GIS temperature field simulation method and system based on model reduction and data assimilation

Abstract

A GIS temperature field simulation method and system based on model reduction and data assimilation includes: establishing a three-dimensional finite element model of an environmentally friendly GIS; constructing a reduced-order model based on simulation temperature field data vectors with different operating parameters; setting N groups of convective heat transfer coefficients for each measuring point and obtaining the simulation temperature of each measuring point; combining the simulation temperatures and corresponding convective heat transfer coefficients of all measuring points in each group into a state vector; inputting all state vectors into a pre-trained improved Kalman filter algorithm, which outputs an updated state vector; repeating the above steps until the difference between the simulation temperature and the corresponding measured temperature of all measuring points in the updated state vector is less than a set threshold; and performing simulation using the convective heat transfer coefficients of the currently updated state vector. This invention significantly improves simulation efficiency and accuracy, providing reliable technical support for the thermal performance evaluation and optimization design of GIS equipment.

Description

Technical Field

[0001] This invention relates to the field of equipment parameter simulation technology, and in particular to an environmentally friendly GIS temperature field simulation method based on model order reduction and data assimilation. Background Technology

[0002] Gas-insulated metal-included switchgear (GIS) is a complete set of switchgear and control equipment composed of electrical components such as circuit breakers, disconnecting switches, grounding switches, current transformers, voltage transformers, surge arresters, busbars, bushings or cable terminals, combined according to a certain wiring method, using SF6 gas or other gases as the insulating medium.

[0003] Thermal stability is one of the core issues in GIS equipment design, especially at high voltage levels (such as 550kV), where temperature rise caused by electromagnetic losses has a significant impact on equipment safety and performance. Temperature rise not only alters the physical properties of gases but also affects gas flow characteristics and heat transfer efficiency, thus influencing the distribution of temperature and electric fields. Simultaneously, microstructures such as millimeter-scale rounded radii and corrugated structures coexist with the meter-scale overall structure, forming a coupling relationship across spatial scales. In this context, to accurately capture electric field peaks and the thermal conduction effects of the boundary layer, the model must employ high-quality anisotropic meshes at sharp corners, gaps, and insulation interfaces. Furthermore, it must address the challenges of interpolation error and consistency management during data interaction between different solvers. This leads to a surge in the number of degrees of freedom after discretization of the overall model, further making the time and memory consumption for solving single operating conditions a computational bottleneck.

[0004] Meanwhile, in GIS temperature field simulation, the accuracy of the convective heat transfer coefficient directly affects the prediction of temperature rise. Because actual measurement of the convective heat transfer coefficient is difficult, and because the convective heat transfer coefficient varies in different regions under actual operating conditions with changes in temperature and gas velocity, changes in the convective heat transfer coefficient affect the gas flow characteristics, heat transfer process, and temperature field evolution, thus impacting the thermal stability and operational reliability of the equipment. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes an environmentally friendly GIS temperature field simulation method based on model order reduction and data assimilation.

[0006] The present invention adopts the following technical solution.

[0007] The first aspect of this invention proposes a GIS temperature field simulation method based on model order reduction and data assimilation, comprising: A three-dimensional model for environmentally friendly GIS finite element simulation is established; multiple electro-thermal-fluid multiphysics finite element simulations are performed on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient at each point in multiple set areas, to obtain the simulation temperature field data vector corresponding to different operating parameters; a reduced-order model is constructed based on the simulation temperature field data vectors with different operating parameters; the number of principal feature modes of the reduced-order model is determined according to the cumulative energy calculation ratio. Temperature sensors are installed at the corresponding measuring points. The initial current value is set to a constant value. Multiple measuring points are set from all the set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed reduced-order model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature fields is obtained as the simulation temperature of the corresponding measuring point. The simulated temperature and corresponding convective heat transfer coefficient of all measuring points in each group are combined into a state vector; the improved Kalman filter algorithm is pre-trained based on all the state vectors; the improved Kalman filter algorithm outputs an updated state vector; the improved Kalman filter algorithm is improved by introducing the energy accumulation ratio when generating an observation vector for each state vector. Every set period, N sets of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the reduced-order model constructed by inputting the set initial current value and the convective heat transfer coefficient of each set, until the difference between the simulated temperature and the corresponding measured temperature of all measurement points of the updated state vector is less than the set threshold. Using the convective heat transfer coefficient of each measuring point in the currently updated state vector, an electro-thermal-fluid multiphysics finite element simulation is performed to obtain the GIS temperature field simulation results.

[0008] Preferably, the three-dimensional model is meshed, with the fluid domain using a hexahedral mesh and the solid region using a tetrahedral mesh. The mesh density for nonlinear structures is set to be greater than that for existing structures. The nonlinear structures include sharp corners and component connections. Material parameters are set for the three-dimensional model, including specific heat capacity, heat transfer coefficient, electrical conductivity, and relative permittivity. Current inlet and outlet surfaces are set at both ends of the guide rod as excitation sources.

[0009] Preferably, the step of constructing a reduced-order model based on the simulated temperature field data vectors with different operating parameters specifically involves: Calculate the covariance matrix of the simulated temperature field data vector obtained from the simulation of all operating parameters; By performing eigenvalue decomposition on the covariance matrix, multiple eigenvalues ​​and eigenvectors are obtained, and the eigenvalues ​​and corresponding eigenvectors are sorted in descending order of eigenvalue. The predicted temperature field is equal to the sum of the products of the first r eigenvectors and the time evolution coefficients of the corresponding eigenvectors; r is the number of principal feature patterns. The mapping function of the reduced-order model is trained. When the convective heat transfer coefficients of all measuring points are input into the reduced-order model, the time evolution coefficient of each feature vector is obtained according to the mapping function of the reduced-order model, and the corresponding predicted temperature field is calculated.

[0010] Preferably, when constructing the reduced-order matrix, the number of corresponding principal feature modes is determined based on the cumulative energy ratio, specifically: The eigenvalue is equal to the square of the singular value divided by the simulated sample. The sum of the singular values ​​corresponding to the first n eigenvalues ​​is divided by the sum of the singular values ​​corresponding to all eigenvalues, and this is used as the energy accumulation ratio of the first n singular values. Obtain all n whose cumulative energy ratio is greater than a set ratio threshold, and find the number r of the smallest n as the main feature pattern.

[0011] Preferably, the step of setting multiple measuring points from multiple regions and setting N sets of convective heat transfer coefficients for each measuring point specifically involves: The multiple regions are defined as three areas: the GIS straight section, the bend section, and the flange-insulator neighborhood. Multiple measuring points are selected from these three regions. The convective heat transfer coefficient of all measuring points in the first group is the initial value of the convective heat transfer coefficient. The convective heat transfer coefficient of all measuring points in the second to Nth groups is increased by a first proportion of the initial value of the convective heat transfer coefficient compared to the convective heat transfer coefficient of the previous group. N is the number of groups of the convective heat transfer coefficient, and the number of groups is the same as the number of measuring points.

[0012] Preferably, the improved Kalman filter algorithm outputs an updated state vector, specifically: Generate an observation vector from each state vector; construct the observation noise covariance matrix based on the error of the temperature sensor at each measuring point; The difference vector is obtained by subtracting the first observation vector from the column vector composed of the measured temperature values ​​of all measuring points. Construct a predicted state anomaly matrix by calculating the average of all state vectors, and subtracting the average of all state vectors from each state vector to obtain each element of the predicted state anomaly matrix. Construct a predictive observation anomaly matrix by calculating the average of all observation vectors, and subtracting the average of all observation vectors from each observation vector to obtain each element of the predictive observation anomaly matrix; Based on the predicted state anomaly matrix and the predicted observation anomaly matrix, calculate the state-observation cross covariance matrix and the observation covariance matrix, respectively. Calculate the Kalman gain matrix based on the state-observation cross covariance matrix, the observation covariance matrix, and the observation noise covariance matrix; The updated state vector is obtained by adding the Kalman gain matrix to the first state vector and multiplying it by the difference vector.

[0013] Preferably, when generating an observation vector from each state vector, an energy accumulation ratio is introduced, specifically: The observation vector is the set selection matrix multiplied by the corresponding state vector, plus the process noise vector; Choose a matrix as one identity matrix and a It is composed of horizontally spliced ​​zero matrices; The process noise vector is the difference between the set process noise figure multiplied by 1 and the energy accumulation ratio of the first r singular values, multiplied by a factor where all elements are 1. Column vectors.

[0014] Preferably, the observation noise covariance matrix is ​​a diagonal matrix composed of the squares of the errors of the temperature sensors at each measuring point.

[0015] Preferably, the calculation of the Kalman gain matrix based on the state-observation cross-covariance matrix, the observation covariance matrix, and the observation noise covariance matrix specifically involves: The state-observation cross-covariance matrix is: multiplied by the transpose of the predicted state anomaly matrix and then divided by the difference between the number of groups and 1; The observation covariance matrix is ​​calculated by multiplying the predicted observation anomaly matrix by the transpose of the predicted observation anomaly matrix and then dividing by the difference between the number of groups and 1. The Kalman gain matrix is ​​calculated by adding the observation covariance matrix to the observation noise covariance matrix, calculating the inverse of the sum, and then multiplying the state-observation cross covariance matrix by the inverse matrix to obtain the Kalman gain matrix.

[0016] The second aspect of this invention proposes a GIS temperature field simulation system based on model reduction and data assimilation, using the method described in the first aspect of this invention. The system includes a simulation module, a model reduction construction module, a data assimilation module, and a temperature field simulation module, specifically: Simulation module: used to establish a three-dimensional model for environmentally friendly GIS finite element simulation; to perform multiple electro-thermal-fluid multiphysics finite element simulations on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient of each point in multiple set areas, to obtain the simulation temperature field data vector corresponding to different operating parameters; The order reduction model construction module is used to construct an order reduction model based on the simulated temperature field data vectors of different operating parameters. The number of main feature modes of the order reduction model is determined according to the cumulative energy ratio of the calculation. Temperature sensors are installed at the corresponding measuring points, the initial current value is set to a constant value, and multiple measuring points are set from all set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed order reduction model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature field is obtained as the simulation temperature of the corresponding measuring point. The data assimilation module is used to assemble the simulated temperatures and corresponding convective heat transfer coefficients of all measurement points in each group into a state vector. Based on all the state vectors, a pre-trained improved Kalman filter algorithm is input, and the improved Kalman filter algorithm outputs an updated state vector. The improved Kalman filter algorithm is improved by introducing an energy accumulation ratio when generating an observation vector for each state vector. After each set period, the N groups of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the reduced-order model constructed by inputting the set initial current value and the convective heat transfer coefficients of each group, until the difference between the simulated temperature and the corresponding measured temperature of all measurement points in the updated state vector is less than the set threshold. Temperature field simulation module: Used to perform electro-thermal-fluid multiphysics finite element simulation using the convective heat transfer coefficient of each measuring point with the currently updated state vector, and obtain GIS temperature field simulation results.

[0017] The beneficial effects of this invention are as follows: Compared with existing technologies, this invention, by combining order reduction technology and data assimilation methods, can correct the convective heat transfer coefficient in real time under dynamic operating conditions, significantly improving the accuracy of temperature field simulation results. Especially when the convective heat transfer coefficient is difficult to measure directly and varies dynamically across different regions, the algorithm-based localized data assimilation to correct the convective heat transfer coefficient effectively compensates for this deficiency, enabling the simulation results to more realistically reflect actual operating conditions. Simultaneously, by incorporating the truncation error generated by order reduction as a process noise term into the calculation process of ensemble Kalman filter data assimilation, an energy-aware model is formed. Thus, while maintaining accuracy, this invention successfully reduces the computational dimension of the simulation model through order reduction technology, thereby significantly improving computational efficiency. Compared to using an unreduced finite element simulation model for data assimilation, the reduced-order model can complete simulation calculations in a shorter time, greatly reducing the required computational resources. The environmentally friendly GIS temperature field simulation method of this invention can adapt to the practical application needs of environmentally friendly GIS in complex working conditions and variable environments, especially in the areas of alternative gas use and temperature rise control. By correcting model parameters in real time, it can effectively cope with temperature field changes under different environments, ensuring the long-term safe operation of equipment. In summary, this invention not only solves the shortcomings of traditional simulation methods in terms of accuracy, computational efficiency, and practical application, but also provides an efficient and reliable technical solution for the thermal stability assessment and optimization design of environmentally friendly GIS. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0020] like Figure 1 As shown, Embodiment 1 of the present invention proposes an environmentally friendly GIS temperature field simulation method based on model order reduction and data assimilation, including: A three-dimensional model for environmentally friendly GIS finite element simulation is established; multiple electro-thermal-fluid multiphysics finite element simulations are performed on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient at each point in multiple set areas, to obtain the simulation temperature field data vector corresponding to different operating parameters; a reduced-order model is constructed based on the simulation temperature field data vectors with different operating parameters; the number of principal feature modes of the reduced-order model is determined according to the cumulative energy calculation ratio. Temperature sensors are installed at the corresponding measuring points. The initial current value is set to a constant value. Multiple measuring points are set from all the set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed reduced-order model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature fields is obtained as the simulation temperature of the corresponding measuring point. Specifically, in this embodiment, N is set to 8; the initial current value is set to 5000A; The simulated temperature and corresponding convective heat transfer coefficient of all measuring points in each group are combined into a state vector; the improved Kalman filter algorithm is pre-trained based on all the state vectors; the improved Kalman filter algorithm outputs an updated state vector; the improved Kalman filter algorithm is improved by introducing the energy accumulation ratio when generating an observation vector for each state vector. Every set period, N sets of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the input of the set initial current value and the convective heat transfer coefficient of each set into the reduced-order model until the difference between the simulated temperature and the corresponding measured temperature at all measurement points of the updated state vector is less than the set threshold. Specifically, the set period is 30s and the set threshold is 0.1% of the measured temperature.

[0021] Using the convective heat transfer coefficient of each measuring point in the currently updated state vector, an electro-thermal-fluid multiphysics finite element simulation is performed to obtain the GIS temperature field simulation results.

[0022] In this preferred embodiment, the three-dimensional model is meshed, with the fluid domain using a hexahedral mesh and the solid region using a tetrahedral mesh. The mesh density for nonlinear structures is set to be greater than that for existing structures. The nonlinear structures include sharp corners and component connections. Material parameters are set for the three-dimensional model, including specific heat capacity, heat transfer coefficient, electrical conductivity, and relative permittivity. Current inlet and outlet surfaces are set at both ends of the guide rod as excitation sources.

[0023] It should be noted that the conductivity of the guide rod and the housing is set with temperature correction characteristics so that its conductivity changes with temperature, in order to reflect the feedback effect of temperature rise on conductor conductivity under actual conditions.

[0024] In this preferred embodiment, the construction of the reduced-order model based on the simulated temperature field data vectors with different operating parameters specifically involves: Calculate the covariance matrix of the simulated temperature field data vector obtained from the simulation of all operating parameters; covariance matrix Specifically:

[0025] in, For the first i The temperature field data vector of each simulation (the magnitude of the current and the magnitude of the convective heat transfer coefficient in each simulation are not specified). The number of simulations; Transpose symbol By performing eigenvalue decomposition on the covariance matrix, multiple eigenvalues ​​and eigenvectors are obtained. The eigenvalues ​​and their corresponding eigenvectors are then sorted in descending order of eigenvalue; the formula is as follows:

[0026] in, For the first q 1 eigenvector For the corresponding eigenvalues; The predicted temperature field is equal to the sum of the products of the first r eigenvectors and the time evolution coefficients of the corresponding eigenvectors; r is the number of principal feature patterns. The mapping function of the reduced-order model is trained. When the convective heat transfer coefficients of all measuring points are input into the reduced-order model, the time evolution coefficient of each feature vector is obtained according to the mapping function of the reduced-order model, and the corresponding predicted temperature field is calculated. The predicted temperature field (i.e., the temperature field's representation in low-dimensional space) is: :

[0027] in, Represents coordinates; Indicates time; The convective heat transfer coefficients of all measuring points at the current time are obtained by inputting the reduced-order model mapping function. q The time evolution coefficients of each eigenvector; For the first q Each eigenvector The value of the point; In this preferred embodiment, when constructing the reduced-order matrix, the number of corresponding principal feature modes is determined based on the cumulative energy ratio, specifically: The eigenvalue is equal to the square of the singular value divided by the simulated sample. The sum of the singular values ​​corresponding to the first n eigenvalues ​​is divided by the sum of the singular values ​​corresponding to all eigenvalues, and this is used as the energy accumulation ratio of the first n singular values. Obtain all n whose cumulative energy ratio is greater than a set threshold, and find the number r of the smallest n as the primary feature patterns; the formula is:

[0028] in, The energy accumulation ratio of the first r singular values; The singular value corresponding to the j-th eigenvalue; This represents the total number of eigenvalues.

[0029] The ratio threshold set in this embodiment is 0.95.

[0030] In this preferred embodiment, the step of setting multiple measuring points from multiple regions and setting N sets of convective heat transfer coefficients for each measuring point specifically involves: The multiple regions are defined as three areas: the GIS straight section, the bend section, and the flange-insulator neighborhood. Multiple measuring points are selected from these three regions. The convective heat transfer coefficient of all measuring points in the first group is the initial value of the convective heat transfer coefficient. The convective heat transfer coefficient of all measuring points in the second to Nth groups is increased by a first proportion of the initial value of the convective heat transfer coefficient compared to the convective heat transfer coefficient of the previous group. N is the number of groups of the convective heat transfer coefficient, and the number of groups is the same as the number of measuring points.

[0031] It should be noted that the first ratio set in this embodiment is 1%.

[0032] In this preferred embodiment, the improved Kalman filter algorithm outputs an updated state vector, specifically as follows: Generate an observation vector from each state vector; The state vector formula is:

[0033] in, This is the m-th state vector; The simulated temperatures at the 1st, 2nd, ..., 8th measurement points in the m-th state vector; Let be the convective heat transfer coefficient at the 1st, 2nd, ..., 8th measuring points in the m-th state vector; In this preferred embodiment, when generating an observation vector from each state vector, an energy accumulation ratio is introduced, specifically as follows: The observation vector is the set selection matrix multiplied by the corresponding state vector, plus the process noise vector, as shown in the formula:

[0034] in, This is the m-th observation vector; To select the matrix; This is the process noise vector; Choose a matrix as one identity matrix and a The zero matrix is ​​formed by horizontally concatenating the zero matrices, and the formula is:

[0035] The process noise vector is the difference between the set process noise figure multiplied by 1 and the energy accumulation ratio of the first r singular values, multiplied by a factor where all elements are 1. The column vector is given by the formula:

[0036] in, The process noise figure; For elements all equal to 1 Column vectors.

[0037] Construct the observation noise covariance matrix based on the error of the temperature sensor at each measuring point. ; In this preferred embodiment, the observation noise covariance matrix is ​​a diagonal matrix composed of the squares of the errors of the temperature sensors at each measuring point, and the formula is:

[0038] in, To observe the noise covariance matrix; , … These are the errors of the temperature sensors at the 1st, 2nd, ..., 8th measuring points, respectively.

[0039] The difference vector is obtained by subtracting the first observation vector from the column vector composed of the measured temperature values ​​of all measuring points. The formula is as follows:

[0040] in, It is a difference vector; This is a column vector consisting of the measured temperature values ​​of all measuring points; This is the first observation vector; Construct a predicted state anomaly matrix by calculating the average of all state vectors. Subtract the average of all state vectors from each state vector to obtain each element of the predicted state anomaly matrix. The formula is as follows:

[0041] in, For predicting the state anomaly matrix; It is the average value of all state vectors.

[0042] Construct a predictive observation anomaly matrix by calculating the average of all observation vectors. Subtract the average of all observation vectors from each observation vector to obtain each element of the predictive observation anomaly matrix, as shown in the formula:

[0043] in, To predict the observed anomaly matrix; This is the average value of all observed vectors.

[0044] Based on the predicted state anomaly matrix and the predicted observation anomaly matrix, calculate the state-observation cross covariance matrix and the observation covariance matrix, respectively. In this preferred embodiment, the state-observation cross-covariance matrix is: multiplied by the transpose of the predicted observation anomaly matrix and then divided by the difference between the number of groups and 1; the observation covariance matrix is: multiplied by the transpose of the predicted observation anomaly matrix and then divided by the difference between the number of groups and 1. The formula is:

[0045]

[0046] in, The state-observation cross-covariance matrix; To observe the covariance matrix.

[0047] Calculate the Kalman gain matrix based on the state-observation cross covariance matrix, the observation covariance matrix, and the observation noise covariance matrix; In this preferred embodiment, the Kalman gain matrix is ​​calculated as follows: The observation covariance matrix is ​​added to the observation noise covariance matrix, the inverse of the sum is obtained, and the Kalman gain matrix is ​​obtained by multiplying the state-observation cross-covariance matrix by the inverse matrix. The formula is:

[0048] in, The Kalman gain matrix; The updated state vector is obtained by adding the Kalman gain matrix to the first state vector and multiplying by the difference vector, as shown in the formula:

[0049] in, This is the updated state vector; This is the Kalman gain matrix.

[0050] It should be noted that after each set period, N sets of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients, specifically as follows: After the update, the convective heat transfer coefficient of all measuring points is used as the convective heat transfer coefficient of the first group of corresponding measuring points; the convective heat transfer coefficient of all measuring points from the second group to the Nth group is increased by the first proportion set by the convective heat transfer coefficient of the first group of corresponding measuring points compared with the convective heat transfer coefficient of the previous group.

[0051] Embodiment 2 of the present invention proposes a GIS temperature field simulation system based on model reduction and data assimilation according to the method described in the first aspect of the present invention, including a simulation module, a model reduction construction module, a data assimilation module, and a temperature field simulation module, specifically as follows: Simulation module: used to establish a three-dimensional model for environmentally friendly GIS finite element simulation; to perform multiple electro-thermal-fluid multiphysics finite element simulations on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient of each point in multiple set areas, to obtain the simulation temperature field data vector corresponding to different operating parameters; The order reduction model construction module is used to construct an order reduction model based on the simulated temperature field data vectors of different operating parameters. The number of main feature modes of the order reduction model is determined according to the cumulative energy ratio of the calculation. Temperature sensors are installed at the corresponding measuring points, the initial current value is set to a constant value, and multiple measuring points are set from all set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed order reduction model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature field is obtained as the simulation temperature of the corresponding measuring point. The data assimilation module is used to assemble the simulated temperatures and corresponding convective heat transfer coefficients of all measurement points in each group into a state vector. Based on all the state vectors, a pre-trained improved Kalman filter algorithm is input, and the improved Kalman filter algorithm outputs an updated state vector. The improved Kalman filter algorithm is improved by introducing an energy accumulation ratio when generating an observation vector for each state vector. After each set period, the N groups of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the reduced-order model constructed by inputting the set initial current value and the convective heat transfer coefficients of each group, until the difference between the simulated temperature and the corresponding measured temperature of all measurement points in the updated state vector is less than the set threshold. Temperature field simulation module: Used to perform electro-thermal-fluid multiphysics finite element simulation using the convective heat transfer coefficient of each measuring point with the currently updated state vector, and obtain GIS temperature field simulation results.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A GIS temperature field simulation method based on model order reduction and data assimilation, characterized in that, include: Establish a three-dimensional model for environmentally friendly GIS finite element simulation; Multiple electro-thermal-fluid multiphysics finite element simulations were performed on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient at each point in multiple set regions, to obtain the simulated temperature field data vectors corresponding to different operating parameters; a reduced-order model was constructed based on the simulated temperature field data vectors with different operating parameters; the number of principal characteristic modes of the reduced-order model was determined according to the cumulative energy ratio of the calculation. Temperature sensors are installed at the corresponding measuring points. The initial current value is set to a constant value. Multiple measuring points are set from all the set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed reduced-order model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature fields is obtained as the simulation temperature of the corresponding measuring point. The simulated temperature and corresponding convective heat transfer coefficient of all measuring points in each group are combined into a state vector; the improved Kalman filter algorithm is pre-trained based on all the state vectors; the improved Kalman filter algorithm outputs an updated state vector; the improved Kalman filter algorithm is improved by introducing the energy accumulation ratio when generating an observation vector for each state vector. Every set period, N sets of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the reduced-order model constructed by inputting the set initial current value and the convective heat transfer coefficient of each set, until the difference between the simulated temperature and the corresponding measured temperature of all measurement points of the updated state vector is less than the set threshold. Using the convective heat transfer coefficient of each measuring point in the currently updated state vector, an electro-thermal-fluid multiphysics finite element simulation is performed to obtain the GIS temperature field simulation results.

2. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 1, characterized in that: The three-dimensional model is meshed. The fluid domain in the three-dimensional model is divided into hexahedral meshes, and the solid region is divided into tetrahedral meshes. The mesh density of the nonlinear structure is set to be larger than that of the existing structure. The nonlinear structure includes sharp corners and component connections. Material parameters are set for the three-dimensional model, including specific heat capacity, heat transfer coefficient, electrical conductivity, and relative permittivity. Current inlet and outlet surfaces are set at both ends of the guide rod as excitation sources.

3. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 1, characterized in that: The construction of a reduced-order model based on simulation temperature field data vectors with different operating parameters is specifically as follows: Calculate the covariance matrix of the simulated temperature field data vector obtained from the simulation of all operating parameters; By performing eigenvalue decomposition on the covariance matrix, multiple eigenvalues ​​and eigenvectors are obtained, and the eigenvalues ​​and corresponding eigenvectors are sorted in descending order of eigenvalue. The predicted temperature field is equal to the sum of the products of the first r eigenvectors and the time evolution coefficients of the corresponding eigenvectors; r is the number of principal feature patterns. The mapping function of the reduced-order model is trained. When the convective heat transfer coefficients of all measuring points are input into the reduced-order model, the time evolution coefficient of each feature vector is obtained according to the mapping function of the reduced-order model, and the corresponding predicted temperature field is calculated.

4. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 2, characterized in that: When constructing the reduced-order matrix, the number of corresponding principal feature modes is determined based on the cumulative energy ratio, specifically: The eigenvalue is equal to the square of the singular value divided by the simulated sample. The sum of the singular values ​​corresponding to the first n eigenvalues ​​is divided by the sum of the singular values ​​corresponding to all eigenvalues, and this is used as the energy accumulation ratio of the first n singular values. Obtain all n whose cumulative energy ratio is greater than a set ratio threshold, and find the number r of the smallest n as the main feature pattern.

5. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 4, characterized in that: The process of setting multiple measuring points from multiple regions and setting N sets of convective heat transfer coefficients for each measuring point is as follows: The multiple regions are defined as three areas: the GIS straight section, the bend section, and the flange-insulator neighborhood. Multiple measuring points are selected from these three regions. The convective heat transfer coefficient of all measuring points in the first group is the initial value of the convective heat transfer coefficient. The convective heat transfer coefficient of all measuring points in the second to Nth groups is increased by a first proportion of the initial value of the convective heat transfer coefficient compared to the convective heat transfer coefficient of the previous group. N is the number of groups of the convective heat transfer coefficient, and the number of groups is the same as the number of measuring points.

6. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 5, characterized in that: The improved Kalman filter algorithm outputs an updated state vector, specifically: Generate an observation vector from each state vector; construct the observation noise covariance matrix based on the error of the temperature sensor at each measuring point; The difference vector is obtained by subtracting the first observation vector from the column vector composed of the measured temperature values ​​of all measuring points. Construct a predicted state anomaly matrix by calculating the average of all state vectors, and subtracting the average of all state vectors from each state vector to obtain each element of the predicted state anomaly matrix. Construct a predictive observation anomaly matrix by calculating the average of all observation vectors, and subtracting the average of all observation vectors from each observation vector to obtain each element of the predictive observation anomaly matrix; Based on the predicted state anomaly matrix and the predicted observation anomaly matrix, calculate the state-observation cross covariance matrix and the observation covariance matrix, respectively. Calculate the Kalman gain matrix based on the state-observation cross covariance matrix, the observation covariance matrix, and the observation noise covariance matrix; The updated state vector is obtained by adding the Kalman gain matrix to the first state vector and multiplying it by the difference vector.

7. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 6, characterized in that: When generating an observation vector from each state vector, an energy accumulation ratio is introduced, specifically: The observation vector is the set selection matrix multiplied by the corresponding state vector, plus the process noise vector; Choose a matrix as one identity matrix and a It is composed of horizontally spliced ​​zero matrices; The process noise vector is the difference between the set process noise figure multiplied by 1 and the energy accumulation ratio of the first r singular values, multiplied by a factor where all elements are 1. Column vectors.

8. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 6, characterized in that: The observation noise covariance matrix is ​​a diagonal matrix composed of the squares of the errors of the temperature sensors at each measuring point.

9. The GIS temperature field simulation method based on model order reduction and data assimilation according to claim 8, characterized in that: The calculation of the Kalman gain matrix based on the state-observation cross covariance matrix, the observation covariance matrix, and the observation noise covariance matrix is ​​specifically as follows: The state-observation cross-covariance matrix is: multiplied by the transpose of the predicted state anomaly matrix and then divided by the difference between the number of groups and 1; The observation covariance matrix is ​​calculated by multiplying the predicted observation anomaly matrix by the transpose of the predicted observation anomaly matrix and then dividing by the difference between the number of groups and 1. The Kalman gain matrix is ​​calculated by adding the observation covariance matrix to the observation noise covariance matrix, calculating the inverse of the sum, and then multiplying the state-observation cross covariance matrix by the inverse matrix to obtain the Kalman gain matrix.

10. A GIS temperature field simulation system based on model reduction and data assimilation according to any one of claims 1-9, comprising a simulation module, a model reduction construction module, a data assimilation module, and a temperature field simulation module, characterized in that: Simulation module: used to establish a three-dimensional model for environmentally friendly GIS finite element simulation; to perform multiple electro-thermal-fluid multiphysics finite element simulations on the three-dimensional model with different operating parameters, including the initial current value and the convective heat transfer coefficient of each point in multiple set areas, to obtain the simulation temperature field data vector corresponding to different operating parameters; The order reduction model construction module is used to construct an order reduction model based on the simulated temperature field data vectors of different operating parameters. The number of main feature modes of the order reduction model is determined according to the cumulative energy ratio of the calculation. Temperature sensors are installed at the corresponding measuring points, the initial current value is set to a constant value, and multiple measuring points are set from all set areas. N sets of convective heat transfer coefficients are set for each measuring point. The set initial current value and the convective heat transfer coefficient of each set are input into the constructed order reduction model to obtain the predicted temperature field of each set of convective heat transfer coefficients. The temperature of each measuring point in each set of predicted temperature field is obtained as the simulation temperature of the corresponding measuring point. The data assimilation module is used to assemble the simulated temperatures and corresponding convective heat transfer coefficients of all measurement points in each group into a state vector. Based on all the state vectors, a pre-trained improved Kalman filter algorithm is input, and the improved Kalman filter algorithm outputs an updated state vector. The improved Kalman filter algorithm is improved by introducing an energy accumulation ratio when generating an observation vector for each state vector. After each set period, the N groups of convective heat transfer coefficients are reset using the updated convective heat transfer coefficients. The above process is repeated starting from the reduced-order model constructed by inputting the set initial current value and the convective heat transfer coefficients of each group, until the difference between the simulated temperature and the corresponding measured temperature of all measurement points in the updated state vector is less than the set threshold. Temperature field simulation module: Used to perform electro-thermal-fluid multiphysics finite element simulation using the convective heat transfer coefficient of each measuring point with the currently updated state vector, and obtain GIS temperature field simulation results.

Images