Gas extraction hole intelligent drilling optimization system and method based on dynamic risk modeling

Abstract

The application discloses a gas extraction hole intelligent drilling optimization system and method based on dynamic risk modeling, and the system comprises a while-drilling dynamic sensing module, a dynamic risk modeling and prediction module, a multi-objective optimization module, a digital twin virtual-real interactive verification module and an intelligent decision-making and execution module. The application builds a multi-dimensional dynamic risk prediction model of jetting orifices, sticking, trajectory out-of-control and the like, generates control instructions matched with a current risk situation based on multi-objective optimization and digital twin virtual-real interactive verification, upgrades gas extraction hole drilling risks from traditional passive alarm to active early warning and intervention based on risk evolution prediction, and significantly improves the drilling efficiency and safety level of the gas extraction hole in the coal mine.

Description

Technical Field

[0001] This invention belongs to the field of underground gas drainage drilling construction technology in coal mines, and relates to an intelligent drilling optimization system and method for gas drainage holes based on dynamic risk modeling. Background Technology

[0002] Pre-drainage of gas from underground boreholes in coal mines is a fundamental measure for gas disaster management. The drilling quality and efficiency of gas drainage boreholes directly affect the continuity of safe production and the effectiveness of drainage. However, the complex geological conditions of coal seams, with widespread development of geological anomalies such as faults, folds, and scour zones, pose significant challenges to gas drainage borehole drilling operations. When encountering geological structural zones or high-gas-enriched areas, the instantaneous imbalance between ground stress and gas pressure can easily lead to blowouts and drill bit jacking. The ejected coal dust and gas can not only cause overall borehole damage but may also trigger gas exceedance accidents. When encountering soft-hard interlayers or fractured coal seams, the sudden change in drill rod stress can easily lead to stuck drill bits and drill bit burial accidents, resulting in drill bit loss and project stagnation. Furthermore, when drilling in complex and fractured strata, if the strata or directional drilling tools cannot meet the design trajectory requirements, the actual drilling trajectory will deviate from the design trajectory, causing trajectory loss of control or borehole abandonment, and even the creation of gas drainage blank zones.

[0003] Currently, drilling operations for underground gas extraction wells in coal mines mainly rely on two modes: First, manual operation based on experience. Drillers observe parameters such as feed / pull-out pressure, rotation pressure, pump pressure, and pump volume, and adjust drilling parameters based on experience to assess the situation inside the well. This method is inherently lagging and subjective, often only responding passively after signs of an accident appear, failing to provide preventative measures. Second, automated drilling rigs are used, but their current control logic is primarily based on preset fixed thresholds, unable to adapt to unknown geological changes ahead. This results in delayed responses to risks such as blowouts, stuck drill bits, and trajectory loss of control. Therefore, there is an urgent need to combine limited static geological information with abundant real-time drilling measurement data to construct a predictive model encompassing dynamic risks such as blowouts, stuck drill bits, and trajectory loss of control. Based on this model, intelligent optimization of drilling processes and trajectory adjustment parameters can be performed to achieve proactive risk avoidance and efficient drilling. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide an intelligent drilling optimization system and method for gas drainage holes based on dynamic risk modeling. By constructing multi-dimensional dynamic risk prediction models for issues such as blowouts, stuck drills, and trajectory loss of control, and based on multi-objective optimization and digital twin virtual-real interaction verification, the system generates control commands that match the current risk situation. This upgrades the drilling risk of gas drainage holes from traditional passive alarms to proactive early warning and intervention based on risk evolution prediction, significantly improving the drilling efficiency and safety level of underground gas drainage holes in coal mines.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] The intelligent drilling optimization system for gas drainage holes based on dynamic risk modeling includes: The drilling dynamic sensing module is used to collect multi-source data in real time during the drilling process, including: engineering parameters, geometric parameters, geological and environmental parameters; The dynamic risk modeling and prediction module is connected to the drilling dynamic sensing module. It is used to perform time alignment, filtering and noise reduction, and feature extraction on the collected multi-source data to construct unified time series data. Based on this, a multi-dimensional dynamic risk prediction model including gas blowout risk, stuck drill risk, and trajectory loss of control risk is established, and the evolution trend of drilling risk is output. The multi-objective optimization module, connected to the dynamic risk modeling and prediction module, is used to receive the current risk situation and establish a multi-objective optimization function with blowout prevention, jamming prevention, trajectory control accuracy and drilling efficiency as optimization objectives, and feed rate, rotational speed, tool face angle, pump pressure and pump volume as decision variables. The NSGA-III algorithm is used to solve the problem online and generate a Pareto optimal solution set containing multiple optimal solutions. The digital twin virtual-real interaction verification module is connected to the multi-objective optimization module. It is used to establish a borehole digital twin, perform virtual drilling simulation and safety verification on the Pareto optimal solution set, and feed back the verification results to the multi-objective optimization module. The intelligent decision-making and execution module is connected to the digital twin virtual-real interaction verification module. It is used to comprehensively optimize the results and twin verification conclusions, generate the final control commands, and send them down to the directional drilling rig and guidance tool for execution. Based on the execution feedback, it realizes online updates of the model and strategy.

[0007] The present invention also includes the following technical features: Specifically, the engineering parameters include: drilling pressure, rotation speed, torque, vibration, pump pressure, and pump flow rate; geometric parameters include: dip angle, azimuth angle, hole depth, and tool face angle; geological and environmental parameters include: formation gamma, resistivity, gas concentration, and gas emission rate.

[0008] Specifically, the dynamic risk modeling and prediction module includes three core sub-models: The gas eruption hazard risk identification submodule employs a Transformer-based encoder network. The encoder input consists of gas concentration, emission velocity, drilling speed, and vibration parameters from the previous N time steps. After processing by the encoder, the gas emission risk index is output through a fully connected layer. and nozzle warning level The model is pre-trained based on historical drilling blowout accident data, using the cross-entropy loss function, with Adam as the optimizer and a learning rate of 0.0001. Dynamic assessment submodule for stuck drill risk: Constructing a system based on real-time torque Mechanical drilling speed Resistance to advancement , slag return rate Vibration spectrum The input is a neural network model; the neural network includes two hidden layers, each with 128 neurons, using the ReLU activation function, and the output layer uses the Sigmoid activation function. The final output is the drill jam risk coefficient. ; The trajectory loss-of-control risk prediction submodule is based on a finite element model of the drill string-rock coupled dynamics model. Inputs include feed rate, rotational speed, tool face angle, pump pressure, and pump flow rate. Output is the predicted trajectory deviation value for the next advance. and probability of loss of control .

[0009] Specifically, the drill string-rock coupled dynamics model uses a simplified beam-column model to describe the buckling behavior of the drill string in the wellbore, and the lateral vibration equation of the drill string is expressed as:

[0010] in, For the bending stiffness of the drill string, This is a lateral displacement. For axial force distribution, The contact stiffness of the hole wall.

[0011] Specifically, in the multi-objective optimization module, the multi-objective optimization function is:

[0012] st

[0013]

[0014] in, Characterizes the blowout prevention hole target; Characterizes the target for preventing drill jamming; Characterize the trajectory control target, weights , Adjust dynamically according to risk; This represents the drilling efficiency target; For time-varying feasible regions; Dynamic constraints include real-time upper limits for gas concentration and equipment limits; The multi-objective optimization was solved using the NSGA-III algorithm, with the following parameters: population size 200, maximum number of iterations 500, crossover probability 0.9, and mutation probability 0.1.

[0015] Specifically, the digital twin virtual-real interaction verification module includes three core sub-models: Virtual geological model: Based on the initial three-dimensional geological exploration data, and combined with the borehole trajectory data and gamma and resistivity data uploaded in real time by the drilling dynamic sensing module, the three-dimensional geological model of the target area is dynamically corrected by the Bayesian update algorithm, so that the virtual geological model gradually approaches the real geological conditions. Virtual drilling rig - borehole model: A rigid-flexible coupling model of the directional drilling rig, drill rod, drill bit and borehole wall is established using finite element dynamics simulation software to simulate the interaction between the drill bit and the coal seam; Virtual sensor model: Add a virtual sensor with the same characteristics as a real sensor to the simulation environment, and add measurement noise.

[0016] Specifically, the security verification logic of the digital twin virtual-real interaction verification module is to simulate the execution of control commands in a virtual environment and monitor whether any of the following situations occur during the simulation process: (a) Gas eruption simulation: If the simulated gas concentration is >1.0%, the outflow velocity is >5 m³ / min, and the duration exceeds 3 seconds, the gas eruption risk is triggered; (b) Simulation of stuck drill event: If the simulated torque exceeds 90% of the rated torque of the directional drilling rig, the slag return rate... If the value is below 0.5 and the duration exceeds 5 seconds, it is determined that a stuck drill has occurred. (c) Trajectory runaway simulation: The actual trajectory is calculated based on the drill bit-coal seam interaction model and compared with the design trajectory. If the deviation of the simulated trajectory exceeds 2m, the trajectory is judged to be out of control. If any situation occurs during the simulation, the instruction is determined to have failed the security check, and the specific risk type and degree are fed back to the multi-objective optimization module.

[0017] Specifically, the intelligent decision-making and execution module includes: The instruction parsing and coordination control unit is used to select the solution that matches the current risk situation from the Pareto solution set that has passed the safety verification as the final instruction, that is, when the risk index is high, a solution that is biased towards safety is selected; when the risk index is low, a solution that is biased towards drilling efficiency is selected. Model predictive disturbance rejection controller is used to construct a dynamic response model with the final command as the setpoint, solve for the optimal control law in a rolling manner, and generate the underlying control signal; The execution status monitoring and adaptive compensation unit is used to monitor the response status of the actuator in real time and to activate the adaptive compensation algorithm to eliminate the system deviation when a systematic deviation is detected. The execution result closed-loop feedback interface is used to feed the execution data back to the online learning module to drive model calibration and updates; The risk response strategy library includes at least the following three proactive intervention strategies: proactive pressure relief strategy before nozzle eruption, proactive anti-jamming strategy before drill jamming, and proactive correction strategy for trajectory loss of control.

[0018] Specifically, the proactive pressure relief strategy before the gas eruption: when the gas eruption risk index output by the dynamic risk modeling and prediction module... At that time, the instruction parsing and coordination control unit automatically activates the blowout preventer intervention mode; Proactive anti-jamming strategy before drill bit jamming: When the risk of drill bit jamming is high... If the torque fluctuation exceeds 20% of the normal value and the slag return rate is less than 0.6%, the command parsing and collaborative control unit will automatically activate the anti-jamming intervention mode. Active trajectory loss correction strategy: When the trajectory loss risk prediction submodule outputs the predicted trajectory deviation value... or probability of getting out of control At that time, the instruction parsing and collaborative control unit automatically activates the correction intervention mode.

[0019] A method for optimizing intelligent drilling of gas drainage boreholes based on the aforementioned intelligent drilling optimization system for gas drainage boreholes based on dynamic risk modeling includes the following steps: Step S10: Import the 3D geological model and borehole engineering design model of the target area, initialize the network parameters of the three prediction models of gas venting, stuck drill and trajectory runaway, initialize the digital twin and NSGA-III optimization algorithm parameters, and build an initial knowledge base containing historical cases; Step S20: During the drilling process, the drilling dynamic sensing module is used to collect multi-source data in real time, including drilling pressure, rotation speed, torque, and gas concentration, and to complete outlier removal, time alignment and feature extraction to form a structured time-series data stream; Step S30: Input the data processed in step S20 into the dynamic risk modeling and prediction module, and output the probability distribution of gas eruption risk probability, stuck drill risk coefficient, and trajectory loss of control in the next advance in real time to form a comprehensive risk situation map. Step S40: Based on the current risk level obtained in step S30, dynamically configure the objective weights and constraints of the multi-objective optimization problem, and then use the NSGA-III algorithm to solve it online to obtain a Pareto optimal solution set containing multiple optimal solutions under the current risk situation. Step S50: Input the optimal solution set obtained in step S40 into the digital twin system for security verification; if the candidate instruction fails the security verification in the digital twin simulation, the specific risk type and degree identified in the simulation are fed back to the multi-objective optimization module, triggering it to dynamically adjust the weight of the optimization objective and recalculate a new candidate solution set; repeat this verification-feedback-re-optimization loop until a set of instructions that passes the security verification is obtained; Step S60: Send the instruction set obtained in step S50 that has passed the safety verification to the directional drilling rig and the guiding tool for execution, so as to accurately complete the drilling task of this round of footage. If the current risk situation triggers the preset active intervention strategy, the corresponding pre-borehole pressure relief, pre-jamming jamming prevention or trajectory loss correction program will be automatically activated. Step S70: After the current drilling footage is completed, compare the actual drilling results with the prediction results of step S30 and the optimization expectations of step S40; use the deviation data between the two to drive the knowledge base and online learning module to automatically calibrate and update the key parameters in the risk prediction model and optimization model, and store this experience as a new case in the knowledge base to form a complete perception-decision-execution-learning closed loop.

[0020] Compared with the prior art, the present invention has the following technical effects: This invention upgrades traditional threshold-based passive alarms to proactive early warning and intervention based on risk evolution prediction by constructing a dynamic risk model that integrates mechanisms and data. This achieves intelligent decision-making and control during the drilling process of underground gas drainage boreholes in coal mines, yielding the following significant technical effects: (i) This invention breaks through the limitations of traditional monitoring methods that can only respond passively after the signs of an accident appear. By using the Transformer encoder to identify the risk of gas venting in real time, the neural network to dynamically assess the risk of stuck drill, and the finite element analysis to predict the risk of trajectory loss of control, it achieves quantitative prediction of the three core risks, predicts risk signs 3 to 5 seconds in advance, and wins a valuable time window for proactive risk intervention, fundamentally changing the passive situation of post-event handling.

[0021] (II) This invention constructs a multi-objective optimization model that can automatically adjust the weight ratio among four objectives—blowout prevention, stuck drill bit prevention, trajectory accuracy, and drilling efficiency—based on real-time risk conditions, and uses the NSGA-III algorithm for online solution. In low-risk sections, the system tends to increase drilling speed; when the risk increases, it automatically tilts towards safety objectives, achieving a dynamic balance between construction safety and drilling efficiency.

[0022] (III) This invention constructs an optimization-twin verification dual-insurance decision-making process. The digital twin virtual-real interaction verification module integrates a virtual geological model, a drilling rig-hole dynamics model, and a virtual sensor model, performing a full-process virtual simulation of each set of optimization instructions before actual execution. By setting clear quantitative verification indicators, it provides an indispensable safety pre-verification link for on-site decision-making, greatly reducing construction blindness and safety risks, and reducing the borehole accident rate by more than 25%.

[0023] (iv) The present invention establishes an intelligent decision-making and execution module, which includes a risk response strategy library such as proactive pressure relief strategy before the blowout, proactive anti-jamming strategy before the drill bit gets stuck, and proactive correction strategy for trajectory loss of control. It can automatically activate risk intervention strategies according to the type of risk encountered during drilling. Compared with manual control, it realizes rapid response to dynamic risks such as gas drainage blowout, drill bit jamming, and trajectory loss of control, and effectively improves the level of intelligence in gas drainage hole drilling. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling.

[0025] Figure 2 Flowchart for multi-objective optimization and the NSGA-III algorithm.

[0026] Figure 3 This is a structural diagram of the digital twin virtual-real interaction verification module.

[0027] Figure 4 This is a schematic diagram of the process for optimizing intelligent drilling of gas extraction holes based on dynamic risk modeling. Detailed Implementation

[0028] The following are specific embodiments of the present invention. It should be noted that the present invention is not limited to the following specific embodiments. All equivalent modifications made based on the technical solutions of this application fall within the protection scope of the present invention.

[0029] Example: This embodiment provides an intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling. (See also...) Figure 1 The system includes: a dynamic sensing module during drilling, a dynamic risk modeling and prediction module, a multi-objective optimization module, a digital twin virtual-real interaction verification module, and an intelligent decision-making and execution module; specifically: The drilling dynamic sensing module is used to collect various types of data in real time during the drilling process, including: Engineering parameters: drilling pressure, rotation speed, torque, vibration, pump pressure, pump flow rate; Geometric parameters: tilt angle, azimuth angle, hole depth, tool face angle; Geological and environmental parameters: formation gamma, resistivity, gas concentration, and gas emission rate.

[0030] The dynamic risk modeling and prediction module, connected to the drilling dynamic sensing module, is used to perform time alignment, filtering and noise reduction, and feature extraction on the collected multi-source data to construct unified time series data. Based on this, a multi-dimensional dynamic risk prediction model is established, including gas blowout risk, stuck drill risk, and trajectory loss of control risk, and the evolution trend of drilling risk is output.

[0031] The multi-objective optimization module, connected to the dynamic risk modeling and prediction module, receives the current risk situation and establishes a multi-objective optimization function with the optimization objectives of blowout prevention, drill jamming prevention, trajectory control accuracy, and drilling efficiency, and the decision variables of feed rate, rotational speed, tool face angle, pump pressure, and pump flow rate. The NSGA-III algorithm is used to solve the problem online, generating a Pareto optimal solution set containing multiple optimal solutions.

[0032] The digital twin virtual-real interaction verification module, connected to the multi-objective optimization module, is used to establish a borehole digital twin, perform virtual drilling simulation and safety verification on the Pareto optimal solution set, and feed the verification results back to the multi-objective optimization module.

[0033] The intelligent decision-making and execution module is connected to the digital twin virtual-real interaction verification module. It is used to comprehensively optimize the results and twin verification conclusions, generate the final control commands, and send them down to the directional drilling rig and guidance tool for execution. Based on the execution feedback, it realizes online updates of the model and strategy.

[0034] Specifically, the dynamic risk modeling and prediction module consists of three core sub-models: The gas eruption hazard risk identification submodule employs a Transformer-based encoder network. The encoder input consists of parameters such as gas concentration, outburst velocity, drilling speed, and vibration from the previous N time steps. After processing by the encoder, a gas outburst risk index is output through a fully connected layer. and nozzle warning level The Transformer encoder has an input dimension of 10 (time steps) × 5 (feature dimension), the multi-head attention mechanism has 8 heads, and the feedforward network has a dimension of 2048. The model is pre-trained based on historical drilling nozzle accident data, using the cross-entropy loss function, the Adam optimizer, and a learning rate of 0.0001.

[0035] The dynamic assessment submodule for stuck drill risk is built based on real-time torque. Mechanical drilling speed Resistance to advancement , slag return rate Vibration spectrum The input neural network model is defined as follows: torque and propulsion resistance directly reflect the friction state between the drill bit and the borehole wall; the cuttings return rate (the ratio of the amount of drill cuttings discharged per unit time to the theoretical amount of drill cuttings) characterizes the smoothness of cuttings removal; and the vibration spectrum represents the contact state between the drill pipe and the borehole wall. The neural network contains two hidden layers, each with 128 neurons, using the ReLU activation function. The output layer uses the Sigmoid activation function, and finally outputs the stuck drill risk coefficient. .

[0036] The trajectory loss-of-control risk prediction submodule is based on a finite element model of drill string-rock coupled dynamics. Inputs include feed rate, rotational speed, tool face angle, pump pressure, and pump flow rate. Output is the predicted trajectory deviation for the next drilling pass. and probability of loss of control .

[0037] Specifically, the drill string-coal seam coupled dynamics model uses a simplified beam-column model to describe the buckling behavior of the drill string in the wellbore, and the lateral vibration equation of the drill string is expressed as:

[0038] in, For the bending stiffness of the drill string, This is a lateral displacement. For axial force distribution, The contact stiffness of the hole wall.

[0039] See Figure 2 The multi-objective function in the multi-objective optimization module is:

[0040] st

[0041]

[0042] in, Characterizes the blowout prevention hole target; Characterizes the target for preventing drill jamming; Characterize the trajectory control target, weights , Adjust dynamically according to risk; This represents the drilling efficiency target (the negative sign indicates maximizing the drilling speed). The time-varying feasible region is dynamically defined by the real-time capabilities of the equipment and geological conditions. The constraints are dynamic, including real-time upper limit of gas concentration and equipment limits.

[0043] The multi-objective optimization was solved using the NSGA-III algorithm, with the following parameters: population size 200, maximum number of iterations 500, crossover probability 0.9, and mutation probability 0.1.

[0044] See Figure 3 The digital twin virtual-real interaction verification module consists of three core sub-models: Virtual geological model: Based on the initial 3D geological exploration data, and combined with the borehole trajectory data and gamma and resistivity data uploaded in real time by the drilling dynamic sensing module, the 3D geological model of the target area is dynamically corrected by the Bayesian update algorithm, so that the virtual geological model gradually approaches the real geological conditions.

[0045] Virtual drilling rig - borehole model: A rigid-flexible coupling model of the directional drilling rig, drill rod, drill bit and borehole wall is established using finite element dynamics simulation software to simulate the interaction between the drill bit and the coal seam; Virtual sensor model: Add a virtual sensor with the same characteristics as a real sensor to the simulation environment, and add measurement noise.

[0046] Specifically, the security verification logic of the digital twin virtual-real interaction verification module is to simulate the execution of control commands in a virtual environment and monitor whether any of the following situations occur during the simulation process: (a) Gas eruption simulation: If the simulated gas concentration is >1.0%, the outflow velocity is >5 m³ / min, and the duration exceeds 3 seconds, the gas eruption risk is triggered; (b) Simulation of stuck drill event: If the simulated torque exceeds 90% of the rated torque of the directional drilling rig, the slag return rate... If the value is below 0.5 and the duration exceeds 5 seconds, it is determined that a stuck drill has occurred. (c) Trajectory runaway simulation: The actual trajectory is calculated based on the drill bit-coal seam interaction model and compared with the design trajectory. If the deviation of the simulated trajectory exceeds 2m, the trajectory is judged to be out of control.

[0047] If any situation occurs during the simulation, the instruction is determined to have failed the security check, and the specific risk type and degree are fed back to the multi-objective optimization module.

[0048] Specifically, the intelligent decision-making and execution module further includes: The instruction parsing and coordination control unit is used to select the solution that matches the current risk situation from the Pareto solution set that has passed the safety verification as the final instruction, that is, when the risk index is high, a solution that is biased towards safety is selected; when the risk index is low, a solution that is biased towards drilling efficiency is selected. Model predictive disturbance rejection controller is used to construct a dynamic response model with the final command as the setpoint, solve for the optimal control law in a rolling manner, and generate the underlying control signal; The execution status monitoring and adaptive compensation unit is used to monitor the response status of the actuator in real time and to activate the adaptive compensation algorithm to eliminate the system deviation when a systematic deviation is detected. The execution result closed-loop feedback interface is used to feed the execution data back to the online learning module to drive model calibration and updates.

[0049] The risk response strategy library includes at least the following three proactive intervention strategies: (a) Active pressure relief strategy before the nozzle: When the gas nozzle risk index output by the dynamic risk modeling and prediction module is... At this time, the command parsing and coordination control unit automatically activates the blowout prevention intervention mode. The control commands include: reducing the feed rate to 30%~50% of the current value to reduce further disturbance of the coal seam by the drill bit, while increasing the pump flow rate to more than 90% of the rated maximum value to enhance the discharge of slag in the hole and prevent further gas accumulation; if the risk continues to rise, the drill bit retraction and pressure relief procedure will be further executed, retracting the drill bit 3 meters and continuously flushing the bottom of the hole until the gas pressure drops below the safety line; (b) Proactive anti-jamming strategy before drill bit jamming: When the risk factor of drill bit jamming is high... If the torque fluctuation exceeds 20% of the normal value and the slag return rate is below 0.6%, the command parsing and coordination control unit automatically activates the anti-sticking intervention mode. The control commands include: periodically executing the "pull-up-feed" program, while increasing the rotational speed to 80%~100% of the rated value to enhance the slag removal capacity; if the sticking risk factor continues to increase, execute the short pull-out program to pull out a certain number of drill pipes until the torque and slag return rate return to normal. (c) Active trajectory loss correction strategy: When the trajectory loss risk prediction submodule outputs the predicted trajectory deviation value... or probability of getting out of control When the error occurs, the command parsing and collaborative control unit automatically activates the correction intervention mode. The control commands include: adopting the sliding drilling mode and fully adjusting the tool face angle according to the direction of the deviation; if the deviation continues to increase, a large-scale adjustment is required, retracting the drill bit to a suitable branch point position and side-drilling the branch hole at a feed rate of 0.2m / min until the actual drilling trajectory returns to the deviation range from the design trajectory.

[0050] See Figure 4 The intelligent drilling optimization method for gas drainage boreholes based on the above system includes the following steps: Step S10: Import the 3D geological model and borehole engineering design model of the target area, initialize the network parameters of the three prediction models such as gas venting and trajectory runaway, initialize the digital twin and NSGA-III optimization algorithm parameters, and build an initial knowledge base containing historical cases.

[0051] Step S20: During the drilling process, the drilling dynamic sensing module is used to collect multi-source data such as drilling pressure, rotation speed, torque, and gas concentration in real time, and completes outlier removal, time alignment and feature extraction to form a structured time-series data stream.

[0052] Step S30: Input the data processed in Step S20 into the dynamic risk modeling and prediction module, and output in real time the probability distribution of gas outburst risk, stuck drill risk coefficient, and trajectory loss of control in the next drilling run, forming a comprehensive risk situation map. Optionally, the comprehensive risk situation assessment results in Step S30 may include at least: the curve of gas outburst risk level changing over time in the next drilling run, the stuck drill risk coefficient, and the probability distribution of trajectory deviation.

[0053] Step S40: Based on the current risk level obtained in step S30, dynamically configure the objective weights and constraints of the multi-objective optimization problem, and then use the NSGA-III algorithm to solve it online to obtain a Pareto optimal solution set containing multiple optimal solutions under the current risk situation.

[0054] Step S50: Input the optimal solution set obtained in step S40 into the digital twin system for security verification. If a candidate instruction fails the security verification in the digital twin simulation, the specific risk type and degree identified in the simulation are fed back to the multi-objective optimization module, triggering it to dynamically adjust the weights of the optimization objectives and recalculate a new candidate solution set. This "verification-feedback-re-optimization" cycle is repeated until a set of instructions that passes the security verification is obtained.

[0055] Step S60: The instruction set obtained in step S50 that has passed the safety verification is sent to the directional drilling rig and the guiding tool for execution, so as to accurately complete the drilling task of this round. If the current risk situation triggers the preset active intervention strategy, the corresponding pre-borehole pressure relief, pre-jamming jamming prevention or trajectory loss correction program will be automatically activated.

[0056] Step S70: After the current drilling footage is completed, compare the actual drilling results with the prediction results of Step S30 and the optimization expectations of Step S40. Use the deviation data to drive the knowledge base and online learning module to automatically calibrate and update the key parameters in the risk prediction model and optimization model, and store this experience as a new case in the knowledge base, forming a complete "perception-decision-execution-learning" closed loop.

[0057] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0058] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0059] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A smart drilling optimization system for gas extraction holes based on dynamic risk modeling, characterized in that, include: The drilling dynamic sensing module is used to collect multi-source data in real time during the drilling process, including: engineering parameters, geometric parameters, geological and environmental parameters; The dynamic risk modeling and prediction module is connected to the drilling dynamic sensing module. It is used to perform time alignment, filtering and noise reduction, and feature extraction on the collected multi-source data to construct unified time series data. Based on this, a multi-dimensional dynamic risk prediction model including gas blowout risk, stuck drill risk, and trajectory loss of control risk is established, and the evolution trend of drilling risk is output. The multi-objective optimization module, connected to the dynamic risk modeling and prediction module, is used to receive the current risk situation and establish a multi-objective optimization function with blowout prevention, jamming prevention, trajectory control accuracy and drilling efficiency as optimization objectives, and feed rate, rotational speed, tool face angle, pump pressure and pump volume as decision variables. The NSGA-III algorithm is used to solve the problem online and generate a Pareto optimal solution set containing multiple optimal solutions. The digital twin virtual-real interaction verification module is connected to the multi-objective optimization module. It is used to establish a borehole digital twin, perform virtual drilling simulation and safety verification on the Pareto optimal solution set, and feed back the verification results to the multi-objective optimization module. The intelligent decision-making and execution module is connected to the digital twin virtual-real interaction verification module. It is used to comprehensively optimize the results and twin verification conclusions, generate the final control commands, and send them down to the directional drilling rig and guidance tool for execution. Based on the execution feedback, it realizes online updates of the model and strategy.

2. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, The engineering parameters include: drilling pressure, rotation speed, torque, vibration, pump pressure, and pump flow rate; geometric parameters include: dip angle, azimuth angle, hole depth, and tool face angle; geological and environmental parameters include: formation gamma, resistivity, gas concentration, and gas emission rate.

3. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, The dynamic risk modeling and prediction module includes three core sub-models: The gas eruption hazard risk identification submodule employs a Transformer-based encoder network. The encoder input consists of gas concentration, emission velocity, drilling speed, and vibration parameters from the previous N time steps. After processing by the encoder, the gas emission risk index is output through a fully connected layer. and nozzle warning level The model is pre-trained based on historical drilling blowout accident data, using the cross-entropy loss function, with Adam as the optimizer and a learning rate of 0.0001. Dynamic assessment submodule for stuck drill risk: Constructing a system based on real-time torque Mechanical drilling speed Resistance to advancement , slag return rate Vibration spectrum The input is a neural network model; the neural network includes two hidden layers, each with 128 neurons, using the ReLU activation function, and the output layer uses the Sigmoid activation function. The final output is the drill jam risk coefficient. ; The trajectory loss-of-control risk prediction submodule is based on a finite element model of the drill string-rock coupled dynamics model. Inputs include feed rate, rotational speed, tool face angle, pump pressure, and pump flow rate. Output is the predicted trajectory deviation value for the next advance. and probability of loss of control .

4. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 3, characterized in that, The drill string-rock coupled dynamics model uses a simplified beam-column model to describe the buckling behavior of the drill string in the wellbore. The lateral vibration equation of the drill string is expressed as: in, For the bending stiffness of the drill string, This is a lateral displacement. For axial force distribution, The contact stiffness of the hole wall.

5. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, In the multi-objective optimization module, the multi-objective optimization function is: s.t. in, Characterizes the blowout prevention hole target; Characterizes the target for preventing drill bit jamming; Characterize the trajectory control target, weights , Adjust dynamically according to risk; This characterizes the drilling efficiency target; For time-varying feasible regions; Dynamic constraints include real-time upper limits for gas concentration and equipment limits; The multi-objective optimization was solved using the NSGA-III algorithm, with the following parameters: population size 200, maximum number of iterations 500, crossover probability 0.9, and mutation probability 0.

1.

6. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, The digital twin virtual-real interaction verification module includes three core sub-models: Virtual geological model: Based on the initial three-dimensional geological exploration data, and combined with the borehole trajectory data and gamma and resistivity data uploaded in real time by the drilling dynamic sensing module, the three-dimensional geological model of the target area is dynamically corrected by the Bayesian update algorithm, so that the virtual geological model gradually approaches the real geological conditions. Virtual drilling rig - borehole model: A rigid-flexible coupling model of the directional drilling rig, drill rod, drill bit and borehole wall is established using finite element dynamics simulation software to simulate the interaction between the drill bit and the coal seam; Virtual sensor model: Add a virtual sensor with the same characteristics as a real sensor to the simulation environment, and add measurement noise.

7. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, The security verification logic of the digital twin virtual-real interaction verification module is to simulate the execution of control commands in a virtual environment and monitor whether any of the following situations occur during the simulation process: (a) Gas eruption simulation: If the simulated gas concentration is >1.0%, the outflow velocity is >5 m³ / min, and the duration exceeds 3 seconds, the gas eruption risk is triggered; (b) Simulation of stuck drill event: If the simulated torque exceeds 90% of the rated torque of the directional drilling rig, the slag return rate... If the value is below 0.5 and the duration exceeds 5 seconds, it is determined that a stuck drill has occurred. (c) Trajectory runaway simulation: The actual trajectory is calculated based on the drill bit-coal seam interaction model and compared with the design trajectory. If the deviation of the simulated trajectory exceeds 2m, the trajectory is judged to be out of control. If any situation occurs during the simulation, the instruction is determined to have failed the security check, and the specific risk type and degree are fed back to the multi-objective optimization module.

8. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 1, characterized in that, The intelligent decision-making and execution module includes: The instruction parsing and coordination control unit is used to select the solution that matches the current risk situation from the Pareto solution set that has passed the safety verification as the final instruction, that is, when the risk index is high, a solution that is biased towards safety is selected; when the risk index is low, a solution that is biased towards drilling efficiency is selected. Model predictive disturbance rejection controller is used to construct a dynamic response model with the final command as the setpoint, solve for the optimal control law in a rolling manner, and generate the underlying control signal; The execution status monitoring and adaptive compensation unit is used to monitor the response status of the actuator in real time and to activate the adaptive compensation algorithm to eliminate the system deviation when a systematic deviation is detected. The execution result closed-loop feedback interface is used to feed the execution data back to the online learning module to drive model calibration and updates; The risk response strategy library includes at least the following three proactive intervention strategies: proactive pressure relief strategy before nozzle eruption, proactive anti-jamming strategy before drill jamming, and proactive correction strategy for trajectory loss of control.

9. The intelligent drilling optimization system for gas extraction holes based on dynamic risk modeling as described in claim 8, characterized in that, Active pressure relief strategy before the nozzle: When the gas nozzle risk index output by the dynamic risk modeling and prediction module... At that time, the instruction parsing and coordination control unit automatically activates the blowout preventer intervention mode; Proactive anti-jamming strategy before drill bit jamming: When the risk of drill bit jamming is high... If the torque fluctuation exceeds 20% of the normal value and the slag return rate is less than 0.6%, the command parsing and collaborative control unit will automatically activate the anti-jamming intervention mode. Active trajectory loss correction strategy: When the trajectory loss risk prediction submodule outputs the predicted trajectory deviation value... or probability of getting out of control At that time, the instruction parsing and collaborative control unit automatically activates the correction intervention mode.

10. A method for optimizing intelligent drilling of gas drainage boreholes based on the intelligent drilling optimization system for gas drainage boreholes based on dynamic risk modeling as described in any one of claims 1 to 9, characterized in that, Includes the following steps: Step S10: Import the 3D geological model and borehole engineering design model of the target area, initialize the network parameters of the three prediction models of gas venting, stuck drill and trajectory runaway, initialize the digital twin and NSGA-III optimization algorithm parameters, and build an initial knowledge base containing historical cases; Step S20: During the drilling process, the drilling dynamic sensing module is used to collect multi-source data in real time, including drilling pressure, rotation speed, torque, and gas concentration, and to complete outlier removal, time alignment and feature extraction to form a structured time-series data stream. Step S30: Input the data processed in step S20 into the dynamic risk modeling and prediction module, and output the probability distribution of gas blowout risk probability, stuck drill risk coefficient, and trajectory loss of control in the next advance in real time to form a comprehensive risk situation map. Step S40: Based on the current risk level obtained in step S30, dynamically configure the objective weights and constraints of the multi-objective optimization problem, and then use the NSGA-III algorithm to solve it online to obtain a Pareto optimal solution set containing multiple optimal solutions under the current risk situation. Step S50: Input the optimal solution set obtained in step S40 into the digital twin system for security verification; If a candidate instruction fails the security check in the digital twin simulation, the specific risk type and degree identified in the simulation are fed back to the multi-objective optimization module, triggering it to dynamically adjust the weights of the optimization objectives and recalculate a new set of candidate solutions. This verification-feedback-re-optimization cycle is repeated until a set of instructions that passes the security check is obtained. Step S60: Send the instruction set obtained in step S50 that has passed the safety verification to the directional drilling rig and the guiding tool for execution, so as to accurately complete the drilling task of this round of footage. If the current risk situation triggers the preset active intervention strategy, the corresponding pre-borehole pressure relief, pre-jamming jamming prevention or trajectory loss correction program will be automatically activated. Step S70: After the current drilling footage is completed, compare the actual drilling results with the predicted results of step S30 and the optimization expectations of step S40. By leveraging the deviation data between the two, the knowledge base and online learning module are driven to automatically calibrate and update the key parameters in the risk prediction model and optimization model, and this experience is stored in the knowledge base as a new case, forming a complete closed loop of perception-decision-execution-learning.

Images