Intelligent well site comprehensive management system based on digital twinning and AI analysis

Abstract

The invention discloses an intelligent well site comprehensive management system based on digital twinning and AI analysis, and belongs to the technical field of digital twinning. Uniformly representing the multi-physical field data as a tensor data structure including a space dimension, a time dimension and a physical quantity dimension; performing data fusion on the tensor data structure and the three-dimensional geometric model of the well site, and establishing a spatial association relationship of multi-physical field data at grid nodes; establishing physical field control equations based on the fused data, and correlating the equations through a coupling operator to form a multi-physical field coupling equation; solving the coupling equation to obtain a numerical solution of each physical field distribution; generating a control instruction sequence for adjusting the operation parameters of the well site equipment according to the numerical solution; according to the invention, accurate sensing and intelligent regulation and control of the operation state of the well site equipment are realized, and the intelligent level of well site management is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, specifically to a smart well site integrated management system based on digital twin and AI analysis. Background Technology

[0002] With the rapid development of Industrial Internet technology, digital twins, as a key technology for achieving deep integration of the physical world and cyberspace, are increasingly widely used in fields such as industrial manufacturing and energy extraction. Digital twins, by constructing virtual mappings of physical entities, enable real-time monitoring and dynamic simulation of equipment operating status, providing a new technological approach for the intelligent management of industrial systems. In the oil and gas extraction sector, well sites, as crucial production locations, require precise perception and intelligent control of equipment operating status to ensure safe production and improve operational efficiency.

[0003] Currently, existing technical solutions mainly focus on the visualization and basic data monitoring of digital twins. Existing technologies typically employ 3D modeling to construct well site equipment models and connect to sensor monitoring data via data interfaces to visualize equipment operating parameters within a 3D scene. Some solutions further introduce multi-source data integration methods, binding single physical field data such as temperature and pressure to the 3D model to achieve a static visualization of equipment operating status. In addition, some improved solutions attempt to provide early warnings for abnormal equipment conditions by setting threshold rules, or to establish statistical analysis models of equipment performance using historical data.

[0004] However, existing technologies for processing multiphysics data often remain at the level of independent analysis and simple superposition, lacking in-depth modeling of the interactions between multiple physical fields such as temperature, pressure, and vibration fields, making it difficult to accurately describe the overall operating state of the equipment. Existing control strategy generation methods largely rely on empirical rules or single-physics field analysis, lacking a closed-loop optimization mechanism based on multiphysics co-simulation results, leading to insufficient accuracy and adaptability of control commands. This restricts the reliability and effectiveness of intelligent well site management systems under complex operating conditions. Summary of the Invention

[0005] The purpose of this invention is to provide a smart well site integrated management system based on digital twins and AI analysis, solving the following technical problems: Existing intelligent well site management systems have limitations such as isolated analysis and simple superposition, lacking in-depth modeling of the interaction between multiple physics fields, and failing to accurately characterize the overall operating state of the equipment. The control strategy generation method relies on empirical rules or single physics field analysis, and fails to establish a closed-loop optimization mechanism based on the results of multi-physics field co-simulation, resulting in insufficient accuracy and adaptability of control commands.

[0006] The objective of this invention can be achieved through the following technical solutions: A smart well site integrated management system based on digital twin and AI analysis includes: The multiphysics data acquisition module is used to collect multiphysics data during the operation of well site equipment. The multiphysics data includes temperature field data, pressure field data, and vibration field data. The tensor data structure construction module is used to represent multiphysics data in a unified tensor data structure, which includes spatial dimension, time dimension and physical quantity dimension. The multiphysics data fusion module is used to fuse tensor data structures with the well site 3D geometric model and establish spatial relationships of multiphysics data at the grid nodes of the 3D geometric model. The multiphysics coupling modeling module is used to establish control equations based on the fused multiphysics data, and to associate the control equations of each physics field with each other through coupling operators to form multiphysics coupling equations. The coupled equation solving module is used to solve the multiphysics coupled equations and obtain numerical solutions for the temperature field distribution, pressure field distribution, and vibration field distribution. The control command sequence generation module is used to generate control command sequences based on numerical solutions. These control command sequences are used to adjust the operating parameters of well site equipment.

[0007] As a further aspect of the present invention: in the tensor data structure construction module, the specific process of uniformly representing multi-physics data as a tensor data structure is as follows: A dynamic spatiotemporal grid is constructed, which includes spatial and temporal dimensions; temperature field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series; pressure field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series; vibration field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series. Multiphysics data associations are established at each node of the dynamic spatiotemporal grid. Temperature field data, pressure field data, and vibration field data are integrated into a unified multiphysics tensor through these associations. The multiphysics data in the dynamic spatiotemporal grid is updated according to the time dimension.

[0008] As a further aspect of the present invention: in the multiphysics data fusion module, the specific process of fusing the tensor data structure with the well site three-dimensional geometric model is as follows: A multi-level mesh structure is established, which includes a surface mesh layer, a volume mesh layer, and a structural mesh layer. Temperature field data is mapped to the surface mesh nodes of the multi-level mesh structure, pressure field data is mapped to the volume mesh nodes of the multi-level mesh structure, and vibration field data is mapped to the structural mesh nodes of the multi-level mesh structure. Calculate the spatial correspondence between surface mesh nodes and volume mesh nodes, calculate the spatial correspondence between volume mesh nodes and structural mesh nodes, establish data transfer paths between surface mesh layers, volume mesh layers and structural mesh layers, and realize the unified representation of multiphysics data in multi-level mesh structures through data transfer paths.

[0009] As a further aspect of the present invention: the specific process of establishing data transmission paths between multi-level grid structures is as follows: A projection relationship is established between the surface mesh layer and the volume mesh layer. The projection relationship transfers the data of the surface mesh nodes to the adjacent volume mesh nodes. An interpolation relationship is established between the volume mesh layer and the structural mesh layer. The interpolation relationship transfers the data of the volume mesh nodes to the corresponding structural mesh nodes. Calculate the distance weights from surface mesh nodes to volume mesh nodes. The distance weights determine the intensity of data transmission. Calculate the shape function from volume mesh nodes to structure mesh nodes. The shape function determines the distribution of data transmission. By combining the distance weights and shape function, establish the first data transmission matrix from the surface mesh layer to the volume mesh layer. By combining shape function and distance weight, a second data transfer matrix is ​​established from the volume mesh layer to the structure mesh layer. Multiplying the two data transfer matrices yields the complete data transfer path from the surface mesh layer to the structure mesh layer.

[0010] As a further aspect of the present invention: in the multiphysics coupling modeling module, the specific process of establishing control equations based on the fused multiphysics data and intercorrelating the control equations of each physics field through coupling operators is as follows: Based on multi-physics data in a multi-level grid structure, a multi-physics state vector is constructed, which includes temperature field components, pressure field components and vibration field components. An evolution equation for the multiphysics state vector is established, which describes the change of the multiphysics state vector with time. A coupled system matrix is ​​introduced into the evolution equation, which represents the interaction between the temperature field component, pressure field component, and vibration field component. The eigenvalue problem of the coupled system matrix is ​​solved to obtain the set of coupled modes. Using the set of coupled modes as basis functions, the evolution equations of the multiphysics state vector are decomposed into modes. Through mode decomposition, a set of mutually coupled mode equations are obtained, which constitute the multiphysics coupling equations.

[0011] As a further aspect of the present invention: the specific process for solving the eigenvalue problem of the coupled system matrix is ​​as follows: Construct a mathematical expression for the coupled system matrix, which includes self-acting terms for the temperature field, pressure field, vibration field, and the cross-acting terms among the three. The characteristic polynomial of the coupled system matrix is ​​calculated. The characteristic polynomial is a high-order algebraic equation about the characteristic parameters. The roots of the characteristic polynomial are solved to obtain the eigenvalue spectrum of the coupled system matrix. The eigenvalues ​​are substituted into the homogeneous equations of the coupled system matrix to solve for the corresponding eigenvectors. The eigenvectors are orthogonalized to obtain the standard orthogonal eigenvector set. The standard orthogonal eigenvector set is sorted according to the eigenvalues ​​to form the coupled mode set.

[0012] As a further aspect of the present invention: the specific process of solving the multiphysics coupling equation in the coupling equation solving module is as follows: The multiphysics coupling equations are converted into a system of algebraic equations. An initial solution vector is set for the system of algebraic equations, which includes initial values ​​for the temperature field, pressure field, and vibration field. An iterative solution process is used to process the system of algebraic equations, and the update amount of the solution vector is calculated in each iteration. In each iteration, the residual value corresponding to the current solution vector is calculated, and the residual value is compared with the preset convergence criterion. When the residual value meets the convergence criterion, the iteration is terminated, and the final numerical solutions of temperature field distribution, pressure field distribution and vibration field distribution are output.

[0013] As a further aspect of the present invention, the specific process of using an iterative solution process to process the system of algebraic equations is as follows: Construct a coefficient matrix for the algebraic equation system. The coefficient matrix contains coefficient blocks for the temperature field, pressure field, vibration field, and inter-field coupling. Extract the diagonal elements of the temperature field coefficient block from the coefficient matrix to form the first submatrix. Extract the diagonal elements of the pressure field coefficient block to form the second submatrix. Extract the diagonal elements of the vibration field coefficient block to form the third submatrix. The first, second, and third submatrices are combined into a block diagonal matrix. The block diagonal matrix is ​​used to preprocess the algebraic equation system. The updated solution vector is obtained by solving the preprocessed algebraic equation system. The updated solution vector is then superimposed on the current solution vector to form a new solution vector.

[0014] As a further aspect of the present invention: in the control command sequence generation module, the specific process of generating the control command sequence based on the numerical solution is as follows: Extract multiphysics state vectors from numerical solutions, establish time evolution sequences of multiphysics state vectors, analyze the changing trends of time evolution sequences, and generate initial control instruction sets based on the changing trends; The initial control command set is input into the digital twin. The calculation process of the multiphysics coupling equation is executed in the digital twin. The changes in the multiphysics state vector during the execution process are recorded. The difference between the state vectors before and after the execution is calculated. An error vector is constructed based on the difference in the state vectors. The control command parameters are adjusted according to the error vector. A corrected control command set is generated through parameter adjustment. The calculation process and parameter adjustment process in the digital twin are repeated until the norm of the error vector is less than a set threshold. Finally, the control command sequence is output.

[0015] The beneficial effects of this invention are: This invention effectively solves the problem of isolated processing of various physical field data in traditional methods by integrating multi-physics field data such as temperature, pressure, and vibration fields into a unified tensor data structure, thus achieving collaborative analysis of multi-source data. By establishing a multi-level grid structure and data transmission paths, multi-physics field data is deeply integrated with a three-dimensional geometric model, forming a high-fidelity digital twin. By constructing a coupled system matrix and solving its eigenvalue problem, the intrinsic relationship between the temperature, pressure, and vibration fields is revealed, establishing a multi-physics field coupled model that accurately describes the overall operating state of the equipment. Through an efficient iterative solution algorithm, complex coupled equations are solved rapidly. Finally, a closed-loop optimization mechanism based on multi-physics field state vectors generates a control command sequence that matches the actual operating state of the equipment. These technologies work together to enable the system to comprehensively grasp the equipment's operating status, make accurate decisions under complex conditions, and significantly improve the reliability and operational efficiency of the intelligent well site management system. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a schematic diagram of the modules of the present invention. Detailed Implementation

[0018] 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, and 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.

[0019] Please see Figure 1 As shown, this invention is a smart well site integrated management system based on digital twin and AI analysis, comprising: The multiphysics data acquisition module is used to collect multiphysics data during the operation of well site equipment. The multiphysics data includes temperature field data, pressure field data, and vibration field data. The tensor data structure construction module is used to represent multiphysics data in a unified tensor data structure, which includes spatial dimension, time dimension and physical quantity dimension. The multiphysics data fusion module is used to fuse tensor data structures with the well site 3D geometric model and establish spatial relationships of multiphysics data at the grid nodes of the 3D geometric model. The multiphysics coupling modeling module is used to establish control equations based on the fused multiphysics data, and to associate the control equations of each physics field with each other through coupling operators to form multiphysics coupling equations. The coupled equation solving module is used to solve the multiphysics coupled equations and obtain numerical solutions for the temperature field distribution, pressure field distribution, and vibration field distribution. The control command sequence generation module is used to generate control command sequences based on numerical solutions. These control command sequences are used to adjust the operating parameters of well site equipment.

[0020] In a preferred embodiment of the present invention, the specific process of uniformly representing multi-physics data as a tensor data structure in the tensor data structure construction module is as follows: When constructing a dynamic spatiotemporal grid, the spatial dimension division rules of the grid must first be determined based on the topological structure of the three-dimensional geometric model of the well site equipment. The spatial dimension can use a Cartesian coordinate system or a cylindrical coordinate system. The spatial density of the grid nodes is adjusted according to the structural characteristics of the well site equipment type (e.g., pumping unit, wellhead equipment, downhole tubing). The grid node density corresponding to critical equipment components (e.g., motor bearings, pumping unit cranks) needs to be higher than that of non-critical areas to ensure data representation accuracy. The temporal dimension sets the sampling interval based on the changing frequency of the multi-physics data. The sampling interval for temperature field data is usually matched to the thermal conductivity characteristics of the equipment; the sampling interval for pressure field data needs to adapt to the wellhead pressure fluctuation cycle; and the sampling interval for vibration field data needs to satisfy the Nyquist sampling theorem for the equipment vibration frequency to ensure complete capture of the dynamic changes of each physical field.

[0021] Temperature field data is collected by distributed fiber optic sensors or contact temperature sensors deployed on the surface of well site equipment and downhole. The continuous analog signals output by the sensors are discretized and sampled according to a set time series. The sampled temperature values ​​are then matched to corresponding spatial nodes in a dynamic spatiotemporal grid based on the spatial deployment coordinates of the sensors using a spatial interpolation algorithm. Pressure field data comes from wellhead pressure transmitters and downhole pressure sensors. After being converted into digital signals by signal conditioning circuits, the data is sampled according to a time series. Based on the installation coordinates of the pressure sensors, the sampled data is mapped to corresponding nodes in the dynamic spatiotemporal grid. Vibration field data is collected by piezoelectric vibration sensors installed on key moving parts of the equipment. The collected vibration acceleration signals are discretized and sampled, then associated with corresponding nodes in the dynamic spatiotemporal grid based on the spatial installation location of the sensors.

[0022] At each node of the dynamic spatiotemporal grid, a data association algorithm is used to establish the correlation between the temperature, pressure, and vibration field data of that node at the same timestamp. The association process requires a time synchronization mechanism to calibrate the acquisition time of each physical field data point, ensuring a one-to-one correspondence between the multi-physics field data of the same node at the same timestamp. Subsequently, the associated multi-physics field data is integrated into a unified multi-physics field tensor. The spatial dimension of the tensor corresponds to the spatial coordinates of the grid node, the time dimension corresponds to the timestamp of data acquisition, and the physical quantity dimension corresponds to the physical quantity types of temperature, pressure, and vibration, respectively. Each element of the tensor directly corresponds to the single physical field data value of the grid node at a specific timestamp. Based on the time dimension's progression rules, the multi-physics field data in the dynamic spatiotemporal grid is updated at a set time step.

[0023] In another preferred embodiment of the present invention, the specific process of fusing the tensor data structure with the well site three-dimensional geometric model in the multiphysics data fusion module is as follows: When establishing a multi-layered mesh structure, a high-precision three-dimensional geometric model of the well site equipment must be used as the basis. Three mesh layers are then created based on the distribution characteristics of the physical field data and the structural features of the equipment. The surface mesh layer uses triangular faceted meshes, with the mesh node density determined by the deployment density of the surface temperature sensors. This ensures that each temperature sensor's acquisition location corresponds to a unique surface mesh node. Simultaneously, the mesh nodes are densified in temperature-sensitive areas such as wellhead flanges and valves. The volume mesh layer uses tetrahedral or hexahedral volume meshes, covering the internal flow channels and cavity structures of the equipment. The mesh division must match the measurement area of ​​the pressure sensors to ensure that pressure field data can be accurately mapped to volume mesh nodes, and the spatial resolution of the volume mesh must meet the requirements for fluid pressure gradient calculation. The structural mesh layer uses unit meshes consistent with the equipment's finite element mechanical model, directly relating to the equipment's mechanical structural components, such as the cranks and connecting rods of the pumping unit. The distribution of mesh nodes must be aligned with the installation positions of the vibration sensors to ensure that vibration field data can be directly mapped to the mesh nodes of key structural parts.

[0024] In the data mapping phase, temperature field data is mapped through spatial matching of sensor coordinates and surface grid node coordinates. A coordinate calibration algorithm is used to eliminate positional offsets caused by sensor installation deviations, ensuring that each temperature sample value is accurately assigned to its corresponding surface grid node. Pressure field data is allocated to corresponding volume grid nodes based on the measurement area of ​​the pressure sensor through region interpolation. If the measurement range of a single pressure sensor covers multiple volume grid nodes, data allocation is performed according to the principle of pressure distribution uniformity. Vibration field data is mapped based on the one-to-one correspondence between the vibration sensor's installation coordinates and the structural grid node coordinates. For sensors installed at structural component connections, structural mechanics analysis is needed to determine the structural grid nodes within their influence range, and vibration data is synchronously mapped to these nodes to reflect the overall vibration state of the structure.

[0025] The specific process for establishing data transfer paths between multi-level grid structures is as follows: When constructing the projection relationship between the surface mesh layer and the volume mesh layer, a spatial neighborhood search algorithm is used to determine the adjacent volume mesh nodes corresponding to each surface mesh node. Typically, a search domain with a fixed spatial radius is set with the surface mesh node as the center, and volume mesh nodes within the domain are selected as associated nodes. Orthogonal projection is preferred for projection relationships, projecting the data of the surface mesh node along the normal vector direction of the device surface to the adjacent volume mesh node. If there is no volume mesh node in the projection direction, the nearest neighbor projection method is used, selecting the volume mesh node closest to the surface mesh node as the projection target. Distance weight calculation is based on the Euclidean distance between the surface mesh node and its adjacent volume mesh nodes, using an inverse distance ratio method to determine the weight value. Volume mesh nodes that are closer are assigned higher weights. Simultaneously, weight normalization is performed to ensure that the sum of the weights of all adjacent volume mesh nodes corresponding to a single surface mesh node is 1, thus guaranteeing energy conservation during data transmission.

[0026] When constructing the interpolation relationship between the volume mesh layer and the structural mesh layer, it is necessary to first determine the set of volume mesh nodes corresponding to each structural mesh node through influence domain analysis, that is, to identify the volume mesh nodes that affect the mechanical state of that structural mesh node. The selection of the shape function needs to match the element types of the volume mesh and the structural mesh. If both are linear elements, a linear interpolation shape function is used; if they are higher-order elements, a quadratic or cubic interpolation shape function is used. The calculation of the shape function needs to be based on the spatial positional relationship between the volume mesh nodes and the structural mesh nodes, reflecting the degree of contribution of the volume mesh node data to the structural mesh node data.

[0027] In the data transfer matrix construction phase, the first data transfer matrix has a dimension of the number of surface grid nodes × the number of volume grid nodes. Each element in the matrix corresponds to the distance weight value from a single surface grid node to a single volume grid node, and the element value for nodes without established projection relationships is set to 0. The second data transfer matrix has a dimension of the number of volume grid nodes × the number of structure grid nodes. The matrix elements are the shape function values ​​from the volume grid node to the structure grid node, and the element value for nodes not within the influence domain is set to 0. The first and second data transfer matrices are multiplied according to the matrix multiplication rules to obtain a complete transfer matrix with a dimension of the number of surface grid nodes × the number of structure grid nodes. Each element of this matrix represents the comprehensive transfer coefficient of surface grid node data transferred to the structure grid node through the volume grid layer, thereby realizing a unified representation of multiphysics data in a multi-level grid structure.

[0028] In another preferred embodiment of the present invention, the specific process of establishing control equations based on the fused multiphysics data and linking the control equations of each physics field to each other through coupling operators in the multiphysics coupling modeling module is as follows: When constructing a multiphysics state vector, data from a multi-layered grid structure must be used as the basis. The vector dimension and element sorting rules are determined according to the physical field type and the spatial distribution of grid nodes. The temperature field component selects temperature sampling values ​​from all nodes in the surface grid layer, arranged sequentially according to the spatial index order of the grid nodes to form a temperature sub-vector. The pressure field component extracts pressure data from all nodes in the volume grid layer, similarly following the spatial topological order of the volume grid to construct a pressure sub-vector. The vibration field component must cover the dynamic mechanical parameters of the structural grid layer nodes, including node vibration displacement, vibration velocity, and vibration acceleration, and is integrated into a vibration sub-vector according to the mechanical association order of the structural components. The temperature, pressure, and vibration sub-vectors are then concatenated sequentially to form a complete multiphysics state vector. The vector dimension is equal to the sum of the number of nodes in the three types of grids, ensuring that the physical quantity data of each grid node corresponds to a unique element in the vector.

[0029] When establishing the evolution equations for the state vectors of a multiphysics field, the time-varying characteristics of the state vectors must be described based on the fundamental physical laws of each physical field. For the temperature field component, the evolution equations must reflect the temperature diffusion law during heat conduction and the influence of heat exchange between adjacent surface grid nodes on the rate of temperature change. For the pressure field component, the evolution equations must conform to the fluid flow law and reflect the effect of the pressure gradient and fluid seepage velocity between volume grid nodes on the rate of pressure change. For the vibration field component, the evolution equations must follow the structural dynamics equations and reflect the constraints of the inertial forces, damping forces, and elastic forces of the structural grid nodes on the rate of change of vibration parameters. When a coupled system matrix is ​​introduced into the evolution equation, the diagonal elements of the matrix are the self-acting terms of each physical field. The self-acting term of the temperature field is related to the thermal conductivity coefficient and the surface grid node spacing, the self-acting term of the pressure field is related to the fluid viscosity and the volume grid permeability, and the self-acting term of the vibration field is related to the structural stiffness coefficient and the damping coefficient. The off-diagonal elements of the matrix are the cross-acting terms, including the influence coefficient of temperature change on fluid viscosity, the influence coefficient of pressure gradient on structural vibration excitation, and the correction coefficient of temperature on structural elastic modulus. These terms characterize the interaction relationships between the physical fields.

[0030] The specific process for solving the eigenvalue problem of the coupled system matrix is ​​as follows: When constructing the mathematical expression of the coupled system matrix, the numerical calculation basis of the matrix elements needs to be determined according to the multiphysics interaction mechanism. The values ​​of self-acting terms need to be calibrated by combining experimental test data and theoretical models. For example, the self-acting terms of the temperature field need to be determined by experimental measurements of the thermal conductivity coefficient of the well site equipment materials, and the self-acting terms of the pressure field need to be calculated based on experimental data of the permeability of the reservoir rock. The values ​​of cross-acting terms need to be obtained through multiphysics coupling experiments. For example, the cross-acting terms of temperature and pressure need to measure the variation of fluid pressure at different temperatures and obtain the correlation coefficient through data fitting.

[0031] When calculating the characteristic polynomial of the coupled system matrix, the determinant expansion method in numerical linear algebra is used. The eigenvalues ​​are substituted into the matrix, and the determinant is calculated. Setting the determinant to zero yields the higher-order algebraic equations about the eigenvalues, i.e., the characteristic polynomial. To solve for the roots of the characteristic polynomial, considering the complexity of analytical solutions to higher-order equations, numerical iterative algorithms, such as the QR decomposition algorithm or the Jacobian algorithm, are used to iteratively approximate the numerical solutions of the eigenvalues, forming the eigenvalue spectrum of the coupled system matrix. When the eigenvalues ​​are substituted into the homogeneous equations of the coupled system matrix, Gaussian elimination is used to solve for non-zero solutions, obtaining the eigenvector corresponding to each eigenvalue. The dimension of the eigenvector is consistent with the multiphysics state vector, and each element corresponds to the modal amplitude of a physical quantity at a grid node.

[0032] When orthogonalizing eigenvectors, the Schmitt orthogonalization method is used. Each eigenvector is orthogonalized sequentially to eliminate linear correlations between different eigenvectors, ensuring that the processed eigenvectors satisfy the orthogonality condition. After orthogonalization, the eigenvectors are normalized to ensure that the magnitude of each eigenvector is 1, forming a set of standard orthogonal eigenvectors. Finally, the set of standard orthogonal eigenvectors is sorted according to the magnitude of their corresponding eigenvalues, typically in ascending order. Each standard orthogonal eigenvector corresponds to a multiphysics coupling mode, and these are combined to form a set of coupling modes.

[0033] In another preferred embodiment of the present invention, the specific process of solving the multiphysics coupling equation in the coupling equation solving module is as follows: When converting multiphysics coupling equations into a system of algebraic equations, the finite volume method or finite element method is used to discretize the coupling equations spatially and temporally. In the spatial discretization stage, based on the node distribution of a multi-level mesh structure, the spatial derivative terms in the coupling equations are replaced with difference expressions for the physical quantities of adjacent mesh nodes. Specifically, the spatial derivative discretization of the temperature field needs to consider the spacing of surface mesh nodes and the direction of heat conduction; the spatial derivative discretization of the pressure field needs to combine the element volume of the volume mesh with the seepage path; and the spatial derivative discretization of the vibration field needs to match the stiffness matrix topology of the structural mesh. In the temporal discretization stage, an explicit or implicit time integration scheme is used to convert the time derivative terms in the coupling equations into difference forms of the physical quantities of mesh nodes at different time steps. The selection of the time step must satisfy numerical stability conditions while balancing computational efficiency and solution accuracy. Through discretization, the continuous form of the multiphysics coupling equations is transformed into a system of algebraic equations with the physical quantities of mesh nodes as unknowns. The dimension of the equation system is consistent with the dimension of the multiphysics state vector.

[0034] When setting the initial solution vector of the algebraic equation system, it is necessary to obtain multiphysics data at the start-up time or steady-state operation of the well site equipment. The initial value of the temperature field comes from the ambient temperature measurement data before the equipment starts up. Combined with the spatial position of the surface grid nodes, the ambient temperature of the corresponding area is assigned to each surface grid node. The initial value of the pressure field uses the pressure sensor sampling data when the equipment is running unloaded. The initial pressure value is allocated according to the association relationship between the volume grid nodes and the sensors. The initial value of the vibration field is based on the vibration baseline measurement results when the equipment is stationary. The initial vibration displacement, velocity, and acceleration of the structural grid nodes are set to zero or the baseline measurement value. The three types of initial values ​​are integrated according to the arrangement order of the multiphysics state vectors to form the initial solution vector of the algebraic equation system.

[0035] The specific process of using an iterative solution method to process a system of algebraic equations is as follows: When constructing the coefficient matrix of the algebraic equation system, the matrix elements need to be determined based on the coefficient calculation results during the discretization process. The temperature field coefficient block is located in the upper left corner of the matrix, and the element values ​​correspond to the coefficients in the temperature field discretization equation, including the ratio of thermal conductivity coefficient to surface mesh node spacing, heat exchange coefficient between adjacent nodes, etc. The pressure field coefficient block is located in the middle of the matrix, and the element values ​​are related to the calculation results of fluid viscosity, volume mesh permeability, and element volume, reflecting the coefficient characteristics of the pressure field discretization equation. The vibration field coefficient block is located in the lower right corner of the matrix, and the element values ​​correspond to the discretization results of structural stiffness coefficient, damping coefficient, and mass matrix, reflecting the coefficient properties of the vibration field discretization equation. The inter-field coupling coefficient block is distributed between the above three types of main diagonal blocks, and the element values ​​are the discretization coefficients of multi-physics interaction terms, such as the correction coefficient of temperature change on fluid viscosity, the excitation coefficient of pressure gradient on structural vibration, etc., used to characterize the coupling effect between different physical fields.

[0036] In the sub-matrix extraction stage, diagonal elements are selected from the temperature field coefficient block and arranged in diagonal order to form the first sub-matrix. Diagonal elements are extracted from the pressure field coefficient block, maintaining their positional correspondence with the diagonal elements of the pressure field coefficient block to construct the second sub-matrix. Diagonal elements are obtained from the vibration field coefficient block and arranged in diagonal order to form the third sub-matrix. The first, second, and third sub-matrixes are placed at the main diagonal positions of the block diagonal matrix, and the remaining off-diagonal positions are filled with zero elements, forming the block diagonal matrix used for preprocessing.

[0037] During preprocessing, the inverse matrix of the block diagonal matrix is ​​calculated using a matrix inversion algorithm. This inverse matrix is ​​then multiplied on the left by both sides of the algebraic equation system to obtain the preprocessed algebraic equation system. An iterative solution algorithm is then used to process the preprocessed equation system. Common algorithms include the generalized minimum residual method or the successive over-relaxation algorithm. The updated value of the solution vector is obtained through iterative calculation. Finally, the updated value is added element-wise to the current solution vector to generate a new solution vector.

[0038] After each iteration, the residual value is calculated. The current solution vector is substituted into the left-hand side of the algebraic equation system, and the difference between the result on the left-hand side and the vector on the right-hand side of the equation system is calculated to form the residual vector. The residual value is obtained by calculating the L2 norm or infinite norm of the residual vector, and this residual value is compared with a preset convergence criterion. The convergence criterion is set according to the accuracy requirements of engineering calculations, and is usually a very small positive number. If the residual value is less than the convergence criterion, the iteration is considered to have converged, and the iteration process is terminated; if the residual value does not meet the convergence criterion, the current new solution vector is used as the initial solution for the next iteration, and the iteration calculation and residual judgment steps are repeated until the convergence condition is met, and finally, the numerical solutions of the temperature field, pressure field, and vibration field distribution are output.

[0039] In another preferred embodiment of the present invention, the specific process of generating the control command sequence based on the numerical solution in the control command sequence generation module is as follows: When extracting multiphysics state vectors from numerical solutions, the numerical solutions for each physical field distribution need to be transformed into a vector structure consistent with the previous modeling, according to the grid node correspondence of the temperature field, pressure field, and vibration field, ensuring a one-to-one correspondence between vector elements and physical quantities of grid nodes. Establishing a time evolution sequence requires arranging the multiphysics state vectors extracted at different times in chronological order according to the time step of the coupled equations, forming a time-series dataset. When analyzing trends, a time-series analysis method is used, fitting the variation curves of each physical field component through a sliding window to identify trend characteristics such as whether the temperature field is experiencing continuous heating or local overheating, whether the pressure field is showing abnormal fluctuations, and whether the vibration field is approaching the structural resonance threshold. This, combined with the controllable parameter range of the well site equipment, generates an initial control command set. The initial control commands must correspond to specific equipment adjustment parameters, including pumping unit stroke frequency, wellhead valve opening, and cooling system power, with parameter values ​​initially set based on trend characteristics.

[0040] After inputting the initial control command set into the digital twin, the digital twin needs to reproduce the physical structure and operating environment of the well site equipment. It executes multiphysics coupling equation calculations according to a preset time step, recording the multiphysics state vectors at each moment during the calculation process. When calculating the difference between the state vectors before and after execution, the state vector before command input is used as a reference, and the difference between the state vector after execution and the reference vector is calculated element by element, integrating them to form an error vector. When adjusting the control command parameters based on the error vector, a gradient optimization algorithm is used, aiming to minimize the error vector norm. The command parameters are corrected along the direction of error reduction; for example, the cooling system power parameters are increased when the temperature field error exceeds the limit, and the pumping unit balance block position parameters are adjusted when the vibration field error exceeds the standard, generating a corrected control command set. The above digital twin simulation calculation and parameter adjustment process is repeated, recalculating the error vector norm after each iteration until the norm is less than a set threshold. Finally, the optimized control commands are arranged in chronological order of execution, outputting a complete control command sequence.

[0041] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A smart well site integrated management system based on digital twin and AI analysis, characterized in that, include: The multiphysics data acquisition module is used to collect multiphysics data during the operation of well site equipment. The multiphysics data includes temperature field data, pressure field data, and vibration field data. The tensor data structure construction module is used to represent multiphysics data in a unified tensor data structure, which includes spatial dimension, time dimension and physical quantity dimension. The multiphysics data fusion module is used to fuse tensor data structures with the well site 3D geometric model and establish spatial relationships of multiphysics data at the grid nodes of the 3D geometric model. The multiphysics coupling modeling module establishes control equations based on the fused multiphysics data, and uses coupling operators to correlate the control equations of each physics field to form a multiphysics coupling equation. The coupled equation solving module is used to solve the multiphysics coupled equations and obtain numerical solutions for the temperature field distribution, pressure field distribution, and vibration field distribution. The control command sequence generation module is used to generate control command sequences based on numerical solutions. These control command sequences are used to adjust the operating parameters of well site equipment.

2. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 1, characterized in that, In the tensor data structure construction module, the specific process of uniformly representing multi-physics data as a tensor data structure is as follows: A dynamic spatiotemporal grid is constructed, which includes spatial and temporal dimensions; temperature field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series; pressure field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series; vibration field data is sampled and mapped to the corresponding nodes of the dynamic spatiotemporal grid according to the time series. Multiphysics data associations are established at each node of the dynamic spatiotemporal grid. Temperature field data, pressure field data, and vibration field data are integrated into a unified multiphysics tensor through these associations. The multiphysics data in the dynamic spatiotemporal grid is updated according to the time dimension.

3. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 1, characterized in that, In the multiphysics data fusion module, the specific process of fusing the tensor data structure with the well site three-dimensional geometric model is as follows: A multi-level mesh structure is established, which includes a surface mesh layer, a volume mesh layer, and a structural mesh layer. Temperature field data is mapped to the surface mesh nodes of the multi-level mesh structure, pressure field data is mapped to the volume mesh nodes of the multi-level mesh structure, and vibration field data is mapped to the structural mesh nodes of the multi-level mesh structure. Calculate the spatial correspondence between surface mesh nodes and volume mesh nodes, calculate the spatial correspondence between volume mesh nodes and structural mesh nodes, establish data transfer paths between surface mesh layers, volume mesh layers and structural mesh layers, and realize the unified representation of multiphysics data in multi-level mesh structures through data transfer paths.

4. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 3, characterized in that, The specific process for establishing data transfer paths between multi-level grid structures is as follows: A projection relationship is established between the surface mesh layer and the volume mesh layer. The projection relationship transfers the data of the surface mesh nodes to the adjacent volume mesh nodes. An interpolation relationship is established between the volume mesh layer and the structural mesh layer. The interpolation relationship transfers the data of the volume mesh nodes to the corresponding structural mesh nodes. Calculate the distance weights from surface mesh nodes to volume mesh nodes. The distance weights determine the intensity of data transmission. Calculate the shape function from volume mesh nodes to structure mesh nodes. The shape function determines the distribution of data transmission. By combining the distance weights and shape function, establish the first data transmission matrix from the surface mesh layer to the volume mesh layer. By combining shape function and distance weight, a second data transfer matrix is ​​established from the volume mesh layer to the structure mesh layer. Multiplying the two data transfer matrices yields the complete data transfer path from the surface mesh layer to the structure mesh layer.

5. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 1, characterized in that, In the multiphysics coupling modeling module, the specific process of establishing control equations based on the fused multiphysics data and intercorrelating the control equations of each physics field through coupling operators is as follows: Based on multi-physics data in a multi-level grid structure, a multi-physics state vector is constructed, which includes temperature field components, pressure field components and vibration field components. An evolution equation for the multiphysics state vector is established, which describes the change of the multiphysics state vector with time. A coupled system matrix is ​​introduced into the evolution equation, which represents the interaction between the temperature field component, pressure field component, and vibration field component. The eigenvalue problem of the coupled system matrix is ​​solved to obtain the set of coupled modes. Using the set of coupled modes as basis functions, the evolution equations of the multiphysics state vector are decomposed into modes. Through mode decomposition, a set of mutually coupled mode equations are obtained, which constitute the multiphysics coupling equations.

6. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 5, characterized in that, The specific process for solving the eigenvalue problem of the coupled system matrix is ​​as follows: Construct a mathematical expression for the coupled system matrix, which includes self-acting terms for the temperature field, pressure field, vibration field, and the cross-acting terms among the three. The characteristic polynomial of the coupled system matrix is ​​calculated. The characteristic polynomial is a high-order algebraic equation about the characteristic parameters. The roots of the characteristic polynomial are solved to obtain the eigenvalue spectrum of the coupled system matrix. The eigenvalues ​​are substituted into the homogeneous equations of the coupled system matrix to solve for the corresponding eigenvectors. The eigenvectors are orthogonalized to obtain the standard orthogonal eigenvector set. The standard orthogonal eigenvector set is sorted according to the eigenvalues ​​to form the coupled mode set.

7. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 1, characterized in that, In the coupling equation solving module, the specific process of solving the multiphysics coupling equation is as follows: The multiphysics coupling equations are converted into a system of algebraic equations. An initial solution vector is set for the system of algebraic equations, which includes initial values ​​for the temperature field, pressure field, and vibration field. An iterative solution process is used to process the system of algebraic equations, and the update amount of the solution vector is calculated in each iteration. In each iteration, the residual value corresponding to the current solution vector is calculated, and the residual value is compared with the preset convergence criterion. When the residual value meets the convergence criterion, the iteration is terminated, and the final numerical solutions of temperature field distribution, pressure field distribution and vibration field distribution are output.

8. The intelligent well site integrated management system based on digital twin and AI analysis according to claim 7, characterized in that, The specific process of using an iterative solution method to process a system of algebraic equations is as follows: Construct a coefficient matrix for the algebraic equation system. The coefficient matrix contains coefficient blocks for the temperature field, pressure field, vibration field, and inter-field coupling. Extract the diagonal elements of the temperature field coefficient block from the coefficient matrix to form the first submatrix. Extract the diagonal elements of the pressure field coefficient block to form the second submatrix. Extract the diagonal elements of the vibration field coefficient block to form the third submatrix. The first, second, and third submatrices are combined into a block diagonal matrix. The block diagonal matrix is ​​used to preprocess the algebraic equation system. The updated solution vector is obtained by solving the preprocessed algebraic equation system. The updated solution vector is then superimposed on the current solution vector to form a new solution vector.

9. A smart well site integrated management system based on digital twin and AI analysis according to claim 5, characterized in that, In the control command sequence generation module, the specific process of generating the control command sequence based on the numerical solution is as follows: Extract multiphysics state vectors from numerical solutions, establish time evolution sequences of multiphysics state vectors, analyze the changing trends of time evolution sequences, and generate initial control instruction sets based on the changing trends; The initial control command set is input into the digital twin. The calculation process of the multiphysics coupling equation is executed in the digital twin. The changes in the multiphysics state vector during the execution process are recorded. The difference between the state vectors before and after the execution is calculated. An error vector is constructed based on the difference in the state vectors. The control command parameters are adjusted according to the error vector. A corrected control command set is generated through parameter adjustment. The calculation process and parameter adjustment process in the digital twin are repeated until the norm of the error vector is less than a set threshold. Finally, the control command sequence is output.

Images