Method and device for reducing order of multi-physics field steady-state simulation of flexible direct-current converter valve sub-module

Abstract

The application discloses a flexible direct-current converter valve submodule multi-physical field steady-state simulation dimension reduction method and device. The method comprises the following steps: constructing a plurality of training samples according to the dimension reduction simulation data and the monitoring derivation mixed data corresponding to each set of operation related data; performing multi-round iterative training on a neural network model based on the plurality of training samples until a preset training end condition is reached to obtain an accurate dimension reduction model; obtaining a current set of operation related data corresponding to a target submodule; determining the dimension reduction simulation data corresponding to the target submodule according to the current set of operation related data corresponding to the target submodule; and inputting the dimension reduction simulation data corresponding to the target submodule into the accurate dimension reduction model to determine the three-dimensional multi-physical field data corresponding to the target submodule.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and apparatus for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule. Background Technology

[0002] High-voltage, high-capacity flexible DC converter valves are core equipment in flexible DC transmission projects, and their operation, maintenance, and repair directly affect the reliability and availability of these projects. Due to the complex structure and numerous components of flexible DC converter valves, it is necessary to monitor the operating status of each sub-module within the flexible DC converter valve to ensure its stable operation.

[0003] Currently, although it is possible to monitor each sub-module in the flexible DC converter valve, it is limited to monitoring the capacitor voltage and whether the components are functioning normally. The acquired status variables are Boolean data rather than numerical data, which can only detect sub-module failures and does not provide early warning. Summary of the Invention

[0004] This application provides a method and apparatus for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule. The main purpose is to realize the rapid and real-time simulation calculation of the three-dimensional multiphysics field of the submodule based on the operation-related data of the submodule in the flexible DC converter valve, thereby realizing the real-time monitoring and analysis of the three-dimensional multiphysics field distribution of the submodule.

[0005] To address the aforementioned technical problems, this application provides the following technical solutions:

[0006] In a first aspect, this application provides a method for determining the order reduction in multiphysics steady-state simulation of a flexible DC converter valve submodule, the method comprising:

[0007] Obtain multiple sets of operation-related data corresponding to each sub-module, and determine the simulation data corresponding to each set of operation-related data based on the multiphysics coupling simulation model corresponding to each sub-module;

[0008] The simulation data is subjected to dimensionality reduction processing according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of running-related data.

[0009] Multiple training samples are constructed based on the dimensionality reduction simulation data and monitoring derivation mixed data corresponding to each of the aforementioned operation-related data sets;

[0010] The neural network model is iteratively trained multiple times based on multiple training samples until a preset training termination condition is reached to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function.

[0011] Obtain the currently running data set corresponding to the target submodule;

[0012] Based on the current running data set corresponding to the target sub-module, determine the dimensionality reduction simulation data corresponding to the target sub-module;

[0013] The dimensionality reduction simulation data corresponding to the target submodule is input into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule.

[0014] Secondly, this application also provides a multiphysics steady-state simulation order reduction device for a flexible DC converter valve submodule, the device comprising:

[0015] The first determining unit is used to acquire multiple sets of operation-related data corresponding to each sub-module, and to determine the simulation data corresponding to each set of operation-related data according to the multiphysics coupling simulation model corresponding to each sub-module.

[0016] The data dimensionality reduction unit is used to perform data dimensionality reduction processing on multiple simulation data according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of running-related data.

[0017] The construction unit is used to construct multiple training samples based on the dimensionality reduction simulation data and monitoring derivation mixed data corresponding to each of the operation-related data sets;

[0018] The training unit is used to perform multiple rounds of iterative training on the neural network model based on multiple training samples until a preset training termination condition is reached in order to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function.

[0019] The acquisition unit is used to acquire the current running-related data set corresponding to the target submodule;

[0020] The second determining unit is used to determine the dimensionality reduction simulation data corresponding to the target sub-module based on the current running relevant data set corresponding to the target sub-module;

[0021] The third determining unit is used to input the dimensionality reduction simulation data corresponding to the target sub-module into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target sub-module.

[0022] Thirdly, embodiments of this application provide a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0023] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0024] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0025] By employing the above-described technical solution, the technical solution provided in this application has at least the following advantages:

[0026] This application provides a method and apparatus for multiphysics steady-state simulation order reduction of a flexible DC converter valve submodule. This application enables the simulation order reduction application to iteratively train a neural network model based on multiple training samples to obtain an accurate order reduction model. When it is necessary to determine the three-dimensional multiphysics data corresponding to the target submodule in the target flexible DC converter valve, the simulation order reduction application first obtains the current operating relevant data set corresponding to the target submodule, then determines the dimensionality reduction simulation data corresponding to the target submodule based on the current operating relevant data set, and finally inputs the dimensionality reduction simulation data corresponding to the target submodule into the accurate order reduction model, thereby determining the three-dimensional multiphysics data corresponding to the target submodule. In this application, dimensionality-reduced simulation data and monitoring-derived mixed data are used as training samples to train a neural network model, thereby obtaining an accurate dimensionality-reduced model. Compared to iteratively training a neural network model using high-dimensional simulation data and monitoring-derived mixed data as training samples, using dimensionality-reduced simulation data and monitoring-derived mixed data as training samples effectively reduces the time and computational complexity consumed in training the neural network model, resulting in faster output of the accurate dimensionality-reduced model. Furthermore, the dimensionality-reduced simulation data retains important information from the high-dimensional simulation data, and the monitoring-derived mixed data is more consistent with actual operating conditions than the dimensionality-reduced simulation data. Therefore, the output of the accurate dimensionality-reduced model obtained through training is closer to the actual three-dimensional multiphysics field data of the submodule. Thus, this application enables rapid and real-time simulation calculation of the three-dimensional multiphysics field of the submodule based on the corresponding operational data of the submodule in the flexible DC converter valve, thereby achieving real-time monitoring and analysis of the three-dimensional multiphysics field distribution of the submodule.

[0027] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0028] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein:

[0029] Figure 1 This paper presents a flowchart of a multiphysics steady-state simulation order reduction method for a flexible DC converter valve submodule according to an embodiment of this application.

[0030] Figure 2 This paper presents a flowchart of another method for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule according to an embodiment of this application.

[0031] Figure 3 This paper shows a block diagram of a multiphysics steady-state simulation order reduction device for a flexible DC converter valve submodule provided in an embodiment of this application.

[0032] Figure 4 This paper presents a block diagram of another flexible DC converter valve submodule multiphysics steady-state simulation order reduction device provided in an embodiment of this application. Detailed Implementation

[0033] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0034] Furthermore, the terms “first,” “second,” and similar terms used in this application do not indicate any order, quantity, or importance, but are merely used to distinguish different parts.

[0035] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0036] Currently, although it is possible to monitor each sub-module in the flexible DC converter valve, it is limited to monitoring the capacitor voltage and whether the components are functioning normally. The acquired status variables are Boolean data rather than numerical data, which can only detect sub-module failures and does not provide early warning.

[0037] To determine the three-dimensional multiphysics data corresponding to each submodule in a flexible DC converter valve, embodiments of this application provide a method for order reduction in multiphysics steady-state simulation of flexible DC converter valve submodules, such as... Figure 1 As shown, the method includes at least 101-107.

[0038] 101. Obtain multiple sets of operation-related data corresponding to each submodule, and determine the simulation data corresponding to each set of operation-related data based on the multiphysics coupling simulation model corresponding to each submodule.

[0039] In the embodiments of this application, the execution entity in each step is a simulation downgrading application running on the target terminal device, wherein the target terminal device may be, but is not limited to, a computer, a server, a laptop, etc.

[0040] The target flexible DC converter valve is one whose three-dimensional multiphysics field data for each of its sub-modules needs to be determined. During the steady-state operation of the target flexible DC converter valve under different operating conditions, various operational-related data need to be collected for each sub-module, resulting in multiple operational-related data sets for each sub-module. For any given sub-module, the operational-related data set may include, but is not limited to: the input current and input voltage of that sub-module, the operational data of the water-cooling system (such as cooling water flow rate and cooling water temperature), and the ambient temperature of that sub-module, etc. For any given sub-module, its corresponding multiphysics field coupled simulation model is based on the operational data of that sub-module. The simulation model is constructed using parameters (such as physical parameters, performance parameters, and technical parameters) and simultaneously considers the interactions of physical fields such as temperature field, magnetic field, and electromagnetic field of the submodule. For any set of operation-related data, the corresponding simulation data may include, but is not limited to, the simulated three-dimensional temperature field data, simulated three-dimensional magnetic field data, and simulated three-dimensional fluid field data of the submodule corresponding to that set of operation-related data. The simulated three-dimensional temperature field data of the submodule includes the simulated temperature values ​​at multiple target locations in the submodule, the simulated three-dimensional magnetic field data of the submodule includes the simulated magnetic induction intensity at multiple target locations in the submodule, and the simulated three-dimensional fluid field data of the submodule includes the simulated fluid data (such as simulated cooling water flow rate, simulated cooling water pressure, etc.) at multiple target locations in the submodule.

[0041] In order to accurately determine the three-dimensional multiphysics field data corresponding to each sub-module in the target flexible DC converter valve during steady-state operation, the simulation reduction application needs to be trained to obtain an accurate reduction model. The first step in training to obtain an accurate reduction model is to acquire multiple operation-related data sets corresponding to each sub-module, and determine the simulation data corresponding to each operation-related data set based on the multiphysics field coupled simulation model corresponding to each sub-module. That is, for any operation-related data set, the multiphysics field coupled simulation model of the sub-module corresponding to the operation-related data set is controlled to run the simulation based on the operation-related data set, and the simulation data corresponding to the operation-related data set is obtained based on the simulation results.

[0042] 102. Perform dimensionality reduction processing on multiple simulation data according to the preset dimensionality reduction algorithm to obtain the dimensionality-reduced simulation data corresponding to each set of running relevant data.

[0043] The preset dimensionality reduction algorithm can be, but is not limited to, the orthogonal decomposition algorithm (POD).

[0044] After determining the simulation data corresponding to each runtime-related dataset, the simulation dimensionality reduction application can perform dimensionality reduction processing on the simulation data corresponding to multiple runtime-related datasets according to a preset dimensionality reduction algorithm, thereby obtaining the dimensionality-reduced simulation data corresponding to each runtime-related dataset.

[0045] It should be noted that the purpose of performing dimensionality reduction processing on multiple simulation data according to the preset dimensionality reduction algorithm is to convert high-dimensional data into low-dimensional data, thereby reducing the complexity of the data while retaining important information.

[0046] 103. Construct multiple training samples based on the dimensionality reduction simulation data and monitoring derivation mixed data corresponding to each set of operation-related data.

[0047] For any given submodule, the multiple target locations within that submodule can be categorized into first-type target locations and second-type target locations. For any first-type target location, the physical field data can be directly measured. However, for any second-type target location, the physical field data cannot be directly measured. Therefore, during the steady-state operation of the target flexible DC converter valve under certain operating conditions, after acquiring the operational data set corresponding to a specific submodule, it is also necessary to measure the actual data (such as actual temperature values, actual magnetic field strength, and actual fluid data) at multiple first-type target locations within that submodule. This process involves obtaining the actual monitoring data corresponding to the operational data set (i.e., the actual data at multiple first-class target locations). Furthermore, based on the actual monitoring data corresponding to the operational data set, it is necessary to determine the monitoring-derived mixed data corresponding to the operational data set. Specifically, this involves first determining the derived data at each second target location based on the spatial relationships between multiple second target locations and multiple first target locations within the submodule corresponding to the operational data set, and the actual data at each first-class target location. Then, based on the actual data at each first target location and the derived data at each second target location, the monitoring-derived mixed data corresponding to the operational data set is determined.

[0048] Therefore, for any set of operation-related data, the corresponding monitoring and derivation mixed data includes monitoring and derivation mixed three-dimensional temperature field data, monitoring and derivation mixed three-dimensional magnetic field data, and monitoring and derivation mixed three-dimensional fluid field data, which are composed of real data at each first type of target location and derivation data at each second target location in the corresponding sub-module.

[0049] Among them, for any set of operation-related data, the corresponding monitoring and derivation hybrid data is more consistent with the actual working conditions than the dimensionality reduction simulation data, and the corresponding monitoring and derivation hybrid data has the same dimension as the dimensionality reduction simulation data.

[0050] After obtaining the dimensionality-reduced simulation data corresponding to each set of runtime-related data, the simulation reduction application can construct multiple training samples based on the dimensionality-reduced simulation data and the monitoring-derived mixed data corresponding to each set of runtime-related data. That is, for any set of runtime-related data, the dimensionality-reduced simulation data and the monitoring-derived mixed data corresponding to that set of runtime-related data are combined into a single training sample, thereby obtaining multiple training samples.

[0051] 104. Perform multiple rounds of iterative training on the neural network model based on multiple training samples until the preset training termination condition is reached, in order to obtain an accurate reduced-order model.

[0052] In this application, the neural network model mentioned in the embodiments can specifically be a feedforward neural network (FNN), a convolutional neural network (CNN), a recurrent neural network (RNN), or other types of neural network models. This application does not specifically limit this type of model. The preset training termination condition can be, but is not limited to, minimizing the loss function, reaching a preset threshold for the number of iterations, reaching a preset threshold for the iteration duration, etc. The loss function corresponding to the neural network function can be, but is not limited to, a mean squared error loss function, a weighted loss function, or a regularized loss function. This application does not specifically limit this type of model.

[0053] After constructing multiple training samples, the simulation reduction application can perform multiple rounds of iterative training on the neural network model based on these training samples until the preset training termination condition is reached, thereby obtaining an accurate reduced-order model.

[0054] In practical applications, a portion of the training samples constructed in step 103 can be used to form a training sample set, and the remaining training samples can be used to form a validation sample set. In this step, the neural network model is iteratively trained using the multiple training samples contained in the training sample set. After obtaining an accurate order reduction model through training, the performance indicators (such as mean square error, R-squared value, etc.) of the accurate order reduction model are calculated based on the validation sample set. The accurate order reduction model is then adjusted according to the performance indicators of the accurate order reduction model, thereby improving the generalization ability of the accurate order reduction model.

[0055] It should be noted that, compared to using a mixture of high-dimensional simulation data and monitoring-derived data as training samples for iterative training of a neural network model, using dimensionality-reduced simulation data and monitoring-derived data as training samples for iterative training can effectively reduce the time and computational complexity consumed in training the neural network model, thereby enabling the accurately reduced-dimensional model to output computational results faster. In addition, the dimensionality-reduced simulation data retains important information from the high-dimensional simulation data, and the monitoring-derived data is more consistent with actual working conditions than the dimensionality-reduced simulation data. Therefore, the output results of the accurately reduced-dimensional model obtained through training are closer to the actual three-dimensional multiphysics data of the submodule.

[0056] 105. Obtain the current running data set corresponding to the target submodule.

[0057] The target submodule can be any one of the submodules in the target flexible DC converter valve.

[0058] After completing steps 101-104 above, an accurate reduced-order model can be trained and obtained. During the steady-state operation of the target flexible DC converter valve under a certain operating condition, when it is necessary to determine the three-dimensional multiphysics data corresponding to the target submodule, the simulation reduction application needs to acquire the current operational data set corresponding to the target submodule. This current operational data set may include, but is not limited to, the input current and input voltage of the target submodule at the current moment, the operating data of the water-cooling system corresponding to the target submodule at the current moment (such as cooling water flow rate, cooling water temperature, etc.), and the ambient temperature corresponding to the target submodule at the current moment, etc.

[0059] 106. Based on the current running data set corresponding to the target submodule, determine the dimensionality reduction simulation data corresponding to the target submodule.

[0060] After obtaining the current running data set corresponding to the target submodule, the simulation reduction application can determine the dimensionality-reduced simulation data corresponding to the target submodule based on the current running data set. The specific process is as follows: First, the multiphysics coupling simulation model corresponding to the target submodule is controlled to run the simulation based on the current running data set; second, the simulation data corresponding to the target submodule is obtained based on the simulation results; finally, the simulation data corresponding to the target submodule is processed by a preset dimensionality reduction algorithm to obtain the dimensionality-reduced simulation data corresponding to the target submodule.

[0061] 107. Input the dimensionality reduction simulation data corresponding to the target submodule into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule.

[0062] After determining the dimensionality reduction simulation data corresponding to the target submodule, the simulation reduction application can input the dimensionality reduction simulation data corresponding to the target submodule into the accurate dimensionality reduction model. After receiving the dimensionality reduction simulation data corresponding to the target submodule, the accurate dimensionality reduction model can determine the 3D multiphysics data corresponding to the target submodule based on the dimensionality reduction simulation data corresponding to the target submodule, and output the 3D multiphysics data corresponding to the target submodule. At this time, the simulation reduction application can obtain the 3D multiphysics data corresponding to the target submodule.

[0063] The target submodule's corresponding three-dimensional multiphysics data may include, but is not limited to, three-dimensional temperature field data, three-dimensional magnetic field data, and three-dimensional fluid field data, etc.

[0064] This application provides a method for reducing the order of a flexible DC converter valve submodule through multiphysics steady-state simulation. This method enables the simulation reduction application to iteratively train a neural network model based on multiple training samples to obtain an accurate reduced-order model. When it is necessary to determine the three-dimensional multiphysics data corresponding to the target submodule in the target flexible DC converter valve, the simulation reduction application first obtains the current operational data set corresponding to the target submodule, then determines the reduced-order simulation data corresponding to the target submodule based on the current operational data set, and finally inputs the reduced-order simulation data corresponding to the target submodule into the accurate reduced-order model, thereby determining the three-dimensional multiphysics data corresponding to the target submodule. In this embodiment, the dimensionality-reduced simulation data and the monitoring-derived mixed data are used as training samples to train the neural network model, thereby obtaining an accurate dimensionality-reduced model. Compared to iteratively training the neural network model using high-dimensional simulation data and monitoring-derived mixed data as training samples, using dimensionality-reduced simulation data and monitoring-derived mixed data as training samples effectively reduces the time and computational complexity consumed in training the neural network model, resulting in a faster output of the calculated results from the trained accurate dimensionality-reduced model. Furthermore, the dimensionality-reduced simulation data retains important information from the high-dimensional simulation data, and the monitoring-derived mixed data is more consistent with actual working conditions than the dimensionality-reduced simulation data. Therefore, the output results of the trained accurate dimensionality-reduced model are closer to the actual three-dimensional multiphysics field data of the submodule. Thus, in this embodiment, based on the operational data corresponding to the submodule in the flexible DC converter valve, rapid and real-time simulation calculations of the three-dimensional multiphysics field of the submodule can be achieved, thereby enabling real-time monitoring and analysis of the three-dimensional multiphysics field distribution of the submodule.

[0065] To illustrate this in more detail, this application provides another method for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule, as detailed below. Figure 2 As shown, the method includes at least 201-208.

[0066] 201. Obtain multiple sets of operation-related data corresponding to each submodule, and determine the simulation data corresponding to each set of operation-related data based on the multiphysics coupling simulation model corresponding to each submodule.

[0067] Specifically, in this step, after obtaining multiple runtime-related data sets corresponding to each submodule, the simulation reduction application determines the simulation data corresponding to each runtime-related data set based on the multiphysics coupling simulation model corresponding to each submodule as follows:

[0068] For any given submodule, firstly, obtain the corresponding operating parameters, such as physical parameters, performance parameters, technical parameters, etc.; secondly, construct the corresponding multiphysics coupling simulation model based on the operating parameters; finally, obtain a set of operating-related data for the submodule, and control the multiphysics coupling simulation model for the submodule to run the simulation based on the set of operating-related data, and obtain the simulation data for the submodule based on the simulation results.

[0069] Based on the above method, the simulation data corresponding to each set of runtime-related data can be determined.

[0070] 202. Perform dimensionality reduction processing on multiple simulation data according to the preset dimensionality reduction algorithm to obtain the dimensionality-reduced simulation data corresponding to each set of running relevant data.

[0071] Specifically, in this step, the simulation dimensionality reduction application performs dimensionality reduction processing on multiple simulation datasets according to a preset dimensionality reduction algorithm to obtain the dimensionality-reduced simulation data corresponding to each relevant dataset. The specific process is as follows:

[0072] For any set of simulation data, firstly, construct the full-order simulation data matrix X corresponding to the simulation data, and then center the full-order simulation data matrix X to obtain the centralized data matrix X. c X c =X-μ, where μ is the mean vector corresponding to the full-order simulation data matrix X; secondly, calculate the centered data matrix X. c The corresponding covariance matrix C is obtained, and eigenvalue decomposition is performed on the covariance matrix C to obtain multiple eigenvalues ​​λ corresponding to the covariance matrix C. i and the eigenvector Φ corresponding to each eigenvalue i Where C=X c X c T / m, where m is the number of simulation data sets, CΦ i =λ i Φ i Furthermore, based on multiple eigenvalues ​​λ i and the eigenvector Φ corresponding to each eigenvalue i Construct a low-dimensional data matrix X POD That is, first process multiple eigenvalues ​​λ i Sort in descending order, then sort by the eigenvectors λ corresponding to the top K eigenvalues. i Construct the eigenvector matrix Φ k Where K is a positive integer; then, based on the full-order simulation data matrix X and the eigenvector matrix Φ k Generate a low-dimensional data matrix X POD , where XPOD =Φ k T X; Finally, based on the low-dimensional data matrix, determine the dimensionality-reduced simulation data corresponding to the relevant data set of the operation to which the simulation data belongs, that is, combine the multiple columns of data contained in the low-dimensional data matrix into the dimensionality-reduced simulation data.

[0073] 203. Determine the monitoring and derivation mixed data corresponding to each set of operation-related data, and construct multiple training samples based on the dimensionality reduction simulation data and monitoring and derivation mixed data corresponding to each set of operation-related data.

[0074] After obtaining the dimensionality-reduced simulation data corresponding to each set of runtime-related data, the simulation reduction application needs to determine the monitoring-derived mixed data corresponding to each set of runtime-related data, and construct multiple training samples based on the dimensionality-reduced simulation data and monitoring-derived mixed data corresponding to each set of runtime-related data.

[0075] Specifically, in this step, for any submodule, the multiple target locations within that submodule can be divided into first-type target locations and second-type target locations. For any first-type target location, the physical field data at that location can be directly measured. However, for any second-type target location, the physical field data cannot be directly measured. Therefore, during the steady-state operation of the target flexible DC converter valve under certain operating conditions, after acquiring the operational data set corresponding to a certain submodule, it is also necessary to measure the actual data (such as actual temperature values, actual magnetic induction intensity, and actual fluid data) at multiple first-type target locations within that submodule to obtain the actual data corresponding to that operational data set. Monitoring data (i.e., real data at multiple first-type target locations); when it is necessary to determine the monitoring-derived mixed data corresponding to the operation-related data set, the real monitoring data corresponding to the operation-related data set can be obtained first, and then the monitoring-derived mixed data corresponding to the operation-related data set can be determined based on the real monitoring data corresponding to the operation-related data set. That is, firstly, based on the spatial relationship between multiple second target locations and multiple first target locations in the sub-module corresponding to the operation-related data set and the real data at each first-type target location, the derived data at each second target location can be determined, and then the monitoring-derived mixed data corresponding to the operation-related data set can be determined based on the real data at each first target location and the derived data at each second target location.

[0076] Therefore, for any set of operation-related data, the corresponding monitoring and derivation mixed data includes monitoring and derivation mixed three-dimensional temperature field data, monitoring and derivation mixed three-dimensional magnetic field data, and monitoring and derivation mixed three-dimensional fluid field data, which are composed of real data at each first type of target location and derivation data at each second target location in the corresponding sub-module.

[0077] 204. Perform multiple rounds of iterative training on the neural network model based on multiple training samples until the preset training termination condition is reached, so as to obtain an accurate reduced-order model.

[0078] The neural network model consists of an input layer, multiple hidden layers, and an output layer. Each hidden layer can be represented as:

[0079] z L =W L h L-1 +b L

[0080] Among them, z L For the Lth hidden layer, W L Let h be the weight matrix of the Lth hidden layer. L-1 = σ (z) L-1 ), σ The activation function can be, but is not limited to, ReLU, Sigmoid, etc. L-1 For the (L-1)th hidden layer, b L This is the bias vector.

[0081] After constructing multiple training samples, the simulation-based order reduction application can perform multiple rounds of iterative training on the neural network model based on these training samples until the preset training termination condition is met, thereby obtaining an accurate order reduction model. The specific process is as follows:

[0082] The neural network model is trained iteratively through multiple rounds using multiple training samples; among them...

[0083] During each training round, the model parameters in the neural network model are optimized and adjusted according to a preset gradient descent algorithm. This preset gradient descent algorithm can be, but is not limited to, SGD, Adam, etc. The model parameters in the neural network model are the weight matrix W and bias vector b in each hidden layer. θ new = θ old -η·▽ θ L ( θ ), θ new To optimize the adjusted model parameters, θold To optimize the model parameters before adjustment, η is the learning rate, ▽ θ L ( θ () represents the gradient of the loss function of the neural network model with respect to the model parameters;

[0084] After each round of training, determine whether the preset training termination condition has been met. The preset training termination condition may include, but is not limited to, minimizing the loss function, reaching a preset threshold for the number of iterations, reaching a preset threshold for the duration of iterations, etc.

[0085] If this is achieved, the neural network model obtained after this round of training will be determined as the accurate reduced-order model;

[0086] If the target is not met, the next round of training will be conducted based on the neural network model with optimized and adjusted model parameters.

[0087] 205. Obtain the current running data set corresponding to the target submodule.

[0088] Regarding step 205, obtaining the current running data set corresponding to the target submodule, please refer to the relevant description of step 105 above. This embodiment of the application will not repeat it here.

[0089] 206. Based on the current running data set corresponding to the target submodule, determine the dimensionality reduction simulation data corresponding to the target submodule.

[0090] Regarding step 206, determining the dimensionality reduction simulation data corresponding to the target sub-module based on the current running relevant data set corresponding to the target sub-module, please refer to the relevant description of step 106 above. This embodiment of the application will not repeat it here.

[0091] 207. Input the dimensionality reduction simulation data corresponding to the target submodule into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule.

[0092] Regarding step 207, which involves inputting the dimensionality reduction simulation data corresponding to the target submodule into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule, please refer to the relevant description of step 107 above. This embodiment of the application will not repeat it here.

[0093] 208. Perform fault warning processing on the target submodule.

[0094] After determining the 3D multiphysics data corresponding to the target submodule, the simulation reduction application can also determine the corresponding fault warning result for the target submodule based on preset rules and the corresponding 3D multiphysics data. This warning result is then displayed on the target terminal device to alert staff whether the target submodule is about to malfunction. This ensures that staff can promptly repair the target submodule when it malfunctions. The preset rules stipulate that when the 3D multiphysics data corresponding to the submodule is within the target range, the data is considered abnormal, and the submodule is in an abnormal operating state. Therefore, when the 3D multiphysics data is within the target range, the submodule is determined to be in an abnormal operating state, and the corresponding fault warning result indicates an impending fault. Conversely, when the 3D multiphysics data is outside the target range, the submodule is considered to be in normal working state, and the corresponding fault warning result indicates no fault will occur.

[0095] Furthermore, as a response to the above Figure 1 and Figure 2 In addition to the implementation of the method shown, another embodiment of this application also provides a multiphysics steady-state simulation order reduction device for a flexible DC converter valve submodule. This device embodiment corresponds to the aforementioned method embodiment. For ease of reading, this device embodiment will not repeat the details of the aforementioned method embodiments one by one, but it should be clear that the device in this embodiment can implement all the contents of the aforementioned method embodiments. This device is used to realize rapid and real-time simulation calculation of the three-dimensional multiphysics field of the submodule based on the operation-related data corresponding to the submodule in the flexible DC converter valve, thereby realizing real-time monitoring and analysis of the three-dimensional multiphysics field distribution of the submodule, specifically as follows... Figure 3 As shown, the device includes:

[0096] The first determining unit 31 is used to acquire multiple sets of operation-related data corresponding to each sub-module, and to determine the simulation data corresponding to each set of operation-related data according to the multiphysics coupling simulation model corresponding to each sub-module.

[0097] Data dimensionality reduction unit 32 is used to perform data dimensionality reduction processing on multiple simulation data according to a preset dimensionality reduction algorithm to obtain dimensionality reduction simulation data corresponding to each set of running related data;

[0098] Construction unit 33 is used to construct multiple training samples based on the dimensionality reduction simulation data and monitoring derivation mixed data corresponding to each of the operation-related data sets;

[0099] Training unit 34 is used to perform multiple rounds of iterative training on the neural network model based on multiple training samples until a preset training termination condition is reached, so as to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function; and to obtain the current running relevant data set corresponding to the target sub-module;

[0100] Acquisition unit 35 is used to acquire the current running data set corresponding to the target submodule;

[0101] The second determining unit 36 ​​is used to determine the dimensionality reduction simulation data corresponding to the target sub-module based on the current running relevant data set corresponding to the target sub-module;

[0102] The third determining unit 37 is used to input the dimensionality reduction simulation data corresponding to the target sub-module into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target sub-module.

[0103] Furthermore, such as Figure 4 As shown, the first determining unit 31 is specifically used to obtain the operating parameters corresponding to the sub-module; construct the multiphysics coupling simulation model corresponding to the sub-module based on the operating parameters corresponding to the sub-module; and control the multiphysics coupling simulation model corresponding to the sub-module to perform simulation operation based on the operating-related data set corresponding to the sub-module, so as to obtain the simulation data corresponding to the sub-module.

[0104] Furthermore, such as Figure 4 As shown, the data dimensionality reduction unit 32 is specifically used to construct a full-order simulation data matrix corresponding to the simulation data; perform centering processing on the full-order simulation data matrix to obtain a centralized data matrix; calculate the covariance matrix corresponding to the centralized data matrix; perform eigenvalue decomposition processing on the covariance matrix to obtain multiple eigenvalues ​​corresponding to the covariance matrix and eigenvectors corresponding to each eigenvalue; construct a low-dimensional data matrix based on the multiple eigenvalues ​​and eigenvectors corresponding to each eigenvalue; and determine the dimensionality-reduced simulation data corresponding to the running-related data set based on the low-dimensional data matrix.

[0105] Furthermore, such as Figure 4 As shown, the data dimensionality reduction unit 32 is specifically used to sort the multiple feature values ​​in descending order; construct a feature vector matrix based on the feature vectors corresponding to the top K feature values; and generate the low-dimensional data matrix based on the full-order simulation data matrix and the feature vector matrix.

[0106] Furthermore, such as Figure 4As shown, the construction unit 33 is specifically used to obtain the real monitoring data corresponding to the operation-related data set; and to determine the monitoring derivation mixed data corresponding to the operation-related data set based on the real monitoring data corresponding to the operation-related data set.

[0107] Furthermore, such as Figure 4 As shown, training unit 34 is specifically used to perform multiple rounds of iterative training on the neural network model based on multiple training samples. During each round of training, the model parameters in the neural network model are optimized and adjusted according to a preset gradient descent algorithm. The neural network model contains multiple hidden layers, and the model parameters in the neural network model are the weight matrix and bias vector in each hidden layer. After each round of training, it is determined whether the preset training termination condition has been met. If it has, the neural network model obtained after this round of training is determined as the accurate reduced-order model. If it has not been met, the next round of training is performed based on the neural network model with optimized and adjusted model parameters.

[0108] Furthermore, such as Figure 4 As shown, the device also includes:

[0109] The early warning unit 38 is used to determine the fault early warning result corresponding to the target sub-module based on preset rules and the three-dimensional multiphysics field data corresponding to the target sub-module; and output and display the fault early warning result corresponding to the target sub-module.

[0110] This application provides a method and apparatus for reducing the order of a flexible DC converter valve submodule through multiphysics steady-state simulation. This method enables the simulation reduction application to iteratively train a neural network model based on multiple training samples to obtain an accurate reduced-order model. When it is necessary to determine the three-dimensional multiphysics data corresponding to a target submodule in the target flexible DC converter valve, the simulation reduction application first obtains the current operating data set corresponding to the target submodule, then determines the reduced-order simulation data corresponding to the target submodule based on the current operating data set, and finally inputs the reduced-order simulation data corresponding to the target submodule into the accurate reduced-order model to determine the three-dimensional multiphysics data corresponding to the target submodule. In this embodiment, the dimensionality-reduced simulation data and the monitoring-derived mixed data are used as training samples to train the neural network model, thereby obtaining an accurate dimensionality-reduced model. Compared to iteratively training the neural network model using high-dimensional simulation data and monitoring-derived mixed data as training samples, using dimensionality-reduced simulation data and monitoring-derived mixed data as training samples effectively reduces the time and computational complexity consumed in training the neural network model, resulting in a faster output of the calculated results from the trained accurate dimensionality-reduced model. Furthermore, the dimensionality-reduced simulation data retains important information from the high-dimensional simulation data, and the monitoring-derived mixed data is more consistent with actual working conditions than the dimensionality-reduced simulation data. Therefore, the output results of the trained accurate dimensionality-reduced model are closer to the actual three-dimensional multiphysics field data of the submodule. Thus, in this embodiment, based on the operational data corresponding to the submodule in the flexible DC converter valve, rapid and real-time simulation calculations of the three-dimensional multiphysics field of the submodule can be achieved, thereby enabling real-time monitoring and analysis of the three-dimensional multiphysics field distribution of the submodule.

[0111] This application provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the above-described method for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule.

[0112] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule.

[0113] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for reducing the order of multiphysics steady-state simulation of a flexible DC converter valve submodule.

[0114] This application provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps:

[0115] Obtain multiple sets of operation-related data corresponding to each sub-module, and determine the simulation data corresponding to each set of operation-related data based on the multiphysics coupling simulation model corresponding to each sub-module;

[0116] The simulation data is subjected to dimensionality reduction processing according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of running-related data.

[0117] Multiple training samples are constructed based on the dimensionality reduction simulation data and monitoring derivation mixed data corresponding to each of the aforementioned operation-related data sets;

[0118] The neural network model is iteratively trained multiple times based on multiple training samples until a preset training termination condition is reached to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function.

[0119] Obtain the currently running data set corresponding to the target submodule;

[0120] Based on the current running data set corresponding to the target sub-module, determine the dimensionality reduction simulation data corresponding to the target sub-module;

[0121] The dimensionality reduction simulation data corresponding to the target submodule is input into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule.

[0122] Furthermore, determining the simulation data corresponding to each set of runtime-related data based on the multiphysics coupling simulation model corresponding to each sub-module includes:

[0123] Obtain the runtime parameters corresponding to the submodule;

[0124] Construct a multiphysics coupling simulation model for the submodule based on the operating parameters corresponding to the submodule;

[0125] The simulation operation is controlled based on the set of operation-related data corresponding to the submodule to obtain the simulation data corresponding to the submodule.

[0126] Furthermore, the step of performing data dimensionality reduction processing on multiple simulation data sets according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of runtime-related data includes:

[0127] Construct the full-order simulation data matrix corresponding to the simulation data;

[0128] The full-order simulation data matrix is ​​centered to obtain a centered data matrix;

[0129] Calculate the covariance matrix corresponding to the centralized data matrix;

[0130] The covariance matrix is ​​subjected to eigenvalue decomposition to obtain multiple eigenvalues ​​corresponding to the covariance matrix and an eigenvector corresponding to each eigenvalue.

[0131] A low-dimensional data matrix is ​​constructed based on the multiple eigenvalues ​​and the eigenvectors corresponding to each eigenvalue.

[0132] The dimensionality-reduced simulation data corresponding to the runtime-related data set is determined based on the low-dimensional data matrix.

[0133] Furthermore, the step of constructing a low-dimensional data matrix based on multiple eigenvalues ​​and the eigenvector corresponding to each eigenvalue includes:

[0134] Sort the multiple feature values ​​in descending order;

[0135] Construct an eigenvector matrix based on the eigenvectors corresponding to the top K eigenvalues ​​in the sorting;

[0136] The low-dimensional data matrix is ​​generated based on the full-order simulation data matrix and the eigenvector matrix.

[0137] Furthermore, before constructing multiple training samples based on the dimensionality-reduced simulation data and monitoring-derived mixed data corresponding to each of the aforementioned operation-related data sets, the method further includes:

[0138] Obtain the actual monitoring data corresponding to the aforementioned operation-related data set;

[0139] Based on the actual monitoring data corresponding to the operation-related data set, determine the monitoring-derived mixed data corresponding to the operation-related data set.

[0140] Furthermore, the step of performing multiple rounds of iterative training on the neural network model based on multiple training samples until a preset training termination condition is met to obtain an accurate reduced-order model includes:

[0141] The neural network model is trained iteratively through multiple rounds based on the training samples mentioned above; wherein...

[0142] During each training round, the model parameters in the neural network model are optimized and adjusted according to a preset gradient descent algorithm. The neural network model contains multiple hidden layers, and the model parameters in the neural network model are the weight matrix and bias vector in each hidden layer.

[0143] After each round of training, determine whether the preset training termination condition has been met;

[0144] If this is achieved, the neural network model obtained after this round of training will be determined as the accurate reduced-order model;

[0145] If the target is not met, the next round of training will be conducted based on the neural network model after optimization and adjustment of the model parameters.

[0146] Furthermore, the method also includes:

[0147] Based on the preset rules and the three-dimensional multiphysics field data corresponding to the target sub-module, the fault warning result corresponding to the target sub-module is determined;

[0148] The output displays the fault warning results corresponding to the target submodule.

[0149] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing program code with the following initialization steps: acquiring multiple sets of runtime-related data corresponding to each sub-module, and determining simulation data corresponding to each set of runtime-related data based on the multiphysics coupling simulation model corresponding to each sub-module; performing dimensionality reduction processing on the multiple sets of simulation data according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of runtime-related data; constructing multiple training samples based on the dimensionality-reduced simulation data and monitoring-derived mixed data corresponding to each set of runtime-related data; performing multiple rounds of iterative training on a neural network model based on the multiple training samples until a preset training termination condition is reached to obtain an accurate dimensionality reduction model, wherein the preset training termination condition is minimizing a loss function; acquiring the current set of runtime-related data corresponding to the target sub-module; determining the dimensionality-reduced simulation data corresponding to the target sub-module based on the current set of runtime-related data corresponding to the target sub-module; inputting the dimensionality-reduced simulation data corresponding to the target sub-module into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target sub-module.

[0150] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0154] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0155] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0156] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0157] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0158] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for order reduction in multiphysics steady-state simulation of a flexible DC converter valve submodule, characterized in that, The method includes: Multiple operation-related data sets corresponding to each sub-module are obtained, and the simulation data corresponding to each operation-related data set is determined according to the multiphysics coupling simulation model corresponding to each sub-module. For any operation-related data set, the simulation data corresponding to the operation-related data set includes the simulated three-dimensional temperature field data, simulated three-dimensional magnetic field data and simulated three-dimensional fluid field data of the sub-module corresponding to the operation-related data set. The simulation data is subjected to dimensionality reduction processing according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of running-related data. Multiple training samples are constructed based on the dimensionality reduction simulation data and monitoring-derived hybrid data corresponding to each of the operation-related data sets. For any operation-related data set, the monitoring-derived hybrid data corresponding to the operation-related data set includes monitoring-derived hybrid three-dimensional temperature field data, monitoring-derived hybrid three-dimensional magnetic field data, and monitoring-derived hybrid three-dimensional fluid field data, which are composed of real data at each first-type target location and derivation data at each second-type target location in the sub-module corresponding to the operation-related data set. For any first-type target location, the physical field data at the first-type target location can be directly measured. For any second-type target location, the physical field data at the second-type target location cannot be directly measured. The derivation data at each second-type target location is determined based on the spatial positional relationship between multiple second-type target locations and multiple first-type target locations and the real data at each first-type target location. The neural network model is iteratively trained multiple times based on multiple training samples until a preset training termination condition is reached to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function. Obtain the currently running data set corresponding to the target submodule; Based on the current running data set corresponding to the target sub-module, determine the dimensionality reduction simulation data corresponding to the target sub-module; The dimensionality reduction simulation data corresponding to the target submodule is input into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target submodule.

2. The method according to claim 1, characterized in that, The step of determining the simulation data corresponding to each set of runtime-related data based on the multiphysics coupling simulation model corresponding to each sub-module includes: Obtain the runtime parameters corresponding to the submodule; Construct a multiphysics coupling simulation model for the submodule based on the operating parameters corresponding to the submodule; The simulation operation is controlled based on the set of operation-related data corresponding to the submodule to obtain the simulation data corresponding to the submodule.

3. The method according to claim 1, characterized in that, The step of performing data dimensionality reduction processing on multiple simulation data according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of runtime-related data includes: Construct the full-order simulation data matrix corresponding to the simulation data; The full-order simulation data matrix is ​​centered to obtain a centered data matrix; Calculate the covariance matrix corresponding to the centralized data matrix; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain multiple eigenvalues ​​corresponding to the covariance matrix and an eigenvector corresponding to each eigenvalue. A low-dimensional data matrix is ​​constructed based on the multiple eigenvalues ​​and the eigenvectors corresponding to each eigenvalue. The dimensionality-reduced simulation data corresponding to the runtime-related data set is determined based on the low-dimensional data matrix.

4. The method according to claim 3, characterized in that, The step of constructing a low-dimensional data matrix based on multiple eigenvalues ​​and the eigenvector corresponding to each eigenvalue includes: Sort the multiple feature values ​​in descending order; Construct an eigenvector matrix based on the eigenvectors corresponding to the top K eigenvalues ​​in the sorting; The low-dimensional data matrix is ​​generated based on the full-order simulation data matrix and the eigenvector matrix.

5. The method according to claim 1, characterized in that, Before constructing multiple training samples based on the dimensionality-reduced simulation data and monitoring-derived mixed data corresponding to each of the aforementioned operation-related datasets, the method further includes: Obtain the actual monitoring data corresponding to the aforementioned operation-related data set; Based on the actual monitoring data corresponding to the operation-related data set, determine the monitoring-derived mixed data corresponding to the operation-related data set.

6. The method according to claim 1, characterized in that, The step of performing multiple rounds of iterative training on the neural network model based on multiple training samples until a preset training termination condition is met to obtain an accurate reduced-order model includes: The neural network model is trained iteratively through multiple rounds based on the training samples mentioned above; wherein... During each training round, the model parameters in the neural network model are optimized and adjusted according to a preset gradient descent algorithm. The neural network model contains multiple hidden layers, and the model parameters in the neural network model are the weight matrix and bias vector in each hidden layer. After each round of training, determine whether the preset training termination condition has been met; If this is achieved, the neural network model obtained after this round of training will be determined as the accurate reduced-order model; If the target is not met, the next round of training will be conducted based on the neural network model after optimization and adjustment of the model parameters.

7. The method according to claim 1, characterized in that, The method further includes: Based on the preset rules and the three-dimensional multiphysics field data corresponding to the target sub-module, the fault warning result corresponding to the target sub-module is determined; The output displays the fault warning results corresponding to the target submodule.

8. A multiphysics steady-state simulation order reduction device for a flexible DC converter valve submodule, characterized in that, The device includes: The first determining unit is used to acquire multiple operation-related data sets corresponding to each sub-module, and determine the simulation data corresponding to each operation-related data set according to the multiphysics coupling simulation model corresponding to each sub-module. For any operation-related data set, the simulation data corresponding to the operation-related data set includes the simulation three-dimensional temperature field data, simulation three-dimensional magnetic field data and simulation three-dimensional fluid field data of the sub-module corresponding to the operation-related data set. The data dimensionality reduction unit is used to perform data dimensionality reduction processing on multiple simulation data according to a preset dimensionality reduction algorithm to obtain dimensionality-reduced simulation data corresponding to each set of running-related data. The construction unit is used to construct multiple training samples based on the dimensionality reduction simulation data and monitoring-derived mixed data corresponding to each of the operation-related data sets. For any operation-related data set, the monitoring-derived mixed data corresponding to the operation-related data set includes monitoring-derived mixed three-dimensional temperature field data, monitoring-derived mixed three-dimensional magnetic field data, and monitoring-derived mixed three-dimensional fluid field data, which are composed of real data at each first-type target location and derivation data at each second-type target location in the sub-module corresponding to the operation-related data set. For any first-type target location, the physical field data at the first-type target location can be directly measured. For any second-type target location, the physical field data at the second-type target location cannot be directly measured. The derivation data at each second-type target location is determined based on the spatial positional relationship between multiple second-type target locations and multiple first-type target locations and the real data at each first-type target location. The training unit is used to perform multiple rounds of iterative training on the neural network model based on multiple training samples until a preset training termination condition is reached in order to obtain an accurate reduced-order model, wherein the preset training termination condition is to minimize the loss function. The acquisition unit is used to acquire the current running-related data set corresponding to the target submodule; The second determining unit is used to determine the dimensionality reduction simulation data corresponding to the target sub-module based on the current running relevant data set corresponding to the target sub-module; The third determining unit is used to input the dimensionality reduction simulation data corresponding to the target sub-module into the accurate dimensionality reduction model to determine the three-dimensional multiphysics data corresponding to the target sub-module.

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

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

Images