A rotary steerable drilling system dynamics analysis system

Abstract

The application relates to the field of rotary steerable drilling system dynamics analysis and control technology, in particular to a rotary steerable drilling system dynamics analysis system, which comprises a data acquisition module, a digital twin modeling module, a noise prediction module, a signal reconstruction module and a cooperative control module. The data acquisition module is used for collecting ground engineering parameters and downhole dynamic parameters in real time and intercepting drilling parameter adjustment instructions. The digital twin modeling module is used for outputting a complete state vector of a digital twin model. The noise prediction module is used for generating a control source noise signal. The signal reconstruction module is used for performing adaptive hedging processing on an original mixed signal to reconstruct a pure geological signal. The cooperative control module is used for predicting vibration risks caused by changes in front strata and generating optimal control instructions for actively avoiding the vibration risks. The application significantly reduces the occurrence probability of malignant vibration, improves drilling efficiency and prolongs the service life of drilling tools.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of dynamic analysis and control of rotary steerable drilling systems, in particular to a dynamic analysis system for rotary steerable drilling systems. BACKGROUND

[0002] In modern rotary steerable drilling engineering, the downhole measurement-while-drilling system is the core technology for realizing accurate geological steering and ensuring operation safety. In order to effectively control the drilling process, the system needs to obtain real-time downhole dynamic parameters reflecting the characteristics of the formation. However, in actual operation, drilling parameter adjustment commands, such as speed increase or pressure increase, issued from the ground will generate and propagate vibrations in the drill string system, forming a certain control source noise. This noise is superimposed with the target geological signal, forming a severely contaminated original mixed signal.

[0003] Existing technical methods mostly adopt passive signal filtering or vibration suppression strategies. These methods process the signal after it has been contaminated, making it difficult to effectively separate the control source noise that overlaps with the geological signal frequency band, resulting in insufficient accuracy and reliability of geological steering. At the same time, due to the inability to obtain pure geological signals, the system cannot accurately predict changes in the formation ahead, thus failing to avoid the risk of stick-slip, impact, and other malignant vibrations caused by sudden changes in the formation. The control strategy is limited to passive response, contributing little to improving drilling efficiency and safety. Therefore, how to establish a dynamic analysis and control method that can prospectively predict and actively eliminate control source noise, reconstruct high-fidelity pure geological signals from a strong noise background, and implement control strategy upgrades from passive vibration suppression to active risk avoidance, is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0004] To solve the above technical problems, the present application provides a dynamic analysis system for rotary steerable drilling systems. Specifically, the technical solution of the present application includes:

[0005] A data acquisition module for real-time acquisition of ground engineering parameters and downhole dynamic parameters, and interception of drilling parameter adjustment commands.

[0006] A digital twin modeling module for constructing and real-time correcting a drill string system dynamics digital twin model based on ground engineering parameters and downhole dynamic parameters, and then outputting a complete state vector of the digital twin model.

[0007] A noise prediction module for calling the digital twin model based on the drilling parameter adjustment commands and the complete state vector to perform forward deduction, thereby generating a control source noise signal.

[0008] a signal reconstruction module, configured to obtain an original mixed signal from the downhole dynamic parameters, and perform adaptive hedging processing on the original mixed signal based on a control source noise signal, to reconstruct a pure geological signal;

[0009] a collaborative control module, configured to predict a vibration risk caused by a change in a formation ahead based on the pure geological signal, and generate optimal control instructions for actively avoiding the vibration risk.

[0010] Preferably, the digital twin modeling module is specifically configured to:

[0011] based on preset parameters of a drill string assembly geometry, material properties and drilling fluid performance, initialize construction of a physical dynamics model with a state space equation as the core;

[0012] adopt a Kalman filter, take real-time collected ground engineering parameters and downhole dynamic parameters as inputs, continuously correct and update the state of the digital twin model by minimizing the error between the model prediction output and the measured value.

[0013] Preferably, the noise prediction module is specifically configured to:

[0014] take the intercepted drilling parameter adjustment instruction as input, and call the complete state vector of the digital twin model at the current time as the initial condition;

[0015] through the transfer function built-in the digital twin model, solve the vibration waveform caused by the drilling parameter adjustment instruction at the position of the downhole sensor, and define the vibration waveform as the control source noise signal.

[0016] Preferably, the signal reconstruction module is specifically configured to:

[0017] use a clock synchronization mechanism to align the control source noise signal and the original mixed signal on the time axis;

[0018] subtract the weighted component of the aligned control source noise signal from the original mixed signal through an adaptive hedging algorithm, to obtain the pure geological signal.

[0019] Preferably, the adaptive hedging algorithm is a least mean square algorithm, and the signal reconstruction module is further configured to:

[0020] according to the principle of minimizing the output signal energy, use the iteration rule of the least mean square algorithm to update the adaptive hedging coefficient for weighting the control source noise signal in real time.

[0021] Preferably, the collaborative control module is specifically configured to:

[0022] input the pure geological signal into a preset lithology-mechanical conversion model to determine a predicted bottom hole torque sequence;

[0023] The predicted bottom hole torque sequence is input into the digital twin model as a future load for forward simulation to obtain a future dynamic response prediction;

[0024] Based on the future dynamic response prediction, a vibration risk index is calculated;

[0025] Based on the vibration risk index, an optimal control instruction is solved and generated by a multi-objective optimizer.

[0026] Preferably, the lithology-mechanical conversion model is a neural network model trained based on machine learning, which is used to predict the drillability or rock strength equivalent parameter of the formation about to be drilled by identifying the characteristic pattern in the pure geological signal, to output the predicted bottom hole torque sequence.

[0027] Preferably, the multi-objective optimizer takes minimizing the future vibration risk index and maximizing the drilling efficiency as optimization objectives, solves a set of drilling parameter adjustment instructions, and defines the instruction as the optimal control instruction.

[0028] Compared with the prior art, the present application has the following beneficial effects:

[0029] 1. The system predicts the deterministic noise waveform generated by the surface control instruction downhole in advance by constructing a drilling string system dynamics digital twin model, and on this basis, actively hedges the noise from the original mixed signal to reconstruct a high-fidelity pure geological signal, greatly improving the accuracy and reliability of geological steering;

[0030] 2. The system uses the reconstructed pure geological signal to predict the lithology change trend of the front formation, and combines the digital twin model for forward simulation, which enables the system to predict the stick-slip, impact and other malignant vibration risks caused by formation changes in advance, realizing the upgrade of the control strategy from passive vibration suppression to active risk avoidance;

[0031] 3. The system establishes an intelligent closed-loop control link from geological signal analysis, vibration risk prediction to optimal control instruction generation, and through the multi-objective optimizer, seeks the best balance between avoiding vibration risks and improving drilling efficiency, thereby significantly reducing the probability of malignant vibration, improving drilling efficiency and prolonging the service life of drilling tools;

[0032] 4. The system combines modeling based on physical mechanism and Kalman filter correction based on real-time data to construct a high-fidelity digital twin model with mechanism accuracy and real-time adaptability, which can accurately reproduce the real dynamic behavior of the downhole drilling string, providing a core foundation for accurate noise prediction and reliable vibration simulation. BRIEF DESCRIPTION OF DRAWINGS

[0033] The application will be further explained in connection with the accompanying drawings and embodiments:

[0034] Figure 1 is a structural diagram of the system of the application. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions and advantages of the application clearer, the application will be further described in detail below with reference to specific embodiments.

[0036] Embodiment 1

[0037] Please refer to Figure 1 A rotary steerable drilling system dynamics analysis system comprises:

[0038] A data acquisition module is configured to acquire ground engineering parameters and downhole dynamic parameters in real time and intercept drilling parameter adjustment instructions.

[0039] A digital twin modeling module is configured to construct and correct a drilling string system dynamics digital twin model in real time based on the ground engineering parameters and the downhole dynamic parameters, and then output a complete state vector of the digital twin model.

[0040] A noise prediction module is configured to call the digital twin model to perform forward deduction based on the drilling parameter adjustment instructions and the complete state vector, so as to generate a control source noise signal.

[0041] A signal reconstruction module is configured to acquire an original mixed signal from the downhole dynamic parameters, and perform adaptive hedging processing on the original mixed signal based on the control source noise signal, so as to reconstruct a pure geological signal.

[0042] A cooperative control module is configured to predict vibration risks caused by changes in front strata based on the pure geological signal, and generate optimal control instructions for actively avoiding the vibration risks.

[0043] The embodiment provides a rotary steerable drilling system dynamics analysis system; in a specific implementation scenario, the system is deployed in a ground control center, and comprises a data acquisition module, a digital twin modeling module, a noise prediction module, a signal reconstruction module and a cooperative control module; these modules work cooperatively to form a complete technical link from data acquisition, model construction, signal processing to closed-loop control.

[0044] The purpose of the data acquisition module is to provide real-time and comprehensive data input for the entire analysis system. It continuously acquires the drive torque and speed of the drive motor through an interface with the drilling rig's surface control system, and obtains surface engineering parameters such as drilling pressure through sensors in the hydraulic system. These surface engineering parameters are macroscopic physical quantities that can be directly measured on the surface and reflect the operating status of the drilling equipment; their function is to serve as the main driving input for the digital twin model. This module receives and analyzes downhole dynamic parameters transmitted from the downhole Measurement While Drilling (MWD) system via a mud pulse telemetry decoding unit. A key function of this module is to monitor the control system's command bus in real time to intercept drilling parameter adjustment commands, such as speed-up or pressurization commands issued by operators or the automatic control system. This instruction is the direct source of subsequent noise prediction;

[0045] The purpose of the digital twin modeling module is to build and maintain a virtual model that can accurately reproduce the dynamic behavior of the actual downhole drill string system. This module receives surface engineering parameters and downhole dynamic parameters from the data acquisition module, and builds and corrects the drill string system dynamics digital twin model based on this real-time data. The core of this model is a state-space expression that can calculate the angular displacement and angular velocity at any position on the drill string with millisecond-level time resolution, thereby outputting the complete state vector of the digital twin model. This state vector is a comprehensive description of the current dynamic state of the entire drill string system and is the basis for subsequent noise prediction and vibration simulation.

[0046] The purpose of the noise prediction module is to transform passive noise response into proactive prediction, providing crucial prior information for subsequent signal cleanup. This module receives drilling parameter adjustment commands intercepted by the data acquisition module. And obtain the real-time complete state vector provided by the digital twin modeling module. It uses a digital twin model to perform forward simulation based on drilling parameter adjustment commands and a complete state vector, simulating how the control commands will propagate through the drill string system and ultimately generate vibrations at the downhole LWD sensor location. The simulation result is defined as a time-series signal, i.e., the control source noise signal. ;

[0047] The purpose of the signal reconstruction module is to accurately reconstruct a clean signal reflecting true formation information from severely contaminated downhole measurement signals; this module obtains the original mixed signal from downhole dynamic parameters. This signal is a superposition of geological information and various types of noise; it receives the control source noise signal generated by the noise prediction module. Based on the predicted signal, adaptive hedging processing is performed on the original mixed signal; through an adaptive filter, from Subtract precisely components, and finally reconstruct the pure geology signal ;

[0048] The purpose of the synergistic control module is to use the high-quality information processed by the front-end module to realize the upgrade of the control strategy from passive vibration suppression to active vibration avoidance; based on the pure geology signal output by the signal reconstruction module , the vibration risk caused by the change of the front formation is predicted; by analyzing the pattern, the upcoming change in formation hardness is predicted, and the stick-slip, impact and other malignant vibrations that may be caused by the change under the current drilling parameters are simulated; the optimizer in the system calculates a set of optimal control parameters that can avoid the risk in advance, and generates optimal control instructions for actively avoiding vibration risks , which are issued to the drilling rig for execution;

[0049] In addition, the system also integrates a health monitoring module for real-time assessment of the prediction error of the digital twin model and the quality of the input signal. When the error exceeds the preset threshold or the sensor data is abnormal, the system will trigger an alarm and can optionally switch to a more conservative control mode or return control to humans to ensure the safety and robustness of the entire drilling operation;

[0050] Through the organic combination of the above-mentioned modules, this embodiment constructs a closed-loop dynamic analysis and control system, which not only extracts high-fidelity geology signals from strong noise background, greatly improving the accuracy and reliability of geosteering, but also significantly reduces the probability of malignant vibration through forward-looking vibration risk prediction and active avoidance, thereby improving drilling efficiency, prolonging the service life of drilling tools, and ensuring operation safety. The system focuses on predicting and eliminating deterministic noise caused by surface control adjustment, and other noise sources with stronger randomness can be further processed through subsequent conventional filtering means.

[0051] Embodiment 2:

[0052] The digital twin modeling module is specifically used for:

[0053] Based on the preset drilling tool combination geometry, material properties and drilling fluid performance parameters, a physical dynamics model with a state space equation as the core is initialized and constructed;

[0054] Using a Kalman filter, real-time collected surface engineering parameters and downhole dynamic parameters are used as inputs to continuously correct and update the state of the digital twin model by minimizing the error between the model prediction output and the measured value.

[0055] This embodiment is a specific implementation of the digital twin modeling module described in embodiment 1;

[0056] The digital twin modeling module performs initialization construction before the drilling operation starts; it initializes the construction of a physical dynamics model with a state space equation as the core based on the preset BHA geometric configuration, material properties and drilling fluid performance parameters; the state space equation is a modern control theory mathematical model for describing the dynamic behavior of a physical system, which has the following form:

[0057]

[0058]

[0059] The physical meaning of the state vector is the collection of angular displacement and angular velocity of each discrete node along the drill string, which is a complete description of the internal dynamics of the system; the input vector is from the data acquisition module and includes the ground driving torque and the torque of the interaction between the drill bit and the rock at the bottom of the well based on the drilling pressure and lithology estimation; the output vector corresponds to the measurable physical quantities of the ground sensor, such as the ground speed, which is used for comparison with the measured value; the initial value of the parameter matrix is derived from the finite element analysis (FEA) calculation results based on the BHA design drawing, which represents the internal physical characteristics of the system.

[0060] During drilling, to cope with the changing downhole conditions, the module uses a Kalman filter to take the real-time collected ground engineering parameters and downhole dynamic parameters as inputs, continuously corrects and updates the state of the digital twin model by minimizing the error between the model predicted output and the measured value; the Kalman filter is an efficient recursive state estimator that can estimate the state of a dynamic system in a series of incomplete and noisy measurements; its role is to compare the predicted output value of the model, such as the predicted ground speed with the actual measured value of the sensor, and use the error between the two to adjust the state vector in real time, so as to ensure that the digital twin model is always highly consistent with the real downhole drill string dynamics behavior;

[0061] Through this two-step method of offline initialization and online correction, the digital twin model constructed in this embodiment has both the mechanism accuracy of the physical model and the real-time adaptability of the data-driven model, and can more accurately reflect the real working conditions, providing a high-fidelity basis for subsequent noise prediction and vibration simulation, thereby improving the analysis accuracy of the entire system.

[0062] Embodiment 3:

[0063] The noise prediction module is specifically used for:

[0064] The intercepted drilling parameter adjustment instruction is taken as input, and the complete state vector of the digital twin model at the current time is called as an initial condition;

[0065] Through the transfer function built in the digital twin model, the vibration waveform caused by the drilling parameter adjustment instruction at the downhole sensor position is calculated, and the vibration waveform is defined as the control source noise signal.

[0066] This embodiment is a specific implementation of the noise prediction module described in Embodiment 1;

[0067] When the data acquisition module intercepts a drilling parameter adjustment instruction to be issued, the noise prediction module is immediately started; it takes the intercepted drilling parameter adjustment instruction as input, which is a vector causing a change in system input ; At the same time, it calls the complete state vector of the digital twin model at the current time as an initial condition, which is provided by the digital twin modeling module in real time and represents the inertial state of the entire drill string system at the moment before the control instruction takes effect;

[0068] The module calculates the vibration waveform caused by the drilling parameter adjustment instruction at the downhole sensor position by forward deduction of the state space equation of the digital twin model, and the new input vector in this process is formed by superimposing the original input vector and the drilling parameter adjustment instruction This process can be mathematically expressed as:

[0069]

[0070] wherein, is the future state vector obtained by solving the following differential equation set:

[0071]

[0072] wherein is the vibration noise time series signal generated at the LWD sensor predicted and generated; the control parameter adjustment instruction intercepted is taken as the future input, and the complete state vector of the digital twin model at the current time is taken as the simulation initial condition, and the downhole vibration response caused by the instruction is calculated by forward deduction of the state space equation; the module defines the calculated vibration waveform as the control source noise signal , and transmits it to the signal reconstruction module;

[0073] The embodiment realizes accurate prediction of the control source noise by forward deduction of the intercepted control instruction as a future input of the digital twin model; the feedforward prediction mechanism enables the system to obtain accurate waveform samples of the noise before the noise actually pollutes the measurement signal, thereby providing a high-quality reference signal for subsequent adaptive hedging.

[0074] Embodiment 4:

[0075] The signal reconstruction module is specifically configured to:

[0076] The clock synchronization mechanism is used to align the control source noise signal and the original mixed signal on the time axis;

[0077] The adaptive hedging algorithm is used to subtract the weighted component of the aligned control source noise signal from the original mixed signal to obtain a pure geological signal;

[0078] The adaptive hedging algorithm is a least mean square algorithm, and the signal reconstruction module is further configured to:

[0079] According to the principle of minimizing the energy of the output signal, the adaptive hedging coefficient for weighting the control source noise signal is updated in real time using the iteration rule of the least mean square algorithm.

[0080] The embodiment is a specific implementation of the signal reconstruction module described in Embodiment 1, which uses an adaptive hedging technology based on the least mean square (LMS) algorithm;

[0081] The signal reconstruction module receives an original mixed signal and a control source noise signal ; to achieve accurate hedging, the module uses a clock synchronization mechanism to align the control source noise signal and the original mixed signal on the time axis; the clock synchronization mechanism ensures that the time stamps of the surface computing unit and the downhole measurement unit are consistent through GPS or other high-precision clock sources, and performs accurate time translation on according to the calibrated total time delay of mud pulse transmission to obtain ;

[0082] After alignment, the module uses an adaptive hedging algorithm to subtract the weighted component of the aligned control source noise signal from the original mixed signal to obtain a pure geological signal; as defined in Embodiment 5, in this embodiment, the adaptive hedging algorithm is a least mean square algorithm, and the actual signal reconstruction process is performed by the following core equation:

[0083]

[0084] wherein is the pure geological signal, which is the key output of the module; is the original mixed signal received from downhole; is time-delayed the aligned predicted noise signal; is the adaptive damping coefficient, a dimensionless, time-varying weighting parameter whose value is updated in real-time by the LMS algorithm;

[0085] To optimize the damping effect, the signal reconstruction module is further configured to update the adaptive damping coefficient for weighting the control source noise signal in real-time according to the principle of minimizing the energy of the output signal, using the iterative rule of the least mean square (LMS) algorithm; the LMS algorithm is a classical adaptive filtering algorithm, whose purpose is to adjust the energy of the reconstructed signal (i.e. the error signal) to be as small as possible through iteration, and its update expression in the discrete-time system is:

[0086]

[0087] wherein, is the updated coefficient value at the next time instant; is the coefficient value at the current time instant, which is a dimensionless parameter; is the reconstructed signal at the current time instant, also known as the error signal in adaptive filtering theory; is the time-delayed aligned predicted noise signal, where is the continuous-time delay is the equivalent delay under discrete sampling;

[0088] To ensure that the iterative formula is strictly consistent in physical dimension, the dimension of the step convergence parameter is set as the inverse of the dimension of the signal product; for example, if the units of the signals and are both volts (V), the unit of is , so that the update term is dimensionless and can be algebraically operated with the dimensionless coefficient ; The optimal value of is determined offline: a calibration dataset containing multiple segments of historical original mixed signals and their corresponding known control source noise signals is used to test a series of candidate values of ; for each candidate value, the LMS algorithm is applied to the entire calibration dataset, and the average residual energy of the output signal is calculated; finally, the value of that minimizes the average residual energy is selected as the optimal step convergence parameter, which ensures the balance between the convergence speed and the steady-state error of the algorithm; the purpose of this iteration process is to continuously adjust values, such that the energy of the output signal tends to be minimized, thereby achieving optimal noise hedging;

[0089] By adopting the LMS adaptive hedging algorithm, the embodiment can dynamically and real-timely adjust the strength of noise hedging, so that the signal reconstruction process has strong robustness and adaptive ability, thereby obtaining a pure geological signal with clearer geological features after significantly suppressing the control source noise.

[0090] Embodiment 5:

[0091] The collaborative control module is specifically used for:

[0092] inputting the pure geological signal into a preset lithology-mechanical conversion model to determine a predicted bottom hole torque sequence;

[0093] inputting the predicted bottom hole torque sequence as a future load into a digital twin model for forward simulation to obtain a future dynamic response prediction;

[0094] calculating a vibration risk index based on the future dynamic response prediction;

[0095] solving and generating optimal control instructions through a multi-objective optimizer based on the vibration risk index;

[0096] The lithology-mechanical conversion model is a neural network model trained based on machine learning, which is used to predict the drillability or rock strength equivalent parameter of the formation to be drilled by identifying the feature mode in the pure geological signal, to output the predicted bottom hole torque sequence;

[0097] The multi-objective optimizer takes minimizing the future vibration risk index and maximizing the drilling efficiency as optimization objectives, solves a set of drilling parameter adjustment instructions, and defines the instructions as the optimal control instructions.

[0098] The embodiment is a specific implementation of the collaborative control module described in Embodiment 1, which integrates machine learning prediction, forward simulation and multi-objective optimization technology to form an intelligent closed-loop control core;

[0099] The collaborative control module receives the high-fidelity pure geological signal output by the signal reconstruction module ; inputting the pure geological signal into a preset lithology-mechanical conversion model to determine a predicted bottom hole torque sequence ; the lithology-mechanical conversion model is a neural network model trained based on machine learning; in the embodiment, the neural network model is a long short-term memory network LSTM, which identifies the feature mode in the pure geological signal through a large number of historical The data is supervised learning trained with corresponding rock mechanics test data; the model is used to predict the drillability or rock strength equivalent parameters of the formation to be drilled by identifying the characteristic patterns in the pure geological signal, to output the predicted bottom hole torque sequence;

[0100] The module obtains the predicted bottom hole torque sequence in the last step as a future load, input into the digital twin model for forward simulation; forward simulation refers to simulating the system dynamic behavior in a future short time window when the drill bit encounters the formation represented by the bottom hole torque sequence ; after simulation, the module obtains the future dynamic response prediction;

[0101] Based on the future dynamic response prediction, the module calculates the vibration risk index ; the vibration risk index is a quantitative index, which defines the amplitude, frequency and duration of the predicted vibration fluctuation, etc. multiple dimensions, used to evaluate the possibility and severity of malignant vibration such as stick-slip and impact; for example, the vibration risk index can be calculated by weighted integration of the normalized predicted angular velocity in the future time window , and one specific form can be expressed as: where, is the average angular velocity in the prediction time window, the first term penalizes the fluctuation of angular velocity, and the second term penalizes the sharp change of angular acceleration, and are the normalization coefficients of angular velocity and angular acceleration, respectively, which can be taken as the standard deviation in historical data or the maximum allowable fluctuation value in engineering, for example, to eliminate the dimension effect, so that both terms in the square brackets are dimensionless, and are dimensionless preset weight coefficients, satisfying ;

[0102] Based on the vibration risk index, the module generates optimal control instructions by a multi-objective optimizer, for example, using non-dominated sorting genetic algorithm NSGA-II to solve and generate; the multi-objective optimizer minimizes the future vibration risk index and maximizes the drilling efficiency, which is defined as the predicted rate of penetration ROP as the optimization objective; the multi-objective optimizer finds a Pareto optimal solution between the two mutually restrictive objectives; the optimizer solves a set of drilling parameter adjustment instructions , and defines the instruction as the optimal control instruction, which is issued to the drilling rig control system for execution;

[0103] To ensure the effectiveness of the vibration risk index, the weight coefficients and Calibration is done by building a database containing multiple sets of historical severe vibration events, annotating each event with its severity level, and iterating over a range of candidate combinations of and , e.g. from 0.1 to 0.9 with a step of 0.1, for each set of weights, compute the predicted vibration risk index for all historical events , select the set of weights that maximizes the discrimination between the risk indices of vibration events of different severity levels as the optimal calibration parameters;

[0104] The closed-loop link of this embodiment, which predicts future loads from geologic signals, simulates future vibrations with digital twins, and generates avoidance strategies with multi-objective optimization, upgrades drilling control from the traditional passive response based on current state to proactive avoidance based on future prediction, enabling the anticipation of risks before severe vibrations occur and the adoption of optimal and gradual control actions to resolve the risks in advance.

[0105] It should be noted that the above-mentioned embodiments are only used to illustrate but not limit the technical solutions of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.​

Claims

1. A dynamic analysis system for a rotary steerable drilling system, characterized in that, include: The data acquisition module is used to collect surface engineering parameters and downhole dynamic parameters in real time, and to intercept drilling parameter adjustment commands. The digital twin modeling module is used to construct and correct the dynamic digital twin model of the drill string system in real time based on surface engineering parameters and downhole dynamic parameters, and then output the complete state vector of the digital twin model. The noise prediction module is used to generate a control source noise signal by invoking a digital twin model for forward extrapolation based on drilling parameter adjustment commands and a complete state vector. Specifically, the noise prediction module is used to: take the intercepted drilling parameter adjustment commands as input and call the complete state vector of the digital twin model at the current moment as initial conditions; calculate the vibration waveform caused by the drilling parameter adjustment commands at the downhole sensor positions through the transfer function built into the digital twin model, and define the vibration waveform as the control source noise signal. The signal reconstruction module is used to obtain the original mixed signal from downhole dynamic parameters and perform adaptive offsetting processing on the original mixed signal based on the control source noise signal to reconstruct a pure geological signal. Specifically, the signal reconstruction module is used to: align the control source noise signal and the original mixed signal on the time axis using a clock synchronization mechanism; and subtract the weighted component of the aligned control source noise signal from the original mixed signal using an adaptive offsetting algorithm to obtain a pure geological signal. The adaptive offsetting algorithm is a least mean square algorithm. The collaborative control module is used to predict the vibration risk caused by changes in the formation ahead based on the pure geological signal, and generate the optimal control command for actively avoiding the vibration risk. Specifically, the collaborative control module is used to: input the pure geological signal into the preset lithology-mechanics conversion model, and predict the drillability or rock strength equivalent parameters of the formation to be encountered by drilling by identifying the characteristic patterns in the pure geological signal, so as to output the predicted bottom hole torque sequence. The predicted bottom hole torque sequence is used as the future load and input into the digital twin model for forward-looking simulation to obtain the future dynamic response prediction. Based on the future dynamic response prediction, the vibration risk index is calculated by weighted integration of the normalized predicted angular velocity within the future time window. Based on the vibration risk index, a multi-objective optimizer is used to solve for a set of drilling parameter adjustment commands with the optimization objectives of minimizing the future vibration risk index and maximizing drilling efficiency. These commands are then defined as the optimal control commands.

2. The dynamic analysis system for a rotary steerable drilling system according to claim 1, characterized in that, The digital twin modeling module is specifically used for: Based on the preset geometric configuration, material properties and drilling fluid performance parameters of the drill string assembly, a physical dynamic model with state-space equations as its core is initialized and constructed. Using a Kalman filter, real-time acquired surface engineering parameters and downhole dynamic parameters are used as inputs. By minimizing the error between the model's predicted output and the measured values, the state of the digital twin model is continuously corrected and updated.

3. The dynamic analysis system for a rotary steerable drilling system according to claim 1, characterized in that, The signal reconstruction module is also used for: Based on the principle of minimizing the output signal energy, the adaptive offset coefficient used to weight the control source noise signal is updated in real time using the iterative rules of the least mean square algorithm.

4. The dynamic analysis system for a rotary steerable drilling system according to claim 1, characterized in that, The lithology-mechanics conversion model is a neural network model trained based on machine learning.

Images