A drilling casing deformation prevention and control method, system, computer device and medium

Abstract

The application provides a drilling casing deformation prevention and control method, system, computer device and medium, and belongs to the technical field of drilling engineering. The method comprises the following steps: collecting drilling dynamic time series data and geological static data, using a deep learning model to extract local features and capture long-range dependencies after spatiotemporal fusion, combining the physical constraints of a three-dimensional geomechanical model to accurately predict the slip amount of potential fracture surfaces around the well. Then, a three-dimensional finite element forward model is constructed by taking the predicted slip amount as the displacement boundary and the in-situ stress field as the load boundary, and the cement ring elastic modulus and thickness are taken as design variables to construct an optimization problem with the minimum casing mechanical response as the target and the engineering feasible range as the constraint, and the optimal parameter combination is automatically output after solving. The method realizes accurate prediction of geological load and efficient optimization of cement ring parameters, significantly improves the reliability and adaptability of casing deformation prevention and control under complex geology, and ensures drilling safety.

Description

Technical Field

[0001] This invention belongs to the field of drilling engineering technology, specifically relating to a method, system, computer equipment, and medium for preventing and controlling drilling casing deformation. Background Technology

[0002] In oil and gas, geothermal, and other drilling projects, the casing serves as a core structure for wellbore stability and formation isolation; its integrity directly determines the success or failure of drilling operations and subsequent production safety. However, drilling areas often contain complex geological structures such as faults and fractures. Affected by dynamic changes in the geostress field and potential slippage of fracture surfaces, the casing is prone to non-uniform loads, leading to deformations such as bending, compression, and tension. In severe cases, this can cause casing rupture and oil and gas leaks, resulting not only in huge economic losses but also potential geological disasters and environmental risks.

[0003] The core approach of existing casing deformation control technology is to combine parameter optimization methods with finite element simulation. By constructing a simplified mechanical model, the mechanical response of the casing under different combinations of cement sheath elastic modulus and thickness is simulated, and a relatively better parameter combination is selected. However, existing finite element simulation-based technical solutions still have significant limitations under complex geological conditions: on the one hand, the accuracy of load prediction is insufficient, and there is a lack of reliable basis for determining key loads such as wellbore fracture surface slippage, often relying on empirical values, resulting in deviations between the load conditions used in the simulation and actual drilling conditions, thus limiting the reliability of the simulation results; on the other hand, the parameter optimization method is singular and inefficient, usually only performing trial calculations around a few cement sheath parameters, with limited control dimensions, making it difficult to fully adapt to the stress characteristics of the casing under complex loads. At the same time, traditional finite element simulation requires calculating multiple sets of parameter combinations one by one, which is time-consuming, and the selection of parameter combinations is easily influenced by experience, which may miss better solutions, resulting in a mismatch between the optimization effect and engineering requirements. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method, system, computer equipment, and medium for preventing and controlling drilling casing deformation.

[0005] To achieve the above objectives, the present invention provides a method for controlling drilling casing deformation, comprising: The system collects dynamic time-series data of the target well and static geological data of the area where the target well is located. The static geological data includes the distance of the well from the fault and fracture, the geostress field value, and rock mechanics parameters. The dynamic time-series data includes the drilling pressure, torque, pump pressure, rotational speed, and flow rate of the target well.

[0006] The dynamic time-series data and geological static data are spatiotemporally fused according to time windows to obtain spatiotemporally fused data. Local time-series features of the spatiotemporally fused data are extracted, and the long-range time-series dependencies of the local time-series features are captured. Key time sequences related to preset risk parameters are located in the dependencies. The preset risk parameters include the fracture activation probability threshold, fracture slip warning value, and construction pressure abnormal deviation rate of the target well. Based on the key time sequences and the physical constraints provided by the preset three-dimensional geomechanical model, the predicted slip of the potential fracture surface around the target well is determined.

[0007] Using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition, a three-dimensional finite element mechanical forward model is constructed for the wellbore structure of the target well. The elastic modulus and thickness of the cement sheath between the wellbore and the casing are used as design variables to obtain the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model for each design variable. With minimizing the casing mechanical response value as the optimization objective and the engineering feasible range of the design variables as constraints, a constrained optimization problem is constructed, solved, and the optimal elastic modulus and thickness of the cement sheath are output. The engineering feasible range includes the lower and upper limits of the elastic modulus of the cement sheath and the thickness of the cement sheath. The lower and upper limits are predetermined based on the test data of the cement slurry material properties and the wellbore geometry of the target well.

[0008] The construction control of the target drilling casing is carried out by utilizing the optimal elastic modulus and thickness of the cement sheath.

[0009] Preferably, before performing spatiotemporal fusion processing of the dynamic time-series data and the geological static data according to the time window, the method further includes: using the variational mode decomposition (VMD) method to decompose the dynamic time-series data for noise reduction, selecting effective modal components for reconstruction, and obtaining the denoised dynamic time-series data; the spatiotemporal fusion processing specifically involves: copying the geological static data along the time dimension and splicing it with the dynamic time-series data within the same time window in terms of feature dimensions to form the spatiotemporal fused data.

[0010] Preferably, local temporal features of the spatiotemporal fusion data are extracted using a temporal convolutional network (TCN), long short-term memory (LSTM) networks are used to capture the long-range temporal dependencies of the local temporal features, and an attention mechanism is used to locate the key time sequences related to preset risk parameters in the dependencies. The key time sequences are then substituted into the physical constraints of the three-dimensional geomechanical model, and the TCN-LSTM-Attention hybrid neural network model outputs the predicted slip of the potential fracture surface around the target well.

[0011] Preferably, the preset three-dimensional geomechanical model is constructed based on seismic interpretation data, well logging data and core experimental data of the target drilling area. The model range is a cylindrical area with a radius of 50-200m centered on the wellbore, and the grid division accuracy is 0.1m-1m.

[0012] Preferably, obtaining the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model under each design variable specifically includes: sampling using the Latin hypercube sampling method within the engineering feasible space of the design variables to generate multiple sample combinations; running the three-dimensional finite element mechanical forward model for each sample combination to calculate the corresponding casing equivalent stress, strain mechanical response values, and displacement mechanical response values, forming a training dataset; training an artificial intelligence proxy model using the training dataset, and obtaining the predicted casing mechanical response value under any design variable combination through the proxy model; the artificial intelligence proxy model is a deep neural network CNN, support vector regression SVR, or Gaussian process regression GPR model.

[0013] Preferably, after the construction control of the target drilling casing is performed, the method further includes: collecting actual effect data after the construction of the target drilling cement sheath, associating it with the corresponding predicted slip and optimal cement sheath parameters and storing it in a knowledge base; when the case data accumulated in the knowledge base reaches a preset scale, triggering the retraining of the artificial intelligence agent model to update the model parameters.

[0014] Preferably, the construction control of the target drilling casing using the optimal elastic modulus and thickness of the cement sheath specifically includes: if the predicted mechanical response value of the casing under the optimal elastic modulus and thickness of the cement sheath is still higher than the safety threshold, then a geometric compensation mechanism is activated. The geometric compensation mechanism includes back-calculating the required cement sheath compensation thickness based on the target response value, simultaneously adjusting the amount of cement slurry, and generating a wellbore enlargement construction plan.

[0015] The present invention also provides a drilling casing deformation prevention and control system, comprising: The data acquisition module is used to acquire dynamic time-series data of the target well and static geological data of the area where the target well is located. The static geological data includes the distance of the well from the fault and fracture, the geostress field value, and rock mechanics parameters. The dynamic time-series data includes the drilling pressure, torque, pump pressure, rotational speed, and flow rate of the target well.

[0016] The prediction module is used to perform spatiotemporal fusion processing on the dynamic time-series data and geological static data according to time windows to obtain spatiotemporal fused data; extract local time-series features of the spatiotemporal fused data, capture the long-range time-series dependencies of the local time-series features, and locate the key time sequence related to preset risk parameters in the dependencies, wherein the preset risk parameters include the fracture activation probability threshold of the target well, the fracture slip volume warning value, and the construction pressure abnormal deviation rate; based on the key time sequence and the physical constraints provided by the preset three-dimensional geomechanical model, determine the predicted slip volume of the potential fracture surface around the target well.

[0017] The application module is used to construct a three-dimensional finite element mechanical forward model for the wellbore structure of the target well, using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition; using the elastic modulus and thickness of the cement sheath between the wellbore and the casing as design variables, it obtains the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model under each design variable; with minimizing the casing mechanical response value as the optimization objective and the engineering feasible range of the design variables as constraints, it constructs a constrained optimization problem, solves the constrained optimization problem, and outputs the optimal cement sheath elastic modulus and thickness; wherein, the engineering feasible range includes the lower and upper limits of the cement sheath elastic modulus and the cement sheath thickness; the lower and upper limits are predetermined based on the cement slurry material performance test data and wellbore geometry of the target well; and the construction control of the target well casing is performed using the optimal cement sheath elastic modulus and thickness.

[0018] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in the drilling casing deformation prevention and control method.

[0019] The present invention also provides a computer-readable storage medium storing a computer program, which, when loaded by a processor, is capable of executing any of the steps in the drilling casing deformation prevention and control method.

[0020] The drilling casing deformation control method provided by this invention has the following beneficial effects: This invention achieves accurate prediction of the slip volume of potential fracture surfaces around the well by spatiotemporally fusing drilling dynamic time-series data with geological static data, and extracting key time-series features and capturing long-range dependencies using a deep learning model. At the same time, it introduces the physical constraints of a three-dimensional geomechanical model. This fundamentally solves the problems of inaccurate load prediction and the disconnect between simulation basis and actual working conditions caused by the lack of dynamic data fusion, lack of scientific fusion mechanism, reliance on empirical estimation of fracture surface slip volume and lack of geomechanical constraints in existing technologies. On this basis, a three-dimensional finite element mechanical forward model is constructed with predicted slip volume and geostress field as boundary conditions. The elastic modulus and thickness of the cement sheath are used as design variables. The optimal parameter combination is automatically obtained by constructing and solving the constraint optimization problem. This effectively overcomes the defects of traditional finite element simulation parameter optimization, such as single dimension, reliance on manual trial calculation, low efficiency of repeated calculations, and easy omission of optimal solutions. This achieves precision and efficiency in casing deformation control under complex geological conditions. Attached Figure Description

[0021] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a drilling casing deformation prevention and control method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the overall process of the intelligent control method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the geological structure and well location distribution of the target work area in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the effect of variational mode decomposition (VMD) noise reduction processing in an embodiment of the present invention. Figure 5 This is a diagram showing the formation pressure response characteristics of the fracture zone during drilling, according to an embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of a real-time diagnostic machine learning model based on a TCN-LSTM-Attention hybrid architecture according to an embodiment of the present invention. Figure 7 This is a schematic diagram of the wellbore enlargement operation according to an embodiment of the present invention; Figure 8 This is a diagram showing the seismic-microseismic fault superposition and well logging interpretation results of a casing deformation according to an embodiment of the present invention. Figure 9 This is a schematic diagram of the cement ring material optimization process according to an embodiment of the present invention; Figure 10 This is a schematic diagram of FEM forward calculus according to an embodiment of the present invention; Figure 11 This is a schematic diagram of the cement ring parameters according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0024] like Figure 2 As shown, this invention provides a closed-loop drilling risk prevention and dynamic control system. It first predicts downhole risks based on geological models and machine learning models, and then monitors whether the working conditions exceed the safety threshold through real-time data streams. When an early warning is triggered, the system enters a reverse design and optimization control process. By optimizing the enlargement size of the risk section, the composition and ratio of the cement sheath, and combining AI proxy models and finite element forward modeling calculations, it quickly selects the optimal parameters and finally generates comprehensive control commands. At the same time, it updates the downhole status and model, forming a complete closed loop from risk prediction, real-time monitoring, dynamic optimization to command generation, so as to ensure safe and efficient drilling.

[0025] Based on this, the present invention provides a method for preventing and controlling drilling casing deformation, specifically as follows: Figure 1 As shown, it includes: S1. Collect dynamic time-series data of the target well and static geological data of the area where the target well is located; the static geological data includes the distance of the well from the fault and fracture, the geostress field value and rock mechanics parameters; the dynamic time-series data includes the drilling pressure, torque, pump pressure, rotation speed and flow rate of the target well.

[0026] Logging-while-drilling sequence data, also known as dynamic time-series data, specifically includes time-series parameters such as drill bit pressure, drill string torque, drilling fluid pump outlet pressure, drill string or drill bit rotation speed, and drilling fluid pump output flow rate, denoted as time series. ,in m For dynamic parameter dimensions, t For time steps, yes t At this moment, the first m Dimensional dynamic temporal characteristics.

[0027] Static geological data includes stratigraphy, fault distance, geostress, pore pressure, etc., and is denoted as static feature vector. ,in n For static parameter dimensions, For the first nThe geological static parameters were analyzed. VMD decomposition was performed on the drilling sequence data for noise reduction. VMD is an adaptive signal decomposition technique that decomposes the original signal into multiple intrinsic mode functions (IMFs) with different center frequencies. By calculating the permutation entropy of each mode component or its correlation with the original signal, components representing the effective signal were selected for reconstruction, thereby filtering out environmental noise and equipment interference to obtain clean time-series data with improved signal-to-noise ratio. .

[0028] Collect 3D seismic data, well logging data (such as sonic logging, density, and resistivity), imaging logging data, and core experimental data for the target block. Core experimental data includes: rock elastic modulus (35-45 GPa), Poisson's ratio (0.22-0.28), compressive strength (280-320 MPa), and tensile strength (8-12 MPa) measured by triaxial compression tests on cored well samples, as well as core porosity tests (4%-6%), permeability tests, and other experimental parameters.

[0029] By applying seismic spectral decomposition and coherence volume techniques, and through joint analysis in the time and frequency domains, the spatial distribution of minute fractures and crack zones can be precisely identified, and their attitude (dip, azimuth) and density can be quantified. For example... Figure 3 The distribution of fractures and well locations is shown in the data, and the analysis results clearly define the spatial relationship between the wellbore trajectory and the main fracture system. Based on well logging data such as wellbore collapse and drilling-induced fractures, and combined with the regional tectonic background, numerical inversion methods (such as finite element inversion) are used to obtain a distribution model of the three-dimensional geostress field (maximum / minimum horizontal principal stress and vertical stress) of the block.

[0030] Referring to the basic rock mechanics parameters in Table 1, the identified fracture network, the inverted in-situ stress field, and the lithological and physical property parameters of the formation are integrated to establish a heterogeneous three-dimensional geomechanical grid model. In this embodiment, the three-dimensional geomechanical model is defined as a cylindrical region with a radius of 100m centered on the wellbore, with a grid resolution of 0.5m. This range is sufficient to cover the influence area of ​​the main fracture system around the well, while the 0.5m grid resolution can effectively analyze the mechanical response characteristics of the cement sheath-casing system, achieving a balance between computational accuracy and efficiency. Each grid cell in the model contains detailed attributes (such as in-situ stress, lithology, porosity, Young's modulus, Poisson's ratio, etc.), providing static feature input for subsequent real-time diagnosis. This structure enables the model to understand transient changes on short timescales while maintaining an overall grasp of long-term evolution trends. Through an attention mechanism, it automatically focuses on risk precursor areas, thus exhibiting higher sensitivity and stronger adaptability in identifying complex risks such as fracture activation, fault slip, and wellbore instability.

[0031] Table 1. Basic Parameters of the Model

[0032] This data compiles second-level drilling engineering parameters (drilling pressure, torque, pump pressure, displacement, etc.) and geological static characteristics of historical wells in the block. For example... Figure 4 As shown, VMD (Variational Mode Decomposition) technology is used to reduce noise in engineering signals. By calculating the permutation entropy of each modal component, effective intrinsic mode functions (IMFs) are selected for signal reconstruction, significantly improving the signal-to-noise ratio. Static features (such as distance from the fracture, geostress value, and rock brittleness index) at corresponding locations are extracted from the geological model constructed in the first step along the actual drilling trajectory.

[0033] These static features are aligned and concatenated with the dynamic time-series data preprocessed by VMD according to a sliding time window to form a spatiotemporal fusion feature sample set. This is then constructed as follows: Figure 6 The TCN-LSTM-Attention hybrid neural network model is shown. This model utilizes TCN's causal dilated convolution to extract local temporal features, LSTM to capture long-range dependencies, and the Attention mechanism to dynamically focus on key risk time steps. The model employs multi-task output, simultaneously predicting crack activation probability, fracture slip, construction pressure curve, and crack geometric parameters (dip angle, width, etc.). The model is trained using historical data, and hyperparameters are optimized using a validation set until the model achieves satisfactory prediction accuracy on the test set (e.g., risk identification accuracy > 90%).

[0034] The preset risk parameters include a crack activation probability threshold, a fracture slip warning value, and an abnormal deviation rate of construction pressure. In this embodiment, based on the historical overlay accident statistics and material safety factor of the target block, the crack activation probability threshold is set to 0.7, the fracture slip warning value is 5mm, and the abnormal deviation rate of construction pressure is set to 15%. When the model output exceeds any preset risk parameter threshold, that moment is determined to be a critical moment sequence, triggering the subsequent reverse design process.

[0035] During the actual drilling of new wells, the system receives data streams in real time. The model combines the static background provided by the geological model with real-time engineering parameters for online inference. For example... Figure 6 As shown, visualizing the Attention weights can intuitively demonstrate which key time steps the model focused on before the risk occurred, and how this corresponds to actual engineering parameter anomalies, such as sudden changes in pump pressure. (See Appendix) Figure 5 The sharp rise in the medium pressure response curve corresponds to the moment of fracture activation, and the risk level can be quantified by the degree of deviation from the normal drilling baseline.

[0036] S2. Perform spatiotemporal fusion processing on the dynamic time-series data and the geological static data according to the time window to obtain spatiotemporal fused data; extract the local time-series features of the spatiotemporal fused data, capture the long-range time-series dependencies of the local time-series features, and locate the key time sequence related to the preset risk parameters in the dependencies; based on the key time sequence and the physical constraints provided by the preset three-dimensional geomechanical model, determine the predicted slip of the potential fault surface around the target well.

[0037] S201. Constructing a high-precision three-dimensional geomechanical model The establishment of a high-precision three-dimensional geomechanical model is mainly based on multi-source heterogeneous data such as seismic, well logging, and drilling data. It employs well-seismic fusion and other techniques to sequentially construct four parts: a structural model, an attribute model, a fracture model, and a stress model. First, basic data is collected, including seismic fault information, structural undulations, well location data, stratification information, and well logging data, to construct the structural model of the study area. Seismic data serves as the core basis throughout the process, and is combined with geological stratification data for structural correction, gradually building models of each sub-layer and ultimately forming a complete platform structural geometric model. Based on this geological model, well logging curves are discretized, and variogram fitting techniques are used to analyze the data, thereby applying Kriging interpolation within the platform area to construct an attribute model. Furthermore, based on data from field outcrops, core inspections, and imaging logging, and combined with ant-body tracking technology in seismic bodies, a fracture model of the study area is further established. Finally, against the backdrop of paleotectonic movement and combined with indoor stress test results, a finite element-based inversion algorithm is used to invert the three-dimensional geostress field. Each grid cell in the three-dimensional geomechanical model established through the above process is assigned detailed attributes, including but not limited to lithology, porosity, permeability, total organic carbon content, triaxial stress, and key rock mechanical parameters (elastic modulus, Poisson's ratio, etc.), providing a solid physical foundation for subsequent intelligent diagnosis.

[0038] Based on the wellbore-fracture relationship, a numerical model of stress and deformation coupling between formation, fracture, wellbore, cement sheath, and casing is constructed. Combined with the fracture-activated slip model, and taking into account factors such as fracture friction coefficient, cement sheath material, and casing strength, the casing deformation is calculated, providing reliable input for subsequent reverse design.

[0039] S202. Construct and train a real-time diagnostic model with time-aware capabilities. Based on massive historical drilling and engineering data, an innovative TCN-LSTM-Attention hybrid neural network model was constructed. This model uses multi-source dynamic parameters during drilling and static formation features as joint inputs. After feature fusion, the data sequentially passes through a temporal convolutional network, a sequence dependency modeling network, and a temporal attention mechanism, forming a progressively deeper and more enhanced feature extraction path. This enables the identification and prediction of critical downhole risk events. The model can simultaneously capture rapid local changes, overall trend evolution, and critical anomalies, making it particularly suitable for real-time risk prediction scenarios in well sections with complex geological structures, severe engineering disturbances, and high signal-noise environments.

[0040] The cleaned time-series data is divided into sliding window segments with a window length of [value missing]. Step size is Generate input samples: ; in, For the first A dynamic data sample matrix for each time window; For the denoised dynamic time series data at time... The value of is a collection A vector of dynamic parameters; For the first The start time of each time window; The length of the sliding window; The number of dimensions for dynamic parameters; This is the index number of the window sample; This matrix is ​​represented as OK A column of real numbers.

[0041] Temporal features are fused with static features. For each time window sample... Static features copy Next, splice together: ; in, This is a matrix formed after copying static geological data; This represents the original geological static feature vector. The number of dimensions of the static parameters; Indicates matrix transpose. The length of the sliding window.

[0042] The two matrices are concatenated along their feature dimensions to obtain the final spatiotemporal fusion data sample: ; in, For the first Spatiotemporal fusion data samples from one window; This represents the total number of feature dimensions after concatenation.

[0043] S203, Deep Learning Model Construction The model employs a three-layer TCN-LSTM-Attention structure, which utilizes spatiotemporal fusion data. As input, local temporal features are extracted through a Temporal Convolutional Network (TCN), then long-term dependencies are captured through a Long Short-Term Memory (LSTM) network, and finally, an attention mechanism dynamically focuses on key time sequences to output the prediction result. The overall mapping relationship is as follows: ; in, For all learnable parameters of the model, For predicting output; For spatiotemporal fusion data samples, Refers to the TCN-LSTM-Attention hybrid neural network model.

[0044] TCN is used to extract local temporal features, employing causal convolution and dilated convolution: ; in, For the first Feature maps output by layer TCN; is the layer index of the TCN network; ReLU is the linear rectified activation function; Indicates the expansion factor as causal convolution; and The first Layer convolution kernel weights and biases.

[0045] The expansion factor increases exponentially with the number of layers: This causes the network's receptive field to expand exponentially: ; in, k The kernel size is... This refers to the number of TCN layers; For the first The expansion factor of the layer.

[0046] The core computations of LSTM at each time step are as follows: The input at the current moment comes from the TCN output, the hidden state at the previous moment, and the cell state.

[0047] ; in, The output of the forget gate; Here is the weight matrix for the forget gate; It is the sigmoid activation function; The product of Hadamard; This is the weight matrix of the input gate; The weight matrix of the candidate states; This is the input for the current moment; It is in a hidden state; This is the state of memory cells; The current state of the candidate memory cells; This is the output of the output gate; For time step index; The output of the input gate; This is the weight matrix of the output gate, used to calculate the output gate. ; The bias vector of the forget gate, The bias vector of the input gate, The bias vector of candidate memory cells, This is the bias vector for the output gate; The hyperbolic tangent activation function maps the input to the (-1,1) interval and is used to generate candidate memory units and output hidden states.

[0048] The attention mechanism is used to focus on key moment sequences, employing additive attention, with attention scores as follows: ; Attention weights (normalized by softmax): ; Context vector (weighted sum): ; In the formula, For time step The original attention score; The total length of the time window; For time step Attention weights; , and These are learnable parameters for the attention mechanism; For LSTM at time t The hidden state; c This is the final context representation vector.

[0049] The TCN (Temporal Convolutional Network) is used to extract local short-term features (local temporal features) of the data. The LSTM (Long Short-Term Memory Network) is used to capture long-range temporal dependencies in the drilling data. The Attention mechanism is combined to dynamically calculate the weight of different time steps for the current risk prediction and determine the predicted slip of the potential fracture surface around the target well.

[0050] Unlike previous models that relied solely on a single convolutional structure or recurrent neural network, this hybrid architecture organically integrates local high-frequency pattern recognition, long-term dependency modeling, and temporal key point focusing. While traditional convolutional models can extract local variation features, they struggle to characterize the long-range correlations of drilling parameters over time; recurrent networks alone are prone to "forgetting" temporal information and exhibiting lag in response to sudden engineering disturbances. This invention introduces a multi-channel causal convolutional structure, making the model more sensitive to abrupt changes in pressure, torque, etc., within a short time window; subsequently, a sequence network with a memory structure is used to stably represent gradual trends, cumulative effects, and slowly changing risk characteristics during drilling; and a temporal attention mechanism automatically highlights the importance of specific time steps, enabling the entire sequence to reinforce critical moments with significant formation changes or prominent risk precursors when converging. Compared to traditional models, this invention's structural design achieves higher prediction accuracy and stronger interpretability while maintaining computational efficiency.

[0051] S3. Using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition, a three-dimensional finite element mechanical forward model is constructed for the wellbore structure of the target well. The elastic modulus and thickness of the cement sheath between the wellbore and the casing are used as design variables to obtain the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model for each design variable. With minimizing the casing mechanical response value as the optimization objective and the engineering feasible range of the design variables as constraints, a constrained optimization problem is constructed. The constrained optimization problem is solved to output the optimal cement sheath elastic modulus and thickness. The construction control of the target well casing is then performed using the optimal cement sheath elastic modulus and thickness.

[0052] The effective normal stress and shear stress at fracture are calculated as follows: ; ; in, The effective normal stress acting on the fracture surface; This refers to the shear stress acting on the fracture surface; This represents the maximum horizontal principal stress. The minimum horizontal principal stress; It is the angle between the normal direction of the fracture surface and the direction of the maximum principal stress.

[0053] The critical activation pressure increment is calculated based on the Coulomb criterion: ; in, The critical pressure increment required to activate the fracture surface; The coefficient of friction of the fracture surface; This represents the original formation pressure.

[0054] The crack activation probability is output using the sigmoid activation function: ; in, This represents the probability of crack activation. For activation functions; This is the weight matrix of the probability output layer; The context vector output by the attention mechanism; For the bias term of the probability output layer; It is a natural constant.

[0055] Predicted slip values ​​are output from linear layers: ; in, The predicted slip at the potential fracture surface; The weight matrix of the output layer is the slip amount; For the bias term of the slip output layer, The context vector output by the attention mechanism.

[0056] In addition to predicting crack activation probability and slip volume, the model achieves forward-looking predictions of construction stress through an additional regression branch. This prediction is based on a fusion of current and historical data. Output the pressure evolution curve for several future time steps: ; in, The predicted construction pressure curve (including key pressure parameters such as pump pressure and drilling pressure). and Weights and biases for the stress prediction branch, c The context vector output by the Attention mechanism is used to compare the predicted pressure with the normal drilling baseline pressure. It can identify abnormal signs in advance and provide early warning of risks.

[0057] Based on the predicted pressure curve The model extracts key features such as the amplitude and duration of pressure anomalies, and combines them with pressure gradients. Abnormal torque and drilling azimuth A quantitative description of the crack is output through a specialized geometric parameter regression layer: ; in, The angle of inclination of the crack; The width of the crack; The length of the crack; For geometric parameter regression function; This is an abnormal amplitude of pressure. For pressure gradient; This is an abnormal torque value. To account for the differences in pressure response at different azimuth angles. P Pressure refers specifically to pressure parameters monitored in real time, such as pump pressure or drilling pressure, in drilling operations. D p The formula represents the well depth. This formula achieves an inverse mapping from engineering response (pressure, torque) to geological characteristics (fracture geometry), enabling the system to perceive fracture status in real time during drilling. This provides quantitative geological risk input for subsequent cement sheath parameter optimization (elastic modulus, thickness), thereby improving the accuracy of casing deformation control.

[0058] The geometric parameter inversion module, together with the leakage probability prediction and slip prediction, constitutes the multi-task output architecture of the model. The predicted crack geometric parameters will be directly input into the cement ring reverse design module, forming a technical flow from risk diagnosis to parameter optimization.

[0059] Reverse design and parameter optimization of cement rings, based on predicted geological risk parameters (especially fracture slip). (and geostress field), constructing an intelligent system for FEM forward modeling, AI proxy modeling, and parameter inverse identification and optimization, aiming to customize optimal cement sheath material parameters (elastic modulus) for specific geological conditions. ) and geometric dimensions (thickness) This reduces the risk of casing deformation.

[0060] Establish a three-dimensional finite element model of the coupled system of formation-fracture-wellbore-cement sheath-casing. For example... Figure 10 As shown, the model includes a cement ring model, a casing model, and a model cross-section, clearly demonstrating the spatial relationships of the various structures. The predicted fracture slip is then calculated. The displacement boundary condition is applied to the fracture surface, and the stress load distribution on the fracture surface is calculated using the geostress tensor. The wellbore-cement sheath-casing system adopts a concentric cylindrical shell model, and the elastic modulus of the cement sheath is... Poisson's ratio ,thickness These are the parameters to be optimized, i.e., the design variables.

[0061] Training samples were generated in the cement sheath parameter space using Latin hypercube sampling (LHS), and FEM simulation was run on each sample to calculate the maximum mechanical response of the casing. A training dataset was constructed. A deep neural network (CNN) was used to fit the input-output relationship to build a model that enables a rapid mapping from cement ring parameters to casing stress.

[0062] As an alternative implementation to the aforementioned AI proxy model, Support Vector Regression (SVR) or Gaussian Process Regression (GPR) can also be used to construct the proxy model. The elastic modulus, thickness, and Poisson's ratio of the cement ring are used as input feature vectors, and the equivalent stress of the casing is used as the output target. A kernel function (such as the radial basis function kernel, RBF) maps the input samples to a high-dimensional feature space, where an optimal regression hyperplane is constructed. The convex quadratic programming problem is solved with the goal of minimizing the ε-insensitive loss function, resulting in a nonlinear regression mapping relationship from input to output. SVR exhibits good generalization ability under small sample conditions, and after training, it can quickly predict the mechanical response value of the casing under a given combination of cement ring parameters. Alternatively, the relationship between the cement ring design parameters and the casing mechanical response can be considered as a Gaussian process prior. An appropriate covariance function (kernel function) is selected to describe the similarity measure between input samples. Based on the training sample set, the posterior prediction distribution is calculated, and the mean and confidence interval of the predicted values ​​are output. GPR can not only provide point estimates of the casing mechanical response, but also provide a quantitative assessment of the prediction uncertainty, providing a risk reference for subsequent optimization decisions.

[0063] The SVR and GPR surrogate models mentioned above have the same input-output mapping relationship as the DNN surrogate model. They can all replace DNN to achieve a fast mapping from cement ring design parameters to casing mechanical response. After training, they can be combined with global optimization algorithms such as genetic algorithm (GA) to solve the constrained optimization problem and output the optimal combination of cement ring parameters.

[0064] The cement ring design is transformed into a constrained optimization problem, with the goal of minimizing the casing mechanical response value while satisfying engineering constraints. : ; ; in, The elastic modulus of the cement ring; Let be the Poisson's ratio of the cement ring; The thickness of the cement ring; This represents the mechanical response value of the casing. For the trained AI agent model. This represents the lower feasible limit for the elastic modulus of cement rings in engineering. The upper limit of the engineering feasible elastic modulus of the cement ring; This represents the lower feasible limit for the thickness of the cement ring. The upper limit of the engineering feasibility for the thickness of the cement ring; This represents the lower limit of the feasible engineering limit for the ring Poisson's ratio of cement. The upper limit of the feasible engineering value for the ring Poisson's ratio of cement; The yield strength of the casing material; For safety factors, the elastic modulus in this example is specifically in the range of 0.5-10 GPa; the cement ring thickness is specifically in the range of 38-50 mm; and the Poisson's ratio is specifically in the range of 0.2-0.35.

[0065] The Genetic Algorithm (GA) is used to perform global optimization on the constrained optimization problem and output the optimal combination of cement ring parameters. The optimization principle is as follows: reducing the elastic modulus enhances deformation coordination, increasing the thickness disperses stress transmission, and the two work together to minimize the stress in the casing.

[0066] The safety threshold is 80% of the yield strength of the casing material. If the predicted maximum equivalent stress of the casing still exceeds the safety threshold after parameter optimization, a geometric compensation mechanism is activated. The wellbore diameter is increased by enlargement technology to enhance the cement sheath thickness. The required thickness is calculated back-calculated based on the target casing stress to determine the enlarged wellbore diameter. Enlargement operations employ direct enlargement, staged enlargement, or selective enlargement strategies depending on geological conditions. By establishing a mechanism of parameter optimization, effect verification, and geometric compensation, a cement sheath design scheme that meets safety requirements can be obtained under various geological conditions.

[0067] When the risks predicted by the diagnostic model (such as slip volume and fracture activation probability) exceed the preset safety threshold, the system automatically triggers the reverse design process. Based on the predicted fracture slip volume and other geological risk parameters, a three-dimensional finite element model of the coupled system of "formation-fracture-cement sheath-casing" is established. The predicted slip volume is applied as a displacement boundary condition to the fracture surface to simulate its impact on the wellbore system under specific geostresses as shown in Table 1. Table 2 shows the assigned parameters for the model input.

[0068] Table 2. Fracture mechanics and seepage properties assigned at different scales.

[0069] Feasible cement ring parameters (elastic modulus) Poisson's ratio ,thickness Within the space, a large number of samples were generated using Latin hypercube sampling (LHS), and the maximum equivalent stress of the casing corresponding to each sample was calculated using the aforementioned FEM model. These data were then used to train a deep neural network (DNN) as an AI proxy model to achieve rapid mapping from cement ring design parameters to casing stress. Subsequently, with the goal of minimizing casing stress and constrained by material properties and construction conditions, a global optimization algorithm, such as a genetic algorithm (GA), was used for inverse optimization to obtain the optimal combination of cement ring parameters. The effectiveness of the optimization scheme can be visually demonstrated through numerical simulation.

[0070] like Figure 9 As shown, this invention achieves inverse optimization of cement sheath parameters by combining a genetic algorithm with a finite element forward model. First, the predicted geological risk parameters are input into a three-dimensional finite element forward model, and multiple sets of forward calculations are performed within the cement sheath parameter space to generate a dataset of casing mechanical responses under different parameter combinations. Then, each set of data is used as input, and a genetic algorithm is employed for global optimization. Based on the casing mechanical response values ​​corresponding to each parameter combination, the optimal cement sheath parameter combination (x, y, z) that minimizes the casing mechanical response is ultimately output. This optimization process, through the global search capability of the genetic algorithm, can efficiently screen the optimal combination within the engineering feasible domain of the cement sheath's elastic modulus and thickness, avoiding the shortcomings of traditional methods that rely on manual trial calculations and repeated calculations. This provides a reliable design basis for casing deformation control under complex geological conditions.

[0071] like Figure 11 As shown, the damage distribution diagrams under different mix proportions (DAMAGE values ​​range from 0 to 0.9343, with the gray level increasing from light to dark corresponding to the increasing DAMAGE value) visually reflect the distribution of the elastic zone (light color) and plastic zone (dark color) of the cement ring, providing a basis for parameter optimization. Multiple sets of deformation simulation results corresponding to different mix proportions were obtained (denoted as R1, R2, R3...R...). n Each set of results corresponds to specific spatial coordinate data (such as (X1,Y1,Z1), (X2,Y2,Z2)...(X...). n ,Y n Z n Using the above simulation results as input data, a genetic algorithm is used for global optimization. The algorithm filters back based on the casing mechanical response values ​​(equivalent stress, strain, and displacement, etc.) corresponding to each set of results, and finally determines the optimal combination of cement ring parameters (denoted as (x,y,z)). This optimal parameter combination can minimize the casing mechanical response value and effectively control casing deformation under complex geological conditions.

[0072] If material ratio optimization alone cannot reduce casing stress below the safety threshold, a geometric compensation mechanism is activated. Studies show that increasing the cement sheath thickness can significantly improve casing's resistance to compression. The wellbore diameter is increased through reaming technology, thereby increasing the cement sheath annular thickness. The system calculates the required thickness based on the target stress, determines the reaming diameter, and recommends specific reaming techniques (such as direct reaming or staged reaming) based on geological conditions. Cement slurry dosage adjustment involves first determining the cement sheath outer diameter based on the calculated cement sheath compensation thickness and the wellbore diameter after reaming; then calculating the annular cross-sectional area based on the casing outer diameter, multiplying it by the sealing section length to obtain the theoretical annular volume; finally, based on the wellbore structure complexity and drilling fluid displacement efficiency, a displacement loss coefficient of 1.15 to 1.25 is introduced to calculate the actual cement slurry dosage. The cement slurry formula is adjusted simultaneously to generate a wellbore reaming construction plan including dosage, ratio, and rheological parameters.

[0073] To achieve efficient eye expansion, such as Figure 7 As shown, the YK series PDC reamer is used for wellbore enlargement. The specific drill string combination includes a 114mm pilot pile, a YK114-140PDC reamer, near-bit geological steering, and non-magnetic pressure-bearing drill pipe, etc., to ensure efficient wellbore enlargement in the build-up and horizontal sections, creating space for increasing the thickness of the cement sheath.

[0074] After well completion, all data from the entire well process (including raw data, model predictions, reverse engineering schemes, and actual construction results) is automatically stored in the system's knowledge base. The preset scale is 100 cases accumulated in the knowledge base, or a quarterly update cycle. Meeting either condition triggers retraining of the AI ​​agent model to update its parameters. By continuously feeding in new field data, the model can adapt to changes in geological conditions and iterative construction techniques, achieving continuous performance improvement. This prediction accuracy can be verified through the overlay and comparison of seismic and microseismic monitoring data, such as... Figure 8 As shown, the predicted fault zone is spatially highly consistent with the microseismic events actually recorded during drilling, and the casing deformation logging interpretation results are located precisely within the high-density microseismic activity zone. This fully demonstrates the accuracy of the method described in this invention in predicting potential risk areas around the well, and provides reliable feedback data for the continuous optimization of the model.

[0075] Actual performance data includes: the cementing index (BI) for cement sheath cementing quality evaluation, the ellipticity change rate interpreted from casing deformation logging, and the leakage rate for wellbore integrity detection, among other measured parameters. The case data is stored in a structured format, including: well number, geological static data, dynamic time-series data, predicted slippage, optimal cement sheath parameters (elastic modulus, thickness), actual performance data, and construction date.

[0076] Based on the same inventive concept, the present invention also provides a drilling casing deformation prevention and control system, comprising: The data acquisition module is used to acquire dynamic time-series data of the target well and static geological data of the area where the target well is located. The static geological data includes the distance of the well from the fault and fracture, the geostress field value, and rock mechanics parameters. The dynamic time-series data includes the drilling pressure, torque, pump pressure, rotational speed, and flow rate of the target well.

[0077] The prediction module is used to perform spatiotemporal fusion processing on the dynamic time-series data and geological static data according to time windows to obtain spatiotemporal fused data; extract local time-series features of the spatiotemporal fused data, capture the long-range time-series dependencies of the local time-series features, and locate the key time sequence related to preset risk parameters in the dependencies, wherein the preset risk parameters include the fracture activation probability threshold of the target well, the fracture slip volume warning value, and the construction pressure abnormal deviation rate; based on the key time sequence and the physical constraints provided by the preset three-dimensional geomechanical model, determine the predicted slip volume of the potential fracture surface around the target well.

[0078] The application module is used to construct a three-dimensional finite element mechanical forward model for the wellbore structure of the target well, using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition; using the elastic modulus and thickness of the cement sheath between the wellbore and the casing as design variables, it obtains the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model under each design variable; with minimizing the casing mechanical response value as the optimization objective and the engineering feasible range of the design variables as constraints, it constructs a constrained optimization problem, solves the constrained optimization problem, and outputs the optimal cement sheath elastic modulus and thickness; wherein, the engineering feasible range includes the lower and upper limits of the cement sheath elastic modulus and the cement sheath thickness; the lower and upper limits are predetermined based on the cement slurry material performance test data and wellbore geometry of the target well; and the construction control of the target well casing is performed using the optimal cement sheath elastic modulus and thickness.

[0079] This invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the drilling casing deformation prevention and control method provided above.

[0080] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the drilling casing deformation prevention and control method provided above.

[0081] Specific limitations regarding the calculation system for drilling casing deformation control methods can be found in the limitations of drilling casing deformation control methods described above, and will not be repeated here. Each module in the aforementioned drilling casing deformation control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0082] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Furthermore, the above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for preventing and controlling drilling casing deformation, characterized in that, include: Collect dynamic time-series data of the target well and static geological data of the area where the target well is located; The geological static data includes the distance between the well and the fault / fracture, the geostress field value, and rock mechanics parameters; the dynamic time-series data includes the target well's drilling pressure, torque, pump pressure, rotational speed, and flow rate. The dynamic time-series data and geological static data are spatiotemporally fused according to time windows to obtain spatiotemporally fused data. Local time-series features of the spatiotemporally fused data are extracted, and the long-range time-series dependencies of the local time-series features are captured. Key time sequences related to preset risk parameters are located in the dependencies. The preset risk parameters include the fracture activation probability threshold, fracture slip warning value, and construction pressure abnormal deviation rate of the target well. Based on the key time sequences and the physical constraints provided by the preset three-dimensional geomechanical model, the predicted slip of the potential fracture surface around the target well is determined. Using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition, a three-dimensional finite element mechanical forward model is constructed for the wellbore structure of the target well. The elastic modulus and thickness of the cement sheath between the wellbore and the casing are used as design variables to obtain the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model for each design variable. Minimizing the casing mechanical response value is the optimization objective, and the engineering feasible range of the design variables is used as a constraint to construct a constrained optimization problem. Solving the constrained optimization problem outputs the optimal elastic modulus and thickness of the cement sheath. The engineering feasible range includes the lower and upper limits of the cement sheath elastic modulus and the cement sheath thickness. The lower and upper limits are predetermined based on the test data of the cement slurry material properties and the wellbore geometry of the target well. The construction control of the target drilling casing is carried out by utilizing the optimal elastic modulus and thickness of the cement sheath.

2. The drilling casing deformation control method according to claim 1, characterized in that, Before performing spatiotemporal fusion processing of the dynamic time-series data and geological static data according to the time window, the method further includes: using the variational mode decomposition (VMD) method to decompose the dynamic time-series data for noise reduction, selecting effective modal components for reconstruction, and obtaining the denoised dynamic time-series data; the spatiotemporal fusion processing specifically involves: copying the geological static data along the time dimension and splicing it with the dynamic time-series data within the same time window in terms of feature dimensions to form the spatiotemporal fused data.

3. The drilling casing deformation control method according to claim 1, characterized in that, Local temporal features of the spatiotemporal fusion data are extracted using a temporal convolutional network (TCN). Long short-term memory (LSTM) networks are used to capture the long-range temporal dependencies of these local temporal features. An attention mechanism is used to locate key time sequences related to preset risk parameters within these dependencies. These key time sequences are then substituted into the physical constraints of a three-dimensional geomechanical model, and the TCN-LSTM-Attention hybrid neural network model outputs the predicted slip of the potential fracture surface around the target well.

4. The drilling casing deformation control method according to claim 1, characterized in that, The preset three-dimensional geomechanical model is constructed based on seismic interpretation data, well logging data and core experimental data of the target drilling area. The model range is a cylindrical area with a radius of 50-200m centered on the wellbore, and the grid division accuracy is 0.1m-1m.

5. The drilling casing deformation control method according to claim 1, characterized in that, Obtaining the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model under various design variables specifically includes: sampling within the engineering feasible space of the design variables using the Latin hypercube sampling method to generate multiple sample combinations; running the three-dimensional finite element mechanical forward model for each sample combination to calculate the corresponding casing equivalent stress, strain mechanical response values, and displacement mechanical response values, forming a training dataset; training an artificial intelligence proxy model using the training dataset, and obtaining the predicted casing mechanical response value under any combination of design variables through the proxy model; the artificial intelligence proxy model is a deep neural network CNN, support vector regression SVR, or Gaussian process regression GPR model.

6. The drilling casing deformation control method according to claim 5, characterized in that, After the construction control of the target drilling casing is carried out, the method further includes: collecting the actual effect data after the construction of the target drilling cement sheath, associating it with the corresponding predicted slip and optimal cement sheath parameters and storing it in the knowledge base; when the case data accumulated in the knowledge base reaches a preset scale, the retraining of the artificial intelligence agent model is triggered to update the model parameters.

7. The drilling casing deformation control method according to claim 1, characterized in that, The method of using the optimal elastic modulus and thickness of the cement sheath to control the construction of the target drilling casing specifically includes: if the predicted mechanical response value of the casing under the optimal elastic modulus and thickness of the cement sheath is still higher than the safety threshold, then a geometric compensation mechanism is activated. The geometric compensation mechanism includes back-calculating the required cement sheath compensation thickness based on the target response value, simultaneously adjusting the amount of cement slurry, and generating a wellbore enlargement construction plan.

8. A drilling casing deformation prevention and control system, characterized in that, include: The data acquisition module is used to collect dynamic time-series data of the target well and static geological data of the area where the target well is located; The geological static data includes the distance between the well and the fault / fracture, the geostress field value, and rock mechanics parameters; the dynamic time-series data includes the target well's drilling pressure, torque, pump pressure, rotational speed, and flow rate. The prediction module is used to perform spatiotemporal fusion processing on the dynamic time-series data and geological static data according to time windows to obtain spatiotemporal fused data; extract local time-series features of the spatiotemporal fused data, capture the long-range time-series dependencies of the local time-series features, and locate the key time sequence related to preset risk parameters in the dependencies, wherein the preset risk parameters include the fracture activation probability threshold of the target well, the fracture slip volume warning value, and the construction pressure abnormal deviation rate; based on the key time sequence and the physical constraints provided by the preset three-dimensional geomechanical model, determine the predicted slip volume of the potential fracture surface around the target well; The application module is used to construct a three-dimensional finite element mechanical forward model for the wellbore structure of the target well, using the predicted slip as the displacement boundary condition and the geostress field as the load boundary condition; using the elastic modulus and thickness of the cement sheath between the wellbore and the casing as design variables, it obtains the casing mechanical response values ​​of the three-dimensional finite element mechanical forward model under each design variable; with minimizing the casing mechanical response value as the optimization objective and the engineering feasible range of the design variables as constraints, it constructs a constrained optimization problem, solves the constrained optimization problem, and outputs the optimal cement sheath elastic modulus and thickness; wherein, the engineering feasible range includes the lower and upper limits of the cement sheath elastic modulus and the cement sheath thickness; the lower and upper limits are predetermined based on the cement slurry material performance test data and wellbore geometry of the target well; and the construction control of the target well casing is performed using the optimal cement sheath elastic modulus and thickness.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 7.

Images