Three-dimensional deformation real-time monitoring method and system based on Beidou and inclination angle data fusion

Abstract

The invention belongs to the technical field of oil and gas storage and transportation safety, structure health monitoring and intelligent sensing fusion, and provides a three-dimensional deformation real-time monitoring method and system based on Beidou and inclination angle data fusion, and the method comprises the steps: collecting coordinate information and attitude angle information of a monitoring point in a three-dimensional space; performing error correction on the coordinate information to obtain a high-precision three-dimensional coordinate of the monitoring point, and performing component processing on the attitude angle information to obtain an overall inclination trend angle of the monitoring point; performing multi-source fusion on the high-precision three-dimensional coordinates and the overall inclination trend angle to construct a real-time three-dimensional deformation monitoring model; according to the real-time three-dimensional deformation monitoring model, displacement and spatial attitude change of the monitoring point are determined; wherein the displacement and space attitude change comprises the lifting or settling volume, the horizontal displacement and the space inclination angle of the monitoring point. Three-dimensional monitoring, abnormal trend recognition and advanced early warning of the gas storage well head device can be achieved under the complex injection and production working conditions, and the injection and production well risk management and control capacity is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of oil and gas storage safety, structural health monitoring and intelligent sensing fusion, and particularly relates to a three-dimensional deformation real-time monitoring method and system based on Beidou and inclination data fusion. BACKGROUND

[0002] As an important infrastructure for natural gas seasonal peak shaving and strategic reserve, the gas storage is prone to axial displacement and relative inclination at the wellhead position due to the differences in the thermal expansion coefficients of the internal components of the wellbore, as well as the combined effects of cementing failure of the casing and cement sheath, and other factors. In particular, under the conditions of strong injection and strong production, large swallowing and large discharge, it is more likely to induce small but continuous axial displacement and spatial inclination. These deformations will gradually be transmitted to the surface Christmas tree, flange connection and control equipment during long-period accumulation, thereby affecting the sealing performance, load bearing capacity and overall stability of the system.

[0003] In current engineering practice, the monitoring of wellhead uplift and deformation mainly uses traditional measuring equipment such as leveling instruments or total stations, mainly relying on manual periodic reading of elevation or displacement data at the measuring points on site. Although this method is simple to deploy at the initial stage, it is heavily dependent on personnel operation and is difficult to achieve continuous observation and rapid response. Especially in the running state of the gas storage with frequent injection and production switching and rapid evolution of deformation, there are problems such as monitoring lag, data discontinuity and error accumulation. At the same time, in the face of such monitoring objects as the wellhead device of the gas storage, which is disturbed by multiple sources, has small deformation amplitude and has a complex structure, the existing scheme is still difficult to meet the comprehensive requirements of "three-dimensional high-precision real-time monitoring + abnormal prediction + intelligent linkage response".

[0004] Therefore, establishing a monitoring system that can realize real-time, accurate and comprehensive perception of three-dimensional deformation of the wellhead is a key technical requirement for ensuring the safe operation of the gas storage. SUMMARY

[0005] To solve the above problems, the application provides a three-dimensional deformation real-time monitoring method based on Beidou and inclination data fusion, which comprises the following steps: Collecting coordinate information and attitude angle information of the monitoring point in three-dimensional space; Performing error correction on the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point, and performing component processing on the attitude angle information to obtain the overall inclination trend angle of the monitoring point; Performing multi-source fusion on the high-precision three-dimensional coordinates and the overall inclination trend angle to construct a real-time three-dimensional deformation monitoring model; Determining the displacement and spatial attitude change of the monitoring point according to the real-time three-dimensional deformation monitoring model; wherein the displacement and spatial attitude change includes the uplift or subsidence amount, horizontal displacement and spatial inclination angle of the monitoring point.

[0006] Furthermore, the method also includes: Construct a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms; The spatiotemporal prediction model is used to dynamically predict the cumulative displacement and spatial attitude change trends of each monitoring point.

[0007] Furthermore, the method also includes: A deformation warning threshold is set. If the predicted value of any measuring point exceeds the deformation warning threshold or the rate of change reaches a critical condition, a warning signal is triggered and an alarm message is generated.

[0008] Furthermore, the high-precision three-dimensional coordinates of the monitoring point are obtained by error correction of the coordinate information, including: The pseudorange data transmitted by BeiDou satellites is corrected by ionospheric error, tropospheric error, satellite clock error and receiver clock error; The corrected pseudorange data is linearized based on Taylor series expansion; The high-precision three-dimensional coordinates of the monitoring points are determined by least squares adjustment of the linearized pseudorange data.

[0009] Furthermore, the overall tilt trend angle of the monitoring point is obtained by performing component processing on the attitude angle information, including: Determine the acceleration components of several tilt sensors in the three-axis coordinate directions; The tilt angle of each monitoring point is determined based on the acceleration components. The overall tilt trend angle of the monitoring points is determined by the tilt angle of each monitoring point and the arctangent function.

[0010] Furthermore, constructing a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms includes: Spatial dependencies between monitoring points are extracted using graph convolutional networks; Spatial dependencies are combined with gated recurrent networks to model time series data, and an attention mechanism is introduced to construct a spatiotemporal prediction model.

[0011] Furthermore, the dynamic prediction of the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model includes: The cumulative displacement is decomposed into trend term displacement and periodic term displacement; The spatiotemporal prediction model is integrated with the double exponential smoothing algorithm to predict the displacement of the trend term, and the displacement prediction model is used to predict the displacement of the periodic term. The correlation trend term displacement prediction and the periodic term displacement prediction can dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point.

[0012] This invention also provides a real-time three-dimensional deformation monitoring system based on the fusion of BeiDou and tilt data. The system includes: an acquisition unit, a calculation unit, a data fusion unit, and a monitoring unit. The acquisition unit is used to acquire the coordinate information and attitude angle information of the monitoring point in three-dimensional space; The calculation unit is used to perform error correction on the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point; it is also used to perform component processing on the attitude angle information to obtain the overall tilt trend angle of the monitoring point. The data fusion unit is used to fuse the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources to construct a real-time three-dimensional deformation monitoring model. The monitoring unit is used to determine the displacement and spatial attitude changes of the monitoring point based on the real-time three-dimensional deformation monitoring model; wherein, the displacement and spatial attitude changes include the lifting or settling amount, horizontal displacement and spatial tilt angle of the monitoring point.

[0013] Furthermore, the system also includes a trend prediction unit. The trend prediction unit is used to build a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms. It is also used to dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model.

[0014] Furthermore, the system also includes an early warning unit. The early warning unit is used to set a deformation early warning threshold. If the predicted value of any measuring point exceeds the deformation early warning threshold or the rate of change reaches a critical condition, an early warning signal is triggered and alarm information is generated.

[0015] Furthermore, the system also includes a sensor deployment unit, wherein the sensors include a BeiDou receiver and an tilt sensor, and the sensor deployment unit is used to deploy no less than three BeiDou receivers and tilt sensors around the monitoring point.

[0016] The present invention also provides an electronic device, the electronic device comprising at least one processor and at least one memory, wherein the processor and the memory are data connected, wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of the present invention.

[0017] The present invention also provides a computer storage medium storing one or more instructions that, when executed by one or more computers, cause the one or more computers to perform the method described in the present invention.

[0018] The present invention relates to a three-dimensional deformation real-time monitoring method and system based on the fusion of BeiDou and tilt angle data. This system enables three-dimensional monitoring, anomaly trend identification, and early warning of gas storage wellhead equipment under complex injection and production conditions, thereby improving the risk management capabilities of injection and production wells. It fully integrates the high-precision advantages of BeiDou static positioning with the spatial attitude perception capabilities of tilt angle sensors, forming an integrated three-dimensional dynamic perception system for wellheads through multi-source data fusion and intelligent calculation algorithms. At the same time, the present invention introduces a deep learning graph neural network framework, which has dynamic trend modeling and anomaly early warning functions, and can adapt to the complex operating conditions of gas storage facilities, achieving a fundamental leap from passive monitoring to active prediction. It has significant engineering promotion value and safety assurance significance.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A schematic diagram of the real-time monitoring method for three-dimensional deformation based on the fusion of BeiDou and tilt data is shown in an embodiment of the present invention. Figure 2 A schematic diagram of the three-dimensional deformation real-time monitoring system based on the fusion of BeiDou and tilt data is shown in an embodiment of the present invention. Figure 3 A schematic diagram of the specific process of the three-dimensional deformation real-time monitoring method based on the fusion of Beidou and tilt data in an embodiment of the present invention is shown. Figure 4(a) shows the early warning judgment and visualization diagram in the schematic diagram of the monitoring point displacement trend in the embodiment of the present invention; Figure 4(b) shows the horizontal trajectory diagram in the schematic diagram of the monitoring point displacement trend in the embodiment of the present invention; Figure 4(c) shows the settlement trajectory diagram in the schematic diagram of the monitoring point displacement trend in the embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] The purpose of this invention is to provide a real-time monitoring and early warning method for three-dimensional deformation of gas storage wellhead devices based on the fusion of Beidou and tilt angle data, so as to solve the problem that the existing technology cannot achieve real-time, high-precision, and all-round monitoring of three-dimensional deformation of gas storage wellhead devices caused by complex working conditions such as injection-production alternation, large intake and output, and strong injection and production. In particular, the existing technology is lacking in real-time performance, spatial perception dimension, data fusion and intelligent early warning capabilities.

[0024] By deploying multiple BeiDou receivers and tilt sensors, three-dimensional displacement and spatial attitude data of the wellhead device are collected. Pseudorange calculation is performed using BeiDou static differential technology to obtain millimeter-level high-precision coordinate change values, and tilt angle change trends are identified through tilt angle calculation. Furthermore, a spatiotemporal fusion model is constructed based on graph convolutional units, gated recurrent units, and attention mechanisms to dynamically predict the cumulative displacement and tilt trend of the wellhead. Multi-level safety thresholds are set by combining historical monitoring data, simulation models, and actual operating conditions to achieve real-time judgment and intelligent early warning of abnormal wellhead uplift, subsidence, or tilt trends. This invention can be widely applied to three-dimensional deformation sensing and proactive risk control of gas storage wellheads under complex service conditions, and has advantages such as high deployment flexibility, high monitoring accuracy, strong data fusion capability, and fast early warning response.

[0025] Figure 1 This diagram illustrates a flowchart of a real-time three-dimensional deformation monitoring method based on the fusion of BeiDou and tilt data, as described in an embodiment of the present invention. Figure 1 The method includes: acquiring coordinate information and attitude angle information of a monitoring point in three-dimensional space; correcting errors in the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point, and performing component processing on the attitude angle information to obtain the overall tilt trend angle of the monitoring point; fusing the high-precision three-dimensional coordinates and the overall tilt trend angle through multi-source fusion to construct a real-time three-dimensional deformation monitoring model; and determining the displacement and spatial attitude changes of the monitoring point based on the real-time three-dimensional deformation monitoring model; wherein the displacement and spatial attitude changes include the heave or subsidence of the monitoring point, horizontal displacement, and spatial tilt angle.

[0026] Specifically, in this embodiment of the invention, the method further includes: Construct a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms; The spatiotemporal prediction model is used to dynamically predict the cumulative displacement and spatial attitude change trends of each monitoring point.

[0027] Specifically, in this embodiment of the invention, the method further includes: A deformation warning threshold is set. If the predicted value of any measuring point exceeds the deformation warning threshold or the rate of change reaches a critical condition, a warning signal is triggered and an alarm message is generated.

[0028] The following provides a detailed description of the specific steps of the three-dimensional deformation real-time monitoring method based on the fusion of BeiDou and tilt data in the embodiments of the present invention: Step S1: Collect the coordinate information and attitude angle information of the monitoring point in three-dimensional space. In this embodiment of the invention, coordinate information in three-dimensional space is collected by a Beidou receiver, and attitude angle information is collected by an inclination sensor. Beidou receivers and inclination sensors are evenly and densely deployed at each wellhead that needs to be monitored. In order to ensure that a rigid measurement network can be formed and the accurate three-dimensional coordinates of the wellhead device can be obtained, the number of Beidou receivers is not less than three. At the same time, in order to cover each monitoring area and ensure that the tilt of each part of the wellhead can be monitored in real time, at least three inclination sensors are evenly deployed around the wellhead that needs to be monitored.

[0029] In this embodiment of the invention, the arrangement of the Beidou receiver and tilt sensor around the wellhead supports simultaneous deployment at multiple points on-site. Both the Beidou receiver and the tilt sensor employ protective structures, possessing corrosion resistance, explosion-proof capabilities, and stable operation under high humidity to adapt to the complex working conditions of the gas storage environment. It should be noted that this embodiment of the invention does not limit the specific structure of the protective structure, as long as it achieves the operational functions of this invention.

[0030] Step S2: Correct the coordinate information to obtain the high-precision three-dimensional coordinates of the monitoring point. Specifically, in this embodiment of the invention, obtaining high-precision three-dimensional coordinates of the monitoring point by error correction of the coordinate information includes: The pseudorange data transmitted by BeiDou satellites is corrected by ionospheric error, tropospheric error, satellite clock error and receiver clock error; The corrected pseudorange data is linearized based on Taylor series expansion; The high-precision three-dimensional coordinates of the monitoring points are determined by least squares adjustment of the linearized pseudorange data.

[0031] The following provides a detailed explanation of the specific calculation process for high-precision three-dimensional coordinates: In this embodiment of the invention, the BeiDou receiver transmits the raw data collected in real time to the correction unit for processing via wired or wireless means. Based on the BeiDou satellite navigation system, the static pseudorange difference method is used to correct the error of the pseudorange data of multiple points of the wellhead device and perform least squares adjustment to obtain the high-precision three-dimensional coordinates of each monitoring point. Specifically, based on the pseudo-random code broadcast by the satellite and the receiver's copy code, the pseudorange data is obtained by calculating the product of the speed of light and the signal propagation time. On this basis, considering the ionospheric error, tropospheric delay, and clock errors of the receiver and the satellite, the pseudorange data is corrected for errors. Then, the nonlinear ranging formula is linearized using Taylor series expansion, and finally, the least squares method is used for adjustment to obtain the high-precision three-dimensional coordinates of each station.

[0032] Step S21: Correct the pseudorange data transmitted by BeiDou satellites using ionospheric error, tropospheric error, satellite clock error, and receiver clock error. The observed pseudorange is expressed as: (1) In equation (1), The pseudorange is represented by r; the geometric distance between the satellite and the receiver is represented by c; Indicates receiver clock bias; Indicates satellite clock bias; Indicates tropospheric error; Indicates ionospheric error. This represents the observation noise / unmodeled error term.

[0033] The geometric distance r between the satellite and the receiver is expressed as: (2) Substituting equation (2) into the pseudorange calculation formula in equation (1), we get: (3) Figure 3 middle, This represents the corrected pseudorange. This represents the residual after linearization / unmodeled error.

[0034] Step S22: Taylor series linearization and design matrix construction Taking a first-order Taylor expansion of the geometric distance function at approximate coordinates (X0, Y0, Z0), we get: (4) (5) Where D represents the geometric distance between the satellite and the receiver calculated at approximate coordinates (X0, Y0, Z0), and V x V y Vz These represent the components (direction cosines) of the unit line-of-sight vector pointing from the receiver to the satellite on the three axes. This is equivalent to the first-order partial derivative of the geometric distance with respect to the coordinates. , , X, Y, and Z represent the three-dimensional coordinates of the satellite in the Earth-Centered, Earth-Fixed coordinate system (ECEF, Earth-Fixed), respectively; X, Y, and Z represent the actual three-dimensional coordinates of the monitoring point (receiver) to be determined; c is the speed of light. This is the receiver clock bias (equivalent distance term).

[0035] When combining multiple star observations: (6) in, ,for The observation vector, where n represents the number of available satellites participating in the solution of this epoch.

[0036] Step S23: Least Squares Adjustment and Iterative Convergence Based on the corrected pseudorange of S21 and the linearization results of S22 (see equations (4)-(6)), the three-dimensional coordinates of the monitoring point and the receiver clock error of each epoch are solved by weighted least squares, and the convergent solution is obtained through the closed loop iteration of "linearization-solution-update-relinearization"; at the same time, the quality label is output to provide weighted input for subsequent BeiDou-tilt data fusion and spatiotemporal prediction.

[0037] Initialization and Modeling: Using the previous epoch result as an approximation, calculate the geometric distance of each satellite according to equation (5) and form the direction cosine and design matrix A. Then, organize the observation vector l=ρ according to equation (6). i D i The weighting array W is adaptively set based on observation quality factors such as satellite elevation angle, signal-to-noise ratio, and multipath indication; at the same time, geometric and observability thresholds (such as effective satellite count, PDOP, and minimum elevation angle) are applied to ensure numerical stability.

[0038] Adjustment and parameter update: At the current approximation, the parameter increment is solved by weighted least squares in one step.

[0039] (7) After obtaining the increment, update the coordinates and clock difference, and calculate the residual vector v for this round: (8) Reconstruct D by substituting the updated parameters back into equations (5)-(6). i A and l are then used to proceed to the next round of linearization and solution.

[0040] Convergence Criteria and Real-Time Control: Iteration stops when the following two thresholds are met: first, the parameter increment reaches millimeter-level accuracy; second, the number of iterations does not exceed a set upper limit (ensuring real-time output of 1 Hz or higher). In engineering implementation, this invention adopts a unified and directly auditable stopping condition: (9) Where ε pos The position convergence threshold is set to 10. 4 On the order of m; ε clk The clock bias convergence threshold (distance equivalence) is set to a distance equivalence threshold of the same order of magnitude. If the geometry of the constellation in this epoch temporarily degenerates, a short-window weak constraint smoothing (exponential forgetting) is triggered to maintain continuous output, and the process automatically exits when the conditions are restored.

[0041] Step S3: Perform component processing on the attitude angle information to obtain the overall tilt trend angle of the monitoring point. Specifically, in this embodiment of the invention, the component processing of the attitude angle information to obtain the overall tilt trend angle of the monitoring point includes: Determine the acceleration components of several tilt sensors in the three-axis coordinate directions; The tilt angle of each monitoring point is determined based on the acceleration components. The overall tilt trend angle of the monitoring points is determined by the tilt angle of each monitoring point and the arctangent function.

[0042] The following provides a detailed explanation of the calculation process for the overall tilt trend angle of the monitoring points: In this embodiment of the invention, the tilt sensor transmits the raw data collected in real time to the angle determination unit for processing via wired or wireless means. The angle determination unit processes the acceleration components in the x-axis, y-axis, and z-axis directions output by the tilt sensor, calculates the tilt angle values ​​of each monitoring point of the wellhead device, and derives the overall tilt trend of the wellhead. The tilt sensor calculates the tilt angle of each point by measuring the acceleration components in the x-axis, y-axis, and z-axis directions using the principle of gravity projection, and further derives the overall tilt information of the wellhead device.

[0043] Step S31: Determine the acceleration components of several tilt sensors in the three-axis coordinate directions. In this embodiment of the invention, three tilt sensors are used as an example. Each tilt sensor can measure the acceleration components of its location in the x-axis, y-axis, and z-axis directions, which are denoted as follows: In this embodiment of the invention, the acceleration component is generated by the projection of gravity onto the inclined plane where the tilt sensor is located; Step S32: Determine the tilt angle of each monitoring point based on the acceleration components, and determine the overall tilt trend angle of the monitoring points based on the tilt angles of each monitoring point and the arctangent function. For the first tilt sensor, its tilt angle The tangent value is expressed as: (10) g represents the acceleration due to gravity, and similarly, we can obtain... , The overall tilt trend angle of the monitoring points is determined by combining the arctangent function. The overall tilt trend angle of the monitoring points is expressed as: (11) Step S4: Perform multi-source fusion of the high-precision three-dimensional coordinates and the overall tilt trend angle to construct a real-time three-dimensional deformation monitoring model; determine the displacement and spatial attitude changes of the monitoring points based on the real-time three-dimensional deformation monitoring model. In this embodiment of the invention, the multi-source fusion process employs synchronous normalization processing of displacement time series and tilt sequence to construct a joint data matrix for collaborative solution.

[0044] Specifically, the three-dimensional coordinate data obtained from the BeiDou receiver is dynamically calculated in real time to generate a displacement change map of the wellhead device. The inclinometer data is converted into tilt angles, and the tilt value at a single point is obtained using the arctangent function. Then, the overall tilt trend of the wellhead is calculated. During multi-source fusion, a cumulative displacement prediction model is established to achieve spatiotemporal correlation between various GNSS (Global Navigation Satellite System) monitoring points. A multi-source data fusion algorithm is used to uniformly process data from different sensors, eliminating the influence of individual errors and constructing a high-precision, real-time dynamic three-dimensional deformation monitoring model for the wellhead. The specific process is as follows: 1) Data acquisition and time alignment: The Beidou receiver and the three-axis tilt sensor sample at 1Hz and synchronously write a unified timestamp; dual constraints of local clock and network clock are used to ensure that the alignment error of multi-source data is ≤0.2 s. Basic quality control (filling in missing measurements, removing obvious jumps / glitches) is performed on the raw data and a unified format is completed.

[0045] 2) Real-time dynamic calculation of GNSS (Global Navigation Satellite System): Following the process in S21–S23, the corrected pseudorange is linearized and subjected to weighted least squares iteration to obtain the high-precision coordinates (X) of the epoch (t). t Y t Z t) and quality indicators (significant stars, PDOP, coordinate standard deviation, etc.). Then, the ECEF coordinates are converted to ENU coordinates under the reference base station (E t N t U t (Site coordinate system / East-North-Sky coordinate system, local Cartesian coordinates coordinate system), and the displacement components are obtained by subtracting the coordinates from the reference epoch (or reference point): (12) 3) Inclination calculation and calibration: The inclinometer outputs triaxial accelerations (ax, ay, az), which are first low-pass filtered (e.g., 2-5Hz cutoff) and calibrated for zero bias / temperature drift, and then the single-point attitude angle is calculated according to the gravity projection model. (13) The attitude information from multiple sensors at the wellhead is combined to form the overall tilt trend angle. It also outputs the tilt angle time series.

[0046] 4) Multi-source fusion and noise reduction: On a unified time axis, consider ΔEt, ΔNt, ΔUt, and θ. x θ y Θ t Perform synchronization and normalization to construct joint feature vectors. (14) in and All values ​​are mass weights (obtained by mapping from PDOP (Position Dilution of Precision), signal-to-noise ratio, residuals / normalized residuals, etc.), and are fused in real time using weighted Kalman / complementary filtering: GNSS dominates the absolute displacement, and tilt angle dominates the attitude and short-term dynamics; the weights are adaptively adjusted according to the mass labels, outputting the denoised 3D displacement and overall tilt angle, and updating the 3D deformation monitoring model in real time.

[0047] 5) Cumulative displacement prediction and its spatiotemporal correlation: The cumulative displacement is decomposed into a "trend term + periodic term". Double exponential smoothing (DES) is used to enhance the robustness of the trend term. For the periodic term and multi-point coupling, a spatiotemporal prediction model using GCN (Graph Neural Network) + GRU (Gated Recurrent Unit) + attention is employed: a graph structure is constructed based on the spatial adjacency relationships (distance / pipeline topology) of the monitoring points, and a sliding time window is used as the input. Perform online updates / inferences, and output the next T hours (e.g., 24 hours). Prediction curve. The model receives quality labels from both steps 2) and 4) as prior weights for prediction confidence.

[0048] 6) Early warning judgment and visualization: The system provides graded early warnings based on thresholds for heave / settlement, horizontal displacement, tilt angle, and rate of change. When these thresholds are exceeded or the rate reaches a critical value, local audio-visual alerts and remote push notifications are triggered, and a linked visualization of "displacement-tilt angle-confidence" (displacement vector diagram, horizontal / settlement trajectory, and tilt angle trend diagram) is generated at the front end.

[0049] 7) Engineering parameters and outputs: The entire link operates at 1Hz; the S23 iteration convergence threshold is set to millimeter level; short-window smoothing is enabled for short-time constellation degradation, and "degraded operation" is marked. The final output includes: denoised 3D displacement (…). (etc.), overall tilt angle Quality labels and health scores, as well as prediction results for the next T hours, serve as unified interface data for the multi-source fusion model and early warning module.

[0050] In this embodiment of the invention, the three-dimensional deformation monitoring model can not only calculate the vertical lift or settlement of the wellhead device, but also accurately obtain its horizontal displacement and spatial tilt angle, and predict its displacement, thus satisfying the requirement for accurate description of the deformation of the entire wellhead area.

[0051] Step S5: Construct a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms. In this embodiment of the invention, a spatiotemporal prediction model including graph convolution units (GC), gated recurrent units, and attention mechanisms is used to dynamically predict the cumulative displacement and spatial attitude change trends of each monitoring point of the wellhead device.

[0052] Specifically, the spatiotemporal prediction model extracts the spatial dependencies between monitoring points through a graph convolutional network and combines it with a GRU network to model the time series. At the same time, an attention mechanism is introduced to improve the ability to identify trend changes, aiming to capture the complex spatiotemporal relationships between GNSS monitoring points at the wellhead. It combines a graph convolutional unit (GC) and a gated recurrent unit (GRU) to jointly construct a cumulative displacement prediction unit to realize the spatiotemporal association of each GNSS monitoring point. The graph convolutional unit (GC) is a key component of the model and is responsible for processing spatial data. Through graph convolution operations, the model can capture the spatial relationships between monitoring points. In practical applications, each monitoring point can be regarded as a node in a graph, and the spatial adjacency between nodes is represented by the edges of the graph. The GC unit effectively extracts the spatial features between monitoring points by performing convolution operations on the graph structure. Its convolution calculation is shown in the following formulas (15)-(18). Three GC units are used to perform convolution operations on the graph structure at time t to extract the spatial features between monitoring points. The iterative calculation method for convolution is as follows: (15) (16) (17) (18) In equations (8)-(11), GC represents a graph convolution unit, and the node feature matrix at time t... Adjacency Matrix

[0053] (19) Where A is the adjacency matrix (N×N) of the graph, determined by the spatial proximity or topological relationships of the monitoring points. Let I be the adjacency matrix after adding self-loops, and let I denote the identity matrix (N×N). for The degree matrix, For trainable weights, the output dimension is R represents the real number field (elements of the matrix / vector are real numbers), used to add self-loops in the adjacency matrix, N represents the number of nodes in the graph (number of monitoring points), and F represents the input feature dimension (number of input features for each node, such as displacement, temperature, etc.). d represents the number of output channels of the graph convolution (feature dimension of the GC unit output). h This represents the hidden state dimension of GRU (the number of hidden units in GRU for each node). , , Both indicate resetting the door parameters. This represents the hidden state at time t-1. ; , This indicates resetting the door and updating the door. ; Indicates the candidate hidden state. This represents the final hidden state at time t. ; Indicates the candidate hidden state. ; Represents the Hadamard product; , , All represent candidate state parameters. , , All of these represent updating gate parameters. Resetting gate parameters, candidate state parameters, and updating gate parameters are all parameters that the model needs to train. This represents the displacement and attribute information of N nodes corresponding to the wellhead at time t. It is an adjacency matrix of N nodes.

[0054] In this embodiment of the invention, based on extracting the spatial dependencies between monitoring points through a graph convolutional network and modeling the time series using a GRU network, an attention mechanism is added. Different time points correspond to different hidden states H. A self-attention mechanism is used to calculate the attention weights corresponding to each hidden state and construct the output vector H', which is expressed as: (20) In equation (20), Q is the query matrix, representing "the retrieval demand for information from each historical moment at the current moment," obtained by linear mapping from H. K is the key matrix, representing the "matchability of features at each historical moment as the target to be retrieved", L represents the time series length (number of historical moments involved in attention calculation / window length), and d v This represents the feature dimension of vector V. V is a value matrix, representing "the weighted aggregated historical features". , , , , For the parameters that need to be trained, It is the feature dimension of vector K, and the attention weights are normalized using the softmax function.

[0055] The hidden state H' of the cumulative displacement prediction unit, which has been processed by the self-attention mechanism, is fully connected through a fully connected layer, and its output is used as the displacement at the next time step. The formula is W ft and b flThese are the weight matrix and bias vector that need to be trained, respectively.

[0056] Step S6: Dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model. In this embodiment of the invention, a cumulative displacement prediction model is proposed based on graph convolutional units, gated recurrent units, and attention mechanisms, and integrated with the double exponential smoothing (DES) algorithm to perform correlation prediction on the displacement of multiple monitoring points on the surface of the sliding body.

[0057] Specifically, the cumulative displacement is decomposed into trend displacement and periodic displacement. The DES algorithm is used to predict the trend displacement, and then the Rayleigh displacement prediction model is used to perform correlation prediction on the periodic displacement of multiple monitoring points at the wellhead, finally obtaining the displacement prediction result at the wellhead.

[0058] Step S7: Set a deformation warning threshold. If the predicted value at any measuring point exceeds the deformation warning threshold or the rate of change reaches a critical condition, a warning signal is triggered, and alarm information is generated. In this embodiment of the invention, a deformation early warning threshold system is set up to compare and analyze real-time monitoring data and prediction results. When any monitoring indicator exceeds the preset threshold or its rate of change reaches the critical condition, an intelligent early warning signal is triggered and alarm information is generated.

[0059] Specifically, the warning thresholds include the maximum allowable displacement of wellhead rise / settlement, the limit of horizontal displacement, the limit of tilt angle, and the critical value of change rate. The warning can be set according to a three-level response.

[0060] After data fusion, real-time data is compared based on preset safety thresholds and dynamic trends. The warning thresholds are set by integrating historical monitoring data, laboratory simulation results, and actual on-site conditions, covering indicators such as maximum allowable displacement, tilt angle, and rate of change. The system employs an intelligent warning model based on time series analysis, fuzzy logic, or artificial intelligence algorithms to predict trends and determine anomalies in the monitoring data. Once the displacement or tilt data of the wellhead device exceeds the safe range, or the rate of change increases sharply, the system will automatically trigger a warning signal. Warning information is transmitted to the monitoring center in real time via a wireless communication module, and simultaneously displays alarms on the on-site screen using visual and audible methods, recording abnormal data and alarm times to provide a basis for subsequent emergency response. The warning signal can also be linked with the automatic control system to guide on-site personnel to take measures such as adjusting injection and production parameters, strengthening wellhead fixation, or initiating on-site maintenance to ensure the overall safe operation of the gas storage facility.

[0061] The three-dimensional deformation real-time monitoring method based on the fusion of Beidou and tilt data described in this embodiment of the invention is applicable to gas storage wellhead monitoring scenarios with periodic injection and production operations, complex wellbore structures, and small deformations but with cumulative effects. It is especially suitable for three-dimensional spatial deformation identification and risk prevention and control of wellhead equipment during long-term operation.

[0062] This invention also provides a three-dimensional deformation real-time monitoring system based on the fusion of BeiDou and tilt data. Figure 2 This diagram illustrates the structure of a three-dimensional deformation real-time monitoring system based on the fusion of BeiDou and tilt data in an embodiment of the present invention. Figure 2 The system includes: an acquisition unit, a calculation unit, a data fusion unit, and a monitoring unit. The acquisition unit is used to acquire the coordinate information and attitude angle information of the monitoring point in three-dimensional space; The calculation unit is used to correct errors in the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point; it is also used to process the attitude angle information to obtain the overall tilt trend angle of the monitoring point; specifically, the calculation unit sends the data of each sensor to the data processing center in real time through a wired or wireless network; the Beidou receiver uses error correction and least squares adjustment to solve the three-dimensional coordinates based on the correlation calculation of pseudo-random code and receiver copy code; the tilt sensor calculates the tilt angle through acceleration components and gravity projection.

[0063] The data fusion unit is used to fuse the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources to construct a real-time three-dimensional deformation monitoring model. Specifically, the data fusion unit synchronizes and normalizes the BeiDou and tilt angle data to construct a unified data matrix, thereby obtaining the uplift, horizontal displacement, and overall tilt angle of the wellhead device. Dynamic calculation and error reduction are achieved through the fusion algorithm.

[0064] The monitoring unit is used to determine the displacement and spatial attitude changes of the monitoring point based on the real-time three-dimensional deformation monitoring model; wherein, the displacement and spatial attitude changes include the lifting or settling amount, horizontal displacement and spatial tilt angle of the monitoring point.

[0065] Specifically, the system also includes a trend prediction unit, used to construct a spatiotemporal prediction model incorporating a graph convolutional network, a gated recurrent network, and an attention mechanism; it is also used to dynamically predict the cumulative displacement and spatial attitude change trends of each monitoring point based on the spatiotemporal prediction model. Specifically, the system utilizes a graph convolutional network to capture the spatial correlation between monitoring points, combines a GRU for time series modeling, and incorporates an attention mechanism to enhance the identification of key change features; it then uses a double exponential smoothing DES and periodic term decomposition method to predict the displacement trend and periodic term, outputting the future deformation trend results.

[0066] Specifically, the system also includes an early warning unit for setting a deformation early warning threshold. If the predicted value at any measuring point exceeds the deformation early warning threshold or the rate of change reaches a critical condition, an early warning signal is triggered, and alarm information is generated. In this embodiment of the invention, the early warning unit is linked with the injection and extraction control system via a wireless communication module, supporting remote early warning issuance and operation and maintenance guidance. It is used to determine anomalies based on set displacement and tilt thresholds and automatically trigger alarm signals and linkage response measures. Specifically, real-time and predicted data are compared using preset safety thresholds. When the monitored data exceeds the limit or the rate of change accelerates abnormally, the system automatically triggers audible and visual alarms and remote alarms, records relevant data, generates a visual report, and guides on-site emergency intervention measures.

[0067] Specifically, the system also includes a sensor deployment unit. The sensors include BeiDou receivers and tilt sensors. The sensor deployment unit deploys at least three BeiDou receivers and tilt sensors around the monitoring points to construct a rigid three-dimensional coordinate measurement network and a spatial attitude monitoring network. It should be noted that the sensor deployment unit deploys at least three BeiDou receivers at each wellhead to obtain high-precision three-dimensional coordinates; simultaneously, it deploys at least three tilt sensors to sense the x / y axis tilt angles in real time, and the protective structure adapts to the high temperature, high humidity, and strong interference environment of the gas storage facility.

[0068] Figure 3 This diagram illustrates the specific process of a real-time three-dimensional deformation monitoring method based on the fusion of BeiDou and tilt data in an embodiment of the present invention. Figure 3In this process, sensors are deployed at the wellhead, including BeiDou receivers and three-axis tilt sensors. At least three BeiDou receivers are deployed at the wellhead to form a BeiDou receiver array, and three-axis tilt sensors are deployed to cover key areas of the wellhead to form a tilt sensor network. It should be noted that protective structures are installed on both the BeiDou receivers and the three-axis tilt sensors. The acquisition unit collects BeiDou data streams by acquiring pseudorange observations, satellite clock error compensation, and atmospheric delay correction; it also simultaneously acquires tilt sensor data streams by acquiring XYZ three-axis acceleration data and calculating using gravity projection. The settlement unit performs correlation calculations based on pseudorandom codes and receiver copy codes on the collected BeiDou data, employing pseudorange linearization processing, error correction, and least squares adjustment to solve for the three-dimensional coordinates. The overall tilt is determined by calculating the tilt angle of a single point; the data fusion unit is used to fuse the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources to construct a real-time three-dimensional deformation monitoring model; the trend prediction unit is used to construct a spatiotemporal prediction model including a graph convolutional network, a gated recurrent network, and an attention mechanism; it is also used to dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model; the early warning unit sets dynamic thresholds, which are determined by the historical working condition database and the boundary values ​​of finite element simulation, and sets multi-level early warning rules: displacement mutation rate alarm, tilt angle over-limit alarm, and composite deformation risk index alarm; and it is linked and controlled through an audible and visual alarm device, a monitoring center push, and an injection and sampling parameter adjustment interface.

[0069] The present invention also provides a detailed description of a method for real-time monitoring and early warning of three-dimensional deformation of gas storage wellhead equipment based on the fusion of BeiDou and tilt data through a specific embodiment: Step 101: Deploy the sensor network Taking a certain well in a gas storage facility as an example, three Beidou receivers and three tilt sensors are installed at the four corners of the wellhead. The three form a rigid triangular network with a height of 0.8m above the ground and a sampling frequency of 1Hz. Step 102: Acquire and Align Multi-Source Data By using differential correction from GNSS reference stations, the BeiDou-calculated coordinates at each moment are obtained. At the same time, the triaxial acceleration data of the tilt sensor is collected, and the data is aligned before fusion using a timestamp synchronization algorithm, with the accuracy error controlled within 0.2 seconds. Step 103: Fusion calculation of BeiDou and tilt data Kalman filtering and least squares method were used to smooth and fit the BeiDou coordinate data to obtain high-precision three-dimensional coordinates for each observation point. Using the gravity vector projection calculation method, the triaxial acceleration is converted into the tilt angle of the wellhead device, forming a set of real-time three-dimensional deformation parameters of the entire structure; A real-time three-dimensional deformation monitoring model is constructed by fusing the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources. Step 104: Monitoring the displacement and spatial attitude changes of the monitoring points The displacement and spatial attitude changes of the monitoring points are determined based on the real-time three-dimensional deformation monitoring model, wherein the displacement and spatial attitude changes include the lifting or settling of the monitoring points, horizontal displacement and spatial tilt angle. Step 105: Construct a deformation trend prediction model Based on measured data, a graph neural network model containing spatial topology is constructed. By integrating time series change information, spatial correlation features between sensor nodes are extracted through GCN. The GRU network is then input to model the time dynamic sequence. With the attention mechanism, the weight allocation is optimized to predict the maximum displacement and tilt angle change trend of the wellhead device in the next 24 hours. Step 106: 3D Deformation Early Warning Judgment and Visualization When the predicted value of any measuring point exceeds the set threshold, a deformation warning is triggered, and a deformation diagnosis report is generated and automatically pushed through the web and mobile terminals. Once the calculation result exceeds the warning line, a warning message is automatically pushed, as shown in the figure below. Figure 4(a) shows the warning judgment and visualization diagram in the monitoring point displacement trend display diagram of the embodiment of the present invention. In Figure 4(a), the horizontal axis represents time, the vertical axis represents the relative difference / mm, and the calculation result includes northward displacement, eastward displacement, and vertical displacement. The warning line range is within Figure 4(b) shows the horizontal trajectory diagram in the schematic diagram of the displacement trend of the monitoring point in the embodiment of the present invention. In Figure 4(b), the horizontal axis represents the east-west displacement, and the vertical axis represents the north-south displacement. Here, the horizontal trajectory of the monitoring point is: eastward displacement. The northward displacement is 1.6 mm; Figure 4(c) shows the settlement trajectory diagram in the schematic diagram of the displacement trend of the monitoring point in the embodiment of the present invention. In Figure 4(c), the horizontal axis represents the east-west displacement, and the vertical axis represents the vertical displacement. Here, the settlement trajectory of the monitoring point is the eastward displacement. Vertical displacement .

[0070] This invention proposes a real-time monitoring and early warning method for three-dimensional deformation of gas storage wellhead devices based on the fusion of BeiDou and tilt angle data. This method enables three-dimensional monitoring, anomaly trend identification, and early warning of gas storage wellhead devices under complex injection and production conditions, improving the risk management capabilities of injection and production wells. It fully integrates the high-precision advantages of BeiDou static positioning with the spatial attitude perception capabilities of tilt angle sensors, forming an integrated three-dimensional dynamic perception system for the wellhead through multi-source data fusion and intelligent calculation algorithms. Furthermore, this invention introduces a deep learning graph neural network framework, possessing dynamic trend modeling and anomaly early warning functions, enabling it to adapt to complex gas storage conditions and achieving a fundamental leap from passive monitoring to proactive prediction. This method has significant engineering application value and safety assurance significance.

[0071] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time monitoring of three-dimensional deformation based on the fusion of BeiDou and tilt data, characterized in that, The method includes: Collect the coordinates and attitude angles of the monitoring points in three-dimensional space; Error correction is performed on the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point, and component processing is performed on the attitude angle information to obtain the overall tilt trend angle of the monitoring point; A real-time three-dimensional deformation monitoring model is constructed by fusing the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources. The displacement and spatial attitude changes of the monitoring points are determined based on the real-time three-dimensional deformation monitoring model; wherein, the displacement and spatial attitude changes include the lifting or settling of the monitoring points, horizontal displacement and spatial tilt angle.

2. The three-dimensional deformation real-time monitoring method based on BeiDou and tilt data fusion as described in claim 1, characterized in that, The method further includes: Construct a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms; The spatiotemporal prediction model is used to dynamically predict the cumulative displacement and spatial attitude change trends of each monitoring point.

3. The three-dimensional deformation real-time monitoring method based on BeiDou and tilt data fusion according to claim 2, characterized in that, The method further includes: A deformation warning threshold is set. If the predicted value of any measuring point exceeds the deformation warning threshold or the rate of change reaches a critical condition, a warning signal is triggered and an alarm message is generated.

4. The three-dimensional deformation real-time monitoring method based on BeiDou and tilt data fusion according to claim 1 or 2, characterized in that, Obtaining high-precision three-dimensional coordinates of the monitoring point by performing error correction on the coordinate information includes: The pseudorange data transmitted by BeiDou satellites is corrected by ionospheric error, tropospheric error, satellite clock error and receiver clock error; The corrected pseudorange data is linearized based on Taylor series expansion; The high-precision three-dimensional coordinates of the monitoring points are determined by least squares adjustment of the linearized pseudorange data.

5. The method for real-time monitoring of three-dimensional deformation based on the fusion of BeiDou and tilt data according to claim 1 or 2, characterized in that, The overall tilt trend angle of the monitoring point is obtained by performing component processing on the attitude angle information, including: Determine the acceleration components of several tilt sensors in the three-axis coordinate directions; The tilt angle of each monitoring point is determined based on the acceleration components. The overall tilt trend angle of the monitoring points is determined by the tilt angle of each monitoring point and the arctangent function.

6. The three-dimensional deformation real-time monitoring method based on BeiDou and tilt data fusion according to claim 2, characterized in that, Constructing a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms includes: Spatial dependencies between monitoring points are extracted using graph convolutional networks; Spatial dependencies are combined with gated recurrent networks to model time series data, and an attention mechanism is introduced to construct a spatiotemporal prediction model.

7. The three-dimensional deformation real-time monitoring method based on BeiDou and tilt data fusion according to claim 6, characterized in that, The dynamic prediction of the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model includes: The cumulative displacement is decomposed into trend term displacement and periodic term displacement; The spatiotemporal prediction model is integrated with the double exponential smoothing algorithm to predict the displacement of the trend term, and the displacement prediction model is used to predict the displacement of the periodic term. The correlation trend term displacement prediction and the periodic term displacement prediction can dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point.

8. A three-dimensional deformation real-time monitoring system based on the fusion of BeiDou and tilt data, characterized in that, The system includes: an acquisition unit, a calculation unit, a data fusion unit, and a monitoring unit. The acquisition unit is used to acquire the coordinate information and attitude angle information of the monitoring point in three-dimensional space; The calculation unit is used to perform error correction on the coordinate information to obtain high-precision three-dimensional coordinates of the monitoring point; it is also used to perform component processing on the attitude angle information to obtain the overall tilt trend angle of the monitoring point. The data fusion unit is used to fuse the high-precision three-dimensional coordinates and the overall tilt trend angle from multiple sources to construct a real-time three-dimensional deformation monitoring model. The monitoring unit is used to determine the displacement and spatial attitude changes of the monitoring point based on the real-time three-dimensional deformation monitoring model; wherein, the displacement and spatial attitude changes include the lifting or settling amount, horizontal displacement and spatial tilt angle of the monitoring point.

9. The three-dimensional deformation real-time monitoring system based on BeiDou and tilt data fusion as described in claim 8, characterized in that, The system also includes a trend prediction unit. The trend prediction unit is used to build a spatiotemporal prediction model that includes graph convolutional networks, gated recurrent networks, and attention mechanisms. It is also used to dynamically predict the cumulative displacement and spatial attitude change trend of each monitoring point based on the spatiotemporal prediction model.

10. The three-dimensional deformation real-time monitoring system based on BeiDou and tilt data fusion according to claim 9, characterized in that, The system also includes an early warning unit. The early warning unit is used to set a deformation early warning threshold. If the predicted value of any measuring point exceeds the deformation early warning threshold or the rate of change reaches a critical condition, an early warning signal is triggered and alarm information is generated.

11. The three-dimensional deformation real-time monitoring system based on BeiDou and tilt data fusion as described in claim 8 or 9, characterized in that, The system also includes a sensor deployment unit, which includes a BeiDou receiver and an tilt sensor. The sensor deployment unit is used to deploy no less than three BeiDou receivers and tilt sensors around the monitoring point.

12. An electronic device, characterized in that, The electronic device includes at least one processor and at least one memory, wherein the processor and the memory are data connected. The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

13. A computer storage medium, characterized in that, The computer storage medium stores one or more instructions, which, when executed by one or more computers, cause the one or more computers to perform the method of any one of claims 1-7.

Images