virtual model is constructed in combination with drilling engineering design parameters; Feature extraction is performed on the pretreated real-time monitoring data to obtain multi-dimensional features, and the multi-dimensional features are reduced in dimension through a principal component analysis algorithm to obtain a core feature vector; The core feature vector is input into a weighted fusion model to obtain a fusion result, and the weighted fusion model has different weights for each modal sensor; An overflow risk index diagnosis result is generated based on the fusion result.

5. The method of claim 4, wherein, The method further comprises the following steps: Real annulus pressure data are acquired based on the measurement-while-drilling system; Simulated and predicted drilling fluid pressure distribution data are acquired based on the three-dimensional downhole working condition virtual model; The drilling fluid pressure distribution data and the real annulus pressure data are compared to obtain a deviation ratio; If the drilling fluid pressure distribution data are normal and the real annulus pressure data are abnormally reduced, prompt information for checking sensor faults or drilling tool faults is generated; If the deviation ratio is less than a preset value, the overflow risk index diagnosis result is not modified; If the deviation ratio is not less than the preset value, model parameters are corrected, and the overflow risk index diagnosis result is recalculated based on the corrected three-dimensional downhole working condition virtual model.

6. The method of claim 4, wherein, The method further comprises the following steps: The core feature vector is matched with historical case features in a historical overflow case feature library to obtain a similarity; If the similarity is greater than a similarity threshold value, the overflow risk index is increased to obtain an updated overflow risk index diagnosis result.

7. The method of claim 6, wherein, If the historical overflow case features in the historical overflow case feature library are more than two, the similarity calculation method comprises: Define the core feature vector as X, the i-th historical case feature in the historical overflow case feature library as , and apply the cosine similarity algorithm to calculate the single-case similarity of X and each : ; wherein the Euclidean norm of X, is the Euclidean norm of Setting a minimum similarity threshold , eliminating < invalid historical case features, retaining valid historical case feature set } and the corresponding valid similarity set }; the similarity is less than the similarity threshold similarity of each valid historical case feature weighted average calculation of all valid similarities ; where n is the number of valid cases.

8. The method of claim 6, wherein, The overflow risk index is improved to obtain an updated overflow risk index diagnosis result, comprising: The original overflow risk index is defined as The updated overflow risk index is The basic adjustment range is Δ, and the The value range of the basic adjustment range is [0, 10], and the The upper limit value of the corresponding risk level is adjusted. If ∈ [0, 3], then ≤ 5, Δ = min (base adjustment range, 5- ). If ∈ [3, 6], then ≤ 7.5, Δ = min (base adjustment range, 7.5- ). If ∈ [6, 8], then ≤ 9, Δ = min (base adjustment range, 9- ). If ∈ [8,10], then Δ = 0, no adjustment ; If there are k≥2 historical overflow case features with similarity greater than a similarity threshold, the basic adjustment range is calculated by weighting wherein, is the basic adjustment range corresponding to the i-th historical overflow case feature. = + .​ 9. The method of claim 4, wherein, The method further comprises: The overflow risk index result: [0, 3] is no risk, [3, 6] is low risk, [6, 8] is medium risk, and [8, 10] is high risk; If the overflow risk index diagnosis result is low risk, the original data sampling frequency is maintained, and a risk state report is output once every preset time; If the overflow risk index diagnosis result is medium risk, a warning information is generated, an automatic push drilling fluid density adjustment suggestion is generated, and the data acquisition frequency is increased to twice the original frequency; If the overflow risk index diagnosis result is high risk, an emergency linkage instruction is sent to the measurement while drilling system, an audible and visual alarm information is sent to the ground monitoring center, and key abnormal parameters are uploaded.

Images