Train high-voltage system model optimization method, system and equipment and storage medium
By building a multi-dimensional physics digital twin model, combining neural networks and artificial bee colony algorithm to optimize parameters, the problem of inefficient computing efficiency of the high-voltage system model of EMU is solved, and efficient online real-time monitoring and fault prediction are achieved.
Patent Information
- Application Number
- CN202510192135.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-20
- Publication Date
- 2025-06-27
AI Technical Summary
The traditional high-voltage system model of EMU is ineffective in processing multi-physics coupling and inflexible parameter adjustments, which leads to the inability to fully capture the dynamic changes of the system and the multi-physics coupling effect, limiting the improvement of the digital twin model in fault analysis and prediction functions.
By building a multi-dimensional physics digital twin model, combining historical monitoring data and simulation monitoring data to generate sample data sets, using neural network models for data fusion and optimization, adopting adaptive dimensionality reduction, causal convolution and expanded convolution to capture timing features, and combining artificial bee colony algorithm to adjust physical parameters to achieve multi-physics simulation configuration optimization.
The calculation efficiency and accuracy of the digital twin model of the high-voltage system of the EMU is improved, real-time online monitoring and fault prediction are realized, and the authenticity and accuracy of fault detection are improved.
Smart Images
Figure CN120217827A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of rail transit, and specifically, to a method, system, device, and storage medium for optimizing a train high-voltage system model. Background Art
[0002] Due to the high operating speed, long driving mileage, high line density, and complex operating environment of high-speed trains, it is necessary to conduct pre-inspections and real-time monitoring of various equipment components of the train. The high-voltage system of the multiple unit train is responsible for providing necessary power for the multiple unit train, and the stability of its performance is directly related to the safe and stable operation of the entire train.
[0003] When the traditional high-voltage system model of the multiple unit train processes the multi-physical field coupling of the high-voltage system, there are often problems such as low calculation efficiency and inflexible parameter adjustment, and it is difficult to achieve real-time effects, resulting in the inability to comprehensively capture the dynamic changes and multi-physical field coupling effects of the system; at the same time, it also limits the improvement of functions such as equipment safety assessment, fault analysis, and prediction of the high-voltage system of the multiple unit train based on the digital twin model.
[0004] Therefore, improving the calculation efficiency and accuracy of the digital twin model of the high-voltage system of the multiple unit train is the key to realizing the online real-time monitoring function. Summary of the Invention
[0005] To solve one of the above technical defects, the embodiments of the present application provide a method, device, and storage medium for optimizing a train high-voltage system model.
[0006] According to the first aspect of the embodiments of the present application, a method for optimizing a train high-voltage system model is provided. The method includes:
[0007] Generating simulation monitoring data of the train high-voltage system according to the multi-physical field digital twin model, and mixing historical monitoring data and simulation monitoring data to construct a sample data set; the multi-dimensional physical field digital twin model is constructed according to the historical monitoring data of the train high-voltage system;
[0008] Establishing a neural network model for each physical field dimension according to the sample data set, and fusing each neural network model to obtain a fused neural network model;
[0009] Setting physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fused neural network model to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model.
[0010] In an optional embodiment of the present application, the step of constructing a multi-dimensional physical field digital twin model according to the historical monitoring data of the train high-voltage system, generating simulation monitoring data of the train high-voltage system according to the multi-physical field digital twin model, and mixing historical monitoring data and simulation monitoring data to construct a sample data set further includes:
[0011] Supplement the missing part in the historical monitoring data with the simulation monitoring data.
[0012] In an optional embodiment of the present application, the steps of separately establishing a neural network model for each physical field dimension according to the sample data set and fusing each neural network model to obtain a fused neural network model further include:
[0013] Perform non-linear dimensionality reduction and preprocessing on the sample data set through an adaptive dimensionality reduction method to divide the sample data set into a training set and a test set, divide the training set into subsets for each physical field dimension, and establish a neural network model according to each subset respectively.
[0014] In an optional embodiment of the present application, the steps of separately establishing a neural network model for each physical field dimension according to the sample data set and fusing each neural network model to obtain a fused neural network model further include:
[0015] Capture the temporal features in the sample data set through causal convolution and dilated convolution to establish the mapping relationship between the sequence data of each physical field dimension and the fault state.
[0016] In an optional embodiment of the present application, the steps of separately establishing a neural network model for each physical field dimension according to the sample data set and fusing each neural network model to obtain a fused neural network model further include:
[0017] Arrange each neural network model in parallel and integrate each neural network model through a random forest network to obtain a fused neural network model.
[0018] In an optional embodiment of the present application, the steps of separately establishing a neural network model for each physical field dimension according to the sample data set and fusing each neural network model to obtain a fused neural network model further include:
[0019] Integrate each neural network model through a random forest network and obtain a fused neural network model by any one of the weighted average and voting mechanisms.
[0020] In an optional embodiment of the present application, the steps of setting the physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fused neural network model to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model further include:
[0021] Use the artificial bee colony algorithm to perform reverse optimization on the prediction calculation results of the fused neural network model to set the physical parameters of the multi-dimensional physical field digital twin model.
[0022] According to the second aspect of the embodiments of the present application, a train high-voltage system model optimization system is provided. The system includes a sample mixing module, a model fusion module, and a model optimization module. Among them,
[0023] The sample mixing module is configured to construct a multi-dimensional physical field digital twin model based on the historical monitoring data of the train high-voltage system, generate simulation monitoring data of the train high-voltage system according to the multi-physical field digital twin model, and mix the historical monitoring data and the simulation monitoring data to construct a sample data set;
[0024] The model fusion module is configured to establish a neural network model for each physical field dimension according to the sample data set, and fuse each neural network model to obtain a fused neural network model;
[0025] The model optimization module is configured to set the physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fused neural network model, so as to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model.
[0026] According to the third aspect of the embodiments of the present application, a computer device is provided, including: a memory; a processor; and a computer program. Among them, the computer program is stored in the memory and is configured to be executed by the processor to implement the train high-voltage system model optimization method according to any one of the first aspects of the embodiments of the present application.
[0027] According to the fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored. The computer program is executed by a processor to implement the train high-voltage system model optimization method according to any one of the first aspects of the embodiments of the present application.
[0028] Adopting the train high-voltage system model optimization method provided in the embodiments of the present application has the following beneficial effects:
[0029] The present application obtains the monitoring data of the high-voltage system of the EMU under real historical working conditions, constructs multi-physical field coupling simulation data based on the multi-physical field digital twin model, and constructs a mixed sample data set by combining the historical monitoring data and simulation data under real working conditions, which can make up for the problem of unbalanced historical data samples, and can further truly reflect the changes of various working conditions and complex environments of the high-voltage system of the EMU, thereby improving the authenticity of fault detection of the digital twin model. In addition, the present application proposes to use a fused neural network model for prediction and complete the fast calculation of on-line detection of the train high-voltage system, which can real-time detect potential fault signs in the data and improve the calculation efficiency and prediction accuracy. Description of the Drawings
[0030] The accompanying drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:
[0031] Figure 1 is a flowchart of the method for optimizing the train high-voltage system model provided by an embodiment of the present application;
[0032] Figure 2 is a structural diagram of the train high-voltage system model optimization system provided by an embodiment of the present application;
[0033] Figure 3 is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners
[0034] In order to make the technical solutions and advantages in the embodiments of the present application clearer and more understandable, the following further details the exemplary embodiments of the present application with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than an exhaustive list of all embodiments. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.
[0035] As of the end of 2023, the total operating mileage of high-speed railways in China reached 35,000 kilometers, accounting for more than two-thirds of the world's total operating mileage. Due to the high operating speed, long driving mileage, high line density, and complex operating environment of high-speed trains, it is necessary to conduct pre-inspections and real-time monitoring of various equipment components of the train. The high-voltage system of the EMU is responsible for providing necessary power for the EMU, and the stability of its performance is directly related to the safe and stable operation of the entire train.
[0036] With the development of society, higher requirements are put forward for the stable operation of the high-voltage system of the EMU. When dealing with the multi-physical field coupling of the high-voltage system, the traditional high-voltage system model of the EMU often has problems such as low calculation efficiency and inflexible parameter adjustment, which bring great difficulties to the realization of real-time effects, resulting in the inability to comprehensively capture the dynamic changes and multi-physical field coupling effects of the system. At the same time, it also restricts the improvement of functions such as equipment safety assessment, fault analysis, and prediction of the high-voltage system of the EMU based on the digital twin model. Therefore, improving the calculation efficiency and accuracy of the digital twin model of the high-voltage system of the EMU is the key to realizing the online real-time monitoring function. The purpose of the present application is to provide a data-driven optimization method for the multi-physical field digital twin model of the high-voltage system of the EMU to improve the above problems.
[0037] As Figure 1 shown, the present application proposes a method for optimizing the train high-voltage system model, Figure 1 is a flowchart of the method for optimizing the train high-voltage system model provided by an embodiment of the present application. Please refer toFigure 1 :
[0038] S1: Generate simulation monitoring data for the train high-voltage system according to the multi-physical-field digital twin model, and construct a sample data set by mixing historical monitoring data and simulation monitoring data; the multi-dimensional physical-field digital twin model is constructed according to the historical monitoring data of the train high-voltage system.
[0039] In some embodiments of the present application, the missing part in the historical monitoring data is supplemented by simulation monitoring data.
[0040] In some embodiments of the present application, the dimensions of the multi-dimensional physical fields include electric field, thermal field, flow field and climate dimension.
[0041] In specific implementation, construct multi-physical-field coupled simulation data based on the multi-physical-field digital twin model, and construct a mixed sample data set by combining historical monitoring data and multi-physical-field coupled simulation data. Specifically, the multi-dimensional physical-field digital twin model is constructed according to the historical monitoring data of the train high-voltage system under real historical working conditions.
[0042] Specifically, the monitoring data under historical working conditions includes electric field, thermal field, flow field and climate parameters.
[0043] Specifically, in the embodiments of the present application, multi-physical-field coupling is used to establish simulation data with multiple dimensions coupled, and further use the simulation data to make up for the data with unbalanced dimensions in historical monitoring. Optionally, the severely missing part in the real data is supplemented with simulation data; construct a mixed sample data set by combining historical monitoring data and multi-physical-field coupled simulation data. Based on this, the mixed sample data set constructed by combining real data and simulation data in the embodiments of the present application makes up for the problem of unbalanced historical data samples, and can truly reflect the changes of various working conditions and complex environments of the high-voltage system of the EMU, thereby improving the authenticity of fault detection of the digital twin model.
[0044] S2: Establish a neural network model for each physical field dimension according to the sample data set, and fuse each neural network model to obtain a fused neural network model.
[0045] Specifically, use the adaptive multi-layer autoencoder dimensionality reduction method to perform non-linear dimensionality reduction on the mixed sample data set to obtain a reduced-dimensional feature input matrix, preprocess the feature input matrix, and establish a target neural network model under different-dimensional features according to the temporal convolutional neural network.
[0046] In some embodiments of the present application, perform non-linear dimensionality reduction and preprocessing on the sample data set through an adaptive dimensionality reduction method to divide the sample data set into a training set and a test set, divide the training set into subsets for each physical field dimension, and establish a neural network model for each subset respectively.
[0047] In a specific implementation, an adaptive dimensionality reduction method is used to perform non-linear dimensionality reduction on the mixed sample data set. The mixed sample data set is mapped to a low-dimensional space through the encoder of the adaptive multi-layer autoencoder to obtain the minimum dimensional representation of the sample data set, and the decoder maps the low-dimensional representation back to the original dimensional structure. The expression of the adaptive multi-layer autoencoder dimensionality reduction method is:
[0048] H = f i (x; w i , b i )
[0049] In the formula, H is the data after dimensionality reduction by the multi-layer autoencoder, f i is the activation function of the i-th layer of the multi-layer autoencoder, x is the mixed sample data of the multi-layer autoencoder, w i is the weight of the i-th hidden layer of the multi-layer autoencoder, and b i is the bias of the i-th hidden layer of the multi-layer autoencoder.
[0050] In this embodiment, the encoder of the adaptive multi-layer autoencoder is composed of hidden layers with gradually decreasing neurons, and the decoder is composed of hidden layers with gradually increasing neurons corresponding to the encoder. The expression of the loss function of the adaptive multi-layer autoencoder is:
[0051]
[0052] In the formula, represents the i-th component of the reconstructed data vector, x i represents the i-th component of the data vector, N represents the dimension of the data, and Loss is the mean square error between the real data and the output data.
[0053] Update the weights and biases of the model according to the mean square error of the loss function, and use the Adam (Adaptive Moment Estimation) algorithm to train the adaptive multi-layer autoencoder, and iterate until the loss function converges to output the final result to achieve data dimensionality reduction.
[0054] Furthermore, preprocess the dimensionality-reduced feature input matrix, where the preprocessing includes data cleaning, standardization, and feature selection, and divide the preprocessed data set into a training set and a test set. Preferably, in this embodiment, the first 80% is divided into the training set and the last 20% is divided into the test set.
[0055] Divide the training set after data analysis into multiple subsets according to different dimensions, including: an electric field subset module, a thermal field subset module, a flow field subset module, and a climate subset module, and use the data of different modules after preprocessing to train different temporal convolutional networks.
[0056] In some embodiments of the present application, causal convolution and dilated convolution are used to capture the temporal features in the sample dataset, so as to establish the mapping relationship between the sequence data of each physical field dimension and the fault state.
[0057] In a specific implementation, a temporal convolutional network model is established, including causal convolution, dilated convolution, and residual connection. Causal convolution is used for sequence modeling, setting a strict time constraint structure, and long temporal dependence information in the sequence data is learned through dilated convolution to avoid the loss of historical data information. Further, the stability of the network is strengthened through residual connection, and the predicted value is output. The forward propagation training of the temporal convolutional network is realized by comparing the real value with the mean square error of the loss function.
[0058] The formula for causal convolution is:
[0059] Y t =Conv(X t ,W)+b
[0060] In the formula, X t is the preprocessed input data, Conv represents the convolution operation, W is the convolution kernel, b is the bias term, and Y t represents the sequence data after causal convolution.
[0061] The formula for dilated convolution is:
[0062]
[0063] In the formula, d is the dilation factor and k is the convolution kernel size.
[0064] The network weights are updated by the backpropagation algorithm of the least mean square error criterion, and the optimal solution is obtained through iteration to obtain the neural network model under different dimensional features. Each temporal convolutional network subset is verified by the test set samples. Preferably, in the embodiments of the present application, it is judged whether the verification result is less than 2%. If it is less than, it is output; if it is greater than, iterative calculation is performed until the verification result is less than 2% and then output.
[0065] In some embodiments of the present application, each neural network model is arranged in parallel, and each neural network model is integrated by a random forest network to obtain a fused neural network model.
[0066] In some embodiments of the present application, each neural network model is integrated by a random forest network, and a fused neural network model is obtained by any one of the weighted average and voting mechanism.
[0067] In specific implementation, a temporal convolutional network for the electric field module, a temporal convolutional network for the thermal field module, a temporal convolutional network for the flow field module, and a temporal convolutional network for the climate module are created. Each temporal convolutional network extracts features of different dimensions and conducts specific analysis on the features of different dimensions.
[0068] Regarding each temporal convolutional network as a base learner, all temporal convolutional networks are fused into a code framework through a random forest network. In the embodiments of the present application, preferably, the number of training rounds of the random forest is set to 100. Each pivot of the random forest is connected to a temporal convolutional network, and a weighted average or voting mechanism is used to operate and integrate a strong learner, thereby obtaining a fused neural network.
[0069] Based on this, in the embodiments of the present application, a fused neural network model for the high-voltage system of the multiple-unit train for big data is proposed based on the parallel processing ability of the temporal convolutional network and the high accuracy of the random forest network, which can quickly process multi-dimensional time-series data, can discover potential fault signs in the data in real time, and improve the calculation efficiency and prediction accuracy.
[0070] S3: Set the physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fused neural network model, so as to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model.
[0071] In some embodiments of the present application, an artificial bee colony algorithm is used to perform reverse optimization on the prediction calculation results of the fused neural network model to set the physical parameters of the multi-dimensional physical field digital twin model.
[0072] In specific implementation, according to the calculation results of the integrated neural network model, combined with the artificial bee colony algorithm, the physical parameters of the first multi-physical field simulation model are adjusted to realize the optimization calculation of the accuracy and efficiency of the simulation model.
[0073] Specifically, analyze the calculation results of the fused neural network, and use the artificial bee colony algorithm to perform reverse optimization on the calculation results of the fused neural network. Initialize the parameters and the positions of each bee, and calculate the fitness value for the solution generated by each bee to evaluate the quality of the solution. The formula for the new solution v ij is as follows:
[0074] v ij = x ij + φ ij ·(x ij - x kj )
[0075] In the formula, v ij is the new solution, φ ij is a number randomly generated within the interval [-1, 1] and is used to control the amplitude of the perturbation. x ij is the solution of the current bee, xkj is the solution of another randomly selected bee.
[0076] Worker bees use the current solution to locally search for the fitness of a new solution. The formula for calculating its fitness is:
[0077]
[0078] In the formula, F is the fitness, f(x) is the value of the objective function, and ε is a small constant.
[0079] Compare with the fitness of the current solution: If the fitness of the new solution is higher, accept the new solution and update the current solution and fitness. If the fitness of the new solution is not high, retain the current solution. At the end of each iteration, information is exchanged among all bees. After all iterations are completed, the solution with the highest fitness in the current iteration is output as the optimal solution, which is the optimal multi - physical - field simulation configuration of the current high - voltage system digital twin model of the EMU.
[0080] Finally, according to the optimal multi - physical - field simulation configuration searched by the artificial bee colony algorithm, verify the parameters of the multi - dimensional physical - field digital twin model, adjust the electric field, thermal field, flow field, and climate parameters in real - time, and dynamically adapt to the changing physical - field conditions, so as to improve the calculation efficiency and monitoring accuracy of the high - voltage system digital twin model of the EMU.
[0081] Based on this, in the embodiments of the present application, the artificial bee colony algorithm is combined to dynamically adjust the changing physical - field conditions, and more accurately capture and optimize the multi - physical - field coupling effect.
[0082] It should be understood that although the steps in the flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the figure may include multiple sub - steps or multiple stages. These sub - steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub - steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub - steps or stages of other steps.
[0083] Please refer to Figure 2 , an embodiment of the present application provides a train high - voltage system model optimization system, which includes a sample mixing module 10, a model fusion module 20, and a model optimization module 30; among them,
[0084] A sample mixing module 10 is configured to construct a multi - dimensional physical field digital twin model based on the historical monitoring data of the train high - voltage system, generate simulation monitoring data of the train high - voltage system according to the multi - physical field digital twin model, and mix the historical monitoring data and the simulation monitoring data to construct a sample data set;
[0085] A model fusion module 20 is configured to establish a neural network model for each physical field dimension according to the sample data set, and fuse each neural network model to obtain a fused neural network model;
[0086] A model optimization module 30 is configured to set the physical parameters of the multi - dimensional physical field digital twin model according to the prediction calculation results of the fused neural network model, so as to optimize the multi - physical field simulation configuration of the multi - dimensional physical field digital twin model.
[0087] For the specific limitations of the above - mentioned train high - voltage system model optimization system, reference can be made to the limitations of the train high - voltage system model optimization method in the above text, which will not be elaborated here. Each module in the above - mentioned train high - voltage system model optimization system can be implemented in whole or in part by software, hardware, and their combinations. The above - mentioned modules can be embedded in the processor of the computer device in hardware form or be independent of it, or be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above - mentioned modules.
[0088] In one embodiment, a computer device is provided. The internal structure diagram of the computer device can be as Figure 3 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non - volatile storage medium and an internal memory. The non - volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non - volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements the above - mentioned train high - voltage system model optimization method. It includes: a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the above - mentioned train high - voltage system model optimization method.
[0089] In one embodiment, a computer - readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it can implement the above - mentioned train high - voltage system model optimization method.
[0090] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. For example, C language, VHDL language, Verilog language, object-oriented programming language Java, and interpreted scripting language JavaScript, etc.
[0091] The present application is described with reference to the flowcharts and / or block diagrams of systems and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.
[0092] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.
[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.
[0094] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0095] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, this application is also intended to cover these changes and modifications.
Claims
1. A train high voltage system model optimization method, characterized in that: include: Generate simulation monitoring data of the train high-voltage system according to the multi-physics field digital twin model, and mix the historical monitoring data and the simulation monitoring data to construct a sample data set; The multi-dimensional physical field digital twin model is constructed based on historical monitoring data of the train high-voltage system; Establishing a neural network model for each physical field dimension according to the sample data set, and fusing each of the neural network models to obtain a fused neural network model; The physical parameters of the multi-dimensional physical field digital twin model are set according to the prediction calculation results of the fusion neural network model to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model.
2. The train high voltage system model optimization method according to claim 1, characterized in that: The steps of constructing a multi-dimensional physical field digital twin model according to the historical monitoring data of the train high-voltage system, generating simulation monitoring data of the train high-voltage system according to the multi-physical field digital twin model, and mixing the historical monitoring data and the simulation monitoring data to construct a sample data set also include: The missing parts in the historical monitoring data are supplemented by the simulated monitoring data.
3. The train high voltage system model optimization method according to claim 1, characterized in that: The step of establishing a neural network model for each physical field dimension according to the sample data set and fusing each of the neural network models to obtain a fused neural network model also includes: The sample data set is subjected to nonlinear dimensionality reduction and preprocessing by an adaptive dimensionality reduction method, so as to divide the sample data set into a training set and a test set, the training set is divided into subsets for each physical field dimension, and the neural network model is established according to each subset.
4. The train high voltage system model optimization method according to claim 3, characterized in that: The step of establishing a neural network model for each physical field dimension according to the sample data set and fusing each of the neural network models to obtain a fused neural network model also includes: The temporal features in the sample data set are captured by causal convolution and dilated convolution to establish a mapping relationship between each physical field dimension sequence data and the fault state.
5. The train high voltage system model optimization method according to any one of claims 1 to 4, characterized in that: The step of establishing a neural network model for each physical field dimension according to the sample data set and fusing each of the neural network models to obtain a fused neural network model also includes: Each of the neural network models is arranged in parallel, and each of the neural network models is integrated through a random forest network to obtain the fused neural network model.
6. The train high voltage system model optimization method according to claim 5, characterized in that: The step of establishing a neural network model for each physical field dimension according to the sample data set and fusing each of the neural network models to obtain a fused neural network model also includes: Each of the neural network models is integrated through a random forest network, and the fused neural network model is obtained through any one of a weighted average and a voting mechanism.
7. The train high voltage system model optimization method according to claim 6, characterized in that: The step of setting the physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fusion neural network model to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model also includes: The artificial bee colony algorithm is used to reversely optimize the prediction calculation results of the fusion neural network model to set the physical parameters of the multi-dimensional physical field digital twin model.
8. A train high voltage system model optimization system, characterized in that: include: Sample mixing module, model fusion module and model optimization module; among them, A sample mixing module, used to generate simulation monitoring data of the train high-voltage system according to the multi-physics field digital twin model, and mix the historical monitoring data and the simulation monitoring data to construct a sample data set; The multi-dimensional physical field digital twin model is constructed based on historical monitoring data of the train high-voltage system; A model fusion module, used to establish a neural network model for each physical field dimension according to the sample data set, and fuse each of the neural network models to obtain a fused neural network model; The model optimization module is used to set the physical parameters of the multi-dimensional physical field digital twin model according to the prediction calculation results of the fusion neural network model, so as to optimize the multi-physical field simulation configuration of the multi-dimensional physical field digital twin model.
9. A computer device, characterized in that: include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the train high voltage system model optimization method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: A computer program is stored thereon; the computer program is executed by a processor to implement a train high-voltage system model optimization method as claimed in any one of claims 1 to 7.