Model training method and device, pantograph-catenary system simulation method and device, and medium

Abstract

The invention provides a model training method and device, a pantograph-catenary system simulation method and device, and a medium, and relates to the technical field of pantograph-catenary simulation, and the method comprises the steps: obtaining a training sample set, constructing a target loss function comprising a physical loss function, constraining the output of a pantograph-catenary simulation model to accord with a physical rule through the physical loss function, and improving the simulation performance of the pantograph-catenary system. And finally, training the pantograph-catenary simulation model according to the training sample set, and obtaining the trained pantograph-catenary simulation model by minimizing the target loss function. By adding the physical loss function into the target loss function, the physical rule constraint output by the pantograph-catenary simulation model can be realized, and the effect of improving the prediction accuracy of the pantograph-catenary simulation model can be achieved.

Description

Technical Field

[0001] This application relates to the technical field of pantograph-catenary simulation, and particularly to a model training method, device, pantograph-catenary system simulation method, device, and medium. Background Art

[0002] In electrified railways, the interaction performance between the pantograph and the catenary is crucial for maintaining a stable current supply. Currently, the finite element method is commonly used to model and evaluate the dynamic performance of the pantograph-catenary system. However, this method has a high modeling difficulty and complex and large computational requirements. Summary of the Invention

[0003] This application provides a model training method, device, pantograph-catenary system simulation method, device, and medium, which can reduce the computational difficulty in the simulation process and improve the accuracy of dynamic simulation.

[0004] A model training method provided by this application includes: Obtain a training sample set; the training sample set includes multiple training samples, and each training sample includes input features and a sample label. The input features include initial simulation parameters and the pantograph-catenary simulation results at multiple consecutive time steps after performing a finite element simulation on the pantograph-catenary system. The sample label includes the pantograph-catenary simulation result at the next time step after multiple consecutive time steps; Construct an objective loss function for the pantograph-catenary simulation model; the objective loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws; Train the pantograph-catenary simulation model according to the training sample set, and obtain the trained pantograph-catenary simulation model by minimizing the objective loss function Optionally, the pantograph-catenary simulation results include a state matrix, an external force matrix, a mass matrix, a damping matrix, and a stiffness matrix, and the physical loss function includes: ; Wherein, is the loss value of the physical loss function, is the total number of training samples, represents the th training sample, is the physical loss term, , represents the state matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the external force matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the mass matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, denotes the damping matrix obtained according to the input features of the pantograph-catenary simulation model and the th training sample, denotes the stiffness matrix obtained according to the input features of the pantograph-catenary simulation model and the th training sample.

[0005] Optionally, constructing the objective loss function of the pantograph-catenary simulation model includes: Determining the weighted sum of the basic loss function and the physical loss function as the objective loss function.

[0006] Optionally, constructing the objective loss function of the pantograph-catenary simulation model includes: Determining the sum of the basic loss function and the physical loss function as the objective loss function.

[0007] Optionally, the basic loss function is the mean square error loss function.

[0008] Optionally, the pantograph-catenary simulation model includes a time convolutional network, a long short-term memory network, a multi-head attention mechanism, and a fully connected network connected in sequence.

[0009] A pantograph-catenary system simulation method provided by the present application includes: Obtaining the current simulation parameters of the pantograph-catenary system; Determining the current pantograph-catenary simulation result according to the current simulation parameters and the pantograph-catenary simulation model; wherein, the pantograph-catenary simulation model is trained based on one or more of the foregoing methods.

[0010] To achieve the above object and other related objects, the present application provides a model training device, including: A training data acquisition module, configured to acquire a training sample set; the training sample set includes a plurality of training samples, and the training samples include initial simulation parameters and the pantograph-catenary simulation results obtained by performing finite element simulation on the pantograph-catenary system according to the initial simulation parameters; A loss function construction module, configured to construct the objective loss function of the pantograph-catenary simulation model; the objective loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws; A model training module, configured to train the pantograph-catenary simulation model according to the training sample set, and obtain the trained pantograph-catenary simulation model by minimizing the objective loss function.

[0011] To achieve the above object and other related objects, the present application provides a pantograph-catenary system simulation device, including: A current data acquisition module, configured to acquire the current simulation parameters of the pantograph-catenary system; A result determination module, configured to determine the current pantograph-catenary simulation result according to the current simulation parameters and the pantograph-catenary simulation model; wherein, the pantograph-catenary simulation model is trained based on one or more of the foregoing methods.

[0012] To achieve the above object and other related objects, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is enabled to execute one or more of the foregoing methods.

[0013] As described above, a model training method, device, pantograph-catenary system simulation method, device, and medium provided by the present application have the following beneficial effects: A model training method in the present application. The method obtains a training sample set, constructs an objective loss function including a physical loss function, constrains the output of the pantograph-catenary simulation model to conform to physical laws through the physical loss function, and finally trains the pantograph-catenary simulation model according to the training sample set. The trained pantograph-catenary simulation model is obtained by minimizing the objective loss function. By adding a physical loss function to the objective loss function, the physical law constraint on the output of the pantograph-catenary simulation model can be realized, and the effect of improving the prediction accuracy of the pantograph-catenary simulation model can be achieved.

[0014] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings: Figure 1 is a flowchart of a model training method shown in an exemplary embodiment of the present application; Figure 2 is a schematic structural diagram of a pantograph-catenary simulation model shown in an exemplary embodiment of the present application; Figure 3 is a schematic diagram of the prediction results of the pantograph-catenary simulation model and the finite element simulation model; wherein, (a1) is the change of the displacement of the contact wire in the X-axis direction with time, (a2) is the change of the derivative of the displacement of the contact wire in the X-axis direction with time, (b1) is the change of the displacement of the contact wire in the Y-axis direction with time, (b2) is the change of the derivative of the displacement of the contact wire in the Y-axis direction with time, (c1) is the change of the displacement of the contact wire in the Z-axis direction with time, and (c2) is the change of the derivative of the displacement of the contact wire in the Z-axis direction with time; Figure 4 It is a schematic diagram of the prediction error of the pantograph-catenary simulation model. Among them, (a1) is the variation of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the X-axis direction with time, (a2) is the variation of the derivative of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the X-axis direction with time, (b1) is the variation of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the Y-axis direction with time, (b2) is the variation of the derivative of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the Y-axis direction with time, (c1) is the variation of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the Z-axis direction with time, and (c2) is the variation of the derivative of the displacement of the prediction results of the finite element simulation model and the pantograph-catenary simulation model in the Z-axis direction with time; Figure 5 It is the frequency domain feature diagram of the prediction results of the pantograph-catenary simulation model and the finite element simulation model. Among them, (a1) is the amplitude variation of the displacement of the catenary in the X-axis direction in the frequency domain, (a2) is the amplitude variation of the derivative of the displacement of the catenary in the X-axis direction in the frequency domain, (b1) is the amplitude variation of the displacement of the catenary in the Y-axis direction in the frequency domain, (b2) is the amplitude variation of the derivative of the displacement of the catenary in the Y-axis direction in the frequency domain, (c1) is the amplitude variation of the displacement of the catenary in the Z-axis direction in the frequency domain, and (c2) is the amplitude variation of the derivative of the displacement of the catenary in the Z-axis direction in the frequency domain; Figure 6 It is the flowchart of the pantograph-catenary system simulation method shown in an exemplary embodiment of the present application; Figure 7 It is the block diagram of the model training device shown in an exemplary embodiment of the present application; Figure 8 It is the block diagram of the pantograph-catenary system simulation device shown in an exemplary embodiment of the present application. Detailed implementation manners

[0016] The following will describe the implementation manners of the present application with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for illustrating the present application, rather than for limiting the protection scope of the present application.

[0017] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present application. Therefore, only the components related to the present application are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0018] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.

[0019] Please refer to Figure 1 , Figure 1 which is a flowchart of the model training method shown in an exemplary embodiment of the present application. Referring to Figure 1 it can be seen that the model training method may include: Step S110, obtaining a training sample set.

[0020] Among them, the training sample set includes multiple training samples. The training sample includes an input feature and a sample label. The input feature includes initial simulation parameters and the pantograph-catenary simulation results at multiple consecutive time steps after performing a finite element simulation on the pantograph-catenary system. The sample label includes the pantograph-catenary simulation results at the next time step after multiple consecutive time steps.

[0021] In an embodiment of the present application, the process of obtaining the training sample set may include: obtaining initial simulation parameters, which can be input by the user and may include parameters such as span and tension. Initialize the finite element simulation model of the pantograph-catenary system according to each initial simulation parameter, and randomly set the running speed and base uplift force in the finite element simulation model. After simulation using the finite element simulation model, a trajectory is generated. This trajectory characterizes the various response states of the system during the operation of the finite element simulation model as the input data changes. The pantograph-catenary simulation results output by the finite element simulation model can be segmented into multiple training samples with a preset length of sliding window, that is, the training sample set, and the training sample set is stored in the data set buffer : ; ; Among them, represents the state matrix, represents the external force matrix, represents the th sample segmented from the trajectory, N represents the length of the sliding window, Represents the true value. The pantograph-catenary simulation results include the state matrix, external force matrix, mass matrix, damping matrix, and stiffness matrix. Since the elements in the external force matrix, mass matrix, damping matrix, and stiffness matrix are all constants, when solving the loss function, only the relevant loss function is determined for the state matrix, and the true value here only uses the state matrix.

[0022] Then, for the data in the dataset buffer The data in is standardized using the Z-score calculation formula to make the basic dimensions of the initial simulation parameters and the pantograph-catenary simulation results match each other. The specific calculation formula is as follows: ; Among them, is the original element in the matrix , and the subscript represents the row index, and the subscript represents the column index. are the mean and standard deviation of the th row respectively. is the result after standardizing the element in the matrix. After standardizing all the elements in the matrix, we get .

[0023] It should be noted that by using finite element simulation based on the initial simulation parameters to obtain the pantograph-catenary simulation results, a sufficient number of training samples can be obtained, and the quality of the training samples meets the training requirements. The pantograph-catenary system is a power transmission system composed of a pantograph and a catenary.

[0024] Step S120, construct the target loss function of the pantograph-catenary simulation model; among them, the target loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws.

[0025] In an embodiment of the present application, the target loss function can measure the gap between the predicted value and the true value of the pantograph-catenary simulation model. The predicted value therein is the pantograph-catenary simulation result obtained by the pantograph-catenary simulation model according to the initial simulation parameters, and the true value is the pantograph-catenary simulation result obtained by finite element simulation based on the initial simulation parameters. After determining the gap between the predicted value and the true value according to the target loss function, based on the gradient of the loss function, through iterative optimization, the weights of the pantograph-catenary simulation model can be continuously adjusted, and the loss value can be gradually reduced to make the predicted value of the pantograph-catenary simulation model closer to the distribution of the true value.

[0026] In one embodiment, the pantograph-catenary simulation results include a state matrix, an external force matrix, a mass matrix, a damping matrix, and a stiffness matrix. The physical loss function can include: ; Among them, is the loss value of the physical loss function, is the total number of training samples, represents the th training sample, is the physical loss term, , represents the state matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the external force matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the mass matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the damping matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample, represents the stiffness matrix obtained according to the pantograph-catenary simulation model and the input features of the th training sample In one embodiment, the basic loss function is the mean square error loss function. The mean square error loss function may include: ; wherein, is the loss value of the basic loss function, the th state matrix of the pantograph-catenary simulation result in the training sample.

[0027] In one embodiment, the process of constructing the target loss function of the pantograph-catenary simulation model may include: determining the weighted sum of the basic loss function and the physical loss function as the target loss function; or, determining the sum of the basic loss function and the physical loss function as the target loss function.

[0028] In a possible implementation manner, the target loss function may include: ; wherein, is the target loss function, is the first weight corresponding to the basic loss function, is the second weight corresponding to the physical loss function. The first weight and the second weight are in a mutually coupled relationship, and the first weight and the second weight can be set by the operator.

[0029] In another possible implementation manner, the target loss function may include: ; wherein, is the target loss function.

[0030] In one embodiment, the pantograph-catenary simulation model includes a temporal convolutional network, a long short-term memory network, a multi-head attention mechanism, and a fully connected network, which are connected in sequence.

[0031] It should be noted that by constructing a physical loss function, physical constraints can be used to enhance the interpretability, generalization ability, and prediction accuracy of the model. Physical constraints can also limit the output within a reasonable physical range and reduce the propagation of uncertainty.

[0032] Exemplarily, the pantograph-catenary simulation model may include: 1. The Temporal Convolutional Network (TCN) consists of causal convolution, dilated convolution, and residual blocks. The specific functions of each layer are as follows: Causal convolution: Ensure that the output at time only depends on historical inputs. For the input sequence and the filter ; where represents the convolution kernel size, and represents the convolution function.

[0033] Dilated convolution: Expand the receptive field by the dilation factor to extract local features in long sequences. For the dilated convolution operation in the th layer, it is defined as follows: ; ; where is the output sequence, represents the dilation factor of the th layer, represents the convolution kernel size, represents the past convolution direction, and is obtained from the previous causal convolution.

[0034] Residual Block: Add the transformation to the identity mapping of the input of the residual block to avoid the problem of gradient vanishing or gradient explosion in deep networks, making the network training more stable. It consists of two layers of dilated causal convolution, weight normalization, spatial dropout, and ReLU activation function. The specific formula is as follows: ; where represents the original input sequence, represents the features obtained through dilated convolution, and Represents the output of the incomplete block.

[0035] 2. Long Short-Term Memory Network (LSTM) captures long-term dependencies through a gating mechanism, consisting of a forget gate, an input gate, an output gate, and a candidate state. At the same time, a dropout layer is introduced to prevent overfitting of the results. The specific formula is as follows: ; ; ; ; ; ; Among them, , , , and respectively represent the forget gate, input gate, output gate, cell state, and candidate cell state at time . is an element in the matrix, and represents the long-term dependency captured by the Long Short-Term Memory Network. and respectively represent the weight matrix and bias vector corresponding to the gate, and respectively represent the sigmoid function and the hyperbolic tangent function, represents the dot product of vectors.

[0036] 3. Multi-head Attention mechanism calculates the correlation between different positions in the input sequence to establish long-range dependencies. It consists of multiple parallel self-attention heads. Each self-attention head independently calculates the attention weights and generates the output, as shown in the following formula: ; ; ; ; ; ; Among them, , and respectively represent the query matrix, key matrix, and value matrix. respectively represent the corresponding learned weight matrices, represents the number of attention heads, represents the dimension of each head, then represents the result output by the multi-head attention mechanism.

[0037] 4. The fully connected network (TNN) is located in the output layer of the network, and multiple TNNs together form a multi-layer perceptron (MLP). The TNN is responsible for mapping the result output by the multi-head attention mechanism to the final pantograph-catenary simulation result , which is defined as follows: ; wherein, , and respectively represent the weight matrix, bias vector and Relu activation function in the fully connected layer.

[0038] Step S130, training the pantograph-catenary simulation model according to the training sample set, and obtaining the trained pantograph-catenary simulation model by minimizing the objective loss function.

[0039] In an embodiment of the present application, the pantograph-catenary simulation model can be used to obtain the pantograph-catenary simulation result according to the input initial simulation parameters. After obtaining the trained pantograph-catenary simulation model, if it is necessary to simulate the pantograph-catenary system, the initial simulation parameters can be input into the pantograph-catenary simulation model, and the pantograph-catenary simulation model will show the pantograph-catenary simulation result. Compared with the related art that uses the finite element method to simulate the pantograph-catenary system, using the pantograph-catenary simulation model obtained in the embodiment of the present application can reduce the simulation calculation amount, and there is no need to perform finite element simulation modeling on the pantograph-catenary system, which can reduce the complexity of the simulation.

[0040] Exemplarily, the training process may include: Sampling a batch of data from the data set buffer to train the pantograph-catenary simulation model. The specific training process is as follows: 1. Performing local feature extraction on the input sequence through a temporal convolutional network to obtain .

[0041] 2. Inputting into a long short-term memory network to capture long-term dependencies .

[0042] 3. Introducing a multi-head attention mechanism to the long-term dependencies captured by the long short-term memory network to generate .

[0043] 4. Inputting into the fully connected network to map it to the final dynamic response output .​

[0044] 5. Based on After calculating the loss function, the pantograph-catenary simulation model is repeatedly trained using the target loss function, the Adam optimizer, and the gradient descent algorithm until the results meet the expectations.

[0045] Exemplarily, please refer to Figure 2 , which is a schematic structural diagram of the pantograph-catenary simulation model shown in an exemplary embodiment of the present application. Taking the Beijing-Tianjin-Hebei line of the high-speed rail line as an example, the finite element simulation method and the supplier simulation model provided in the embodiments of the present application are used to perform finite element simulations on the Beijing-Tianjin-Hebei line of the high-speed rail line respectively. Figure 3 And Figure 4 is a comparison between the method provided in the embodiments of the present application and the traditional method (finite element method). It can be seen that as the time step progresses, the displacements of the pantograph in the x, y, and z directions predicted by the pantograph-catenary simulation model and the displacement change rates in the x, y, and z directions are all very consistent with the actual results calculated by the finite element simulation model, indicating that the pantograph-catenary simulation model can accurately predict the dynamic response. In Figure 3 , the horizontal axis represents the time step, and the vertical axis represents the displacement, with the unit of m. In Figure 4 , the horizontal axis represents the time step, and the vertical axis represents the prediction error, with the unit of m.

[0046] To fully compare the prediction results of FE (blue line) and the prediction results of the pantograph-catenary simulation model (red line), as Figure 5 shown, which gives the frequency domain characteristics of the node dynamic response. In Figure 5 , the horizontal axis represents the frequency, with the unit of HZ, and the vertical axis represents the amplitude. Through Figure 5 the following conclusions can be drawn: 1. The degree-of-freedom displacements of the catenary nodes indicate that the prediction results of the pantograph-catenary simulation model are very consistent with the actual results calculated by FE in terms of amplitude and shape, indicating that the pantograph-catenary simulation model can accurately predict the dynamic response.

[0047] 2. Although the vibration trends of the dynamic response curves are the same, the high-frequency components are slightly different. Figure 5 The frequency domain diagrams in

[0048] further illustrate this phenomenon. They are the same between 0 - 200 Hz, but slightly different between 200 - 400 Hz.

[0049] Through the above examples, it is illustrated that the pantograph-catenary simulation model provided by the embodiments of the present application improves the previous simulation method and enhances the calculation speed of the model, which has practical significance for the further development of high-speed railways.

[0050] Figure 6 It is a flowchart of the pantograph-catenary system simulation method shown in an exemplary embodiment of the present application. Refer to Figure 6 It can be seen that the pantograph-catenary system simulation method may include: Step S610, obtaining the current simulation parameters of the pantograph-catenary system.

[0051] In an embodiment of the present application, the current simulation parameters may include parameters such as span and tension.

[0052] Step S620, determining the current pantograph-catenary simulation result according to the current simulation parameters and the pantograph-catenary simulation model. Among them, the pantograph-catenary simulation model is trained based on the Figure 1 method in the embodiment.

[0053] Figure 7 It is a block diagram of the model training device shown in an exemplary embodiment of the present application. As Figure 7 shown, the exemplary model training device 700 includes: A training data acquisition module 710, configured to acquire a training sample set; the training sample set includes a plurality of training samples, and each training sample includes initial simulation parameters and the pantograph-catenary simulation result obtained by performing finite element simulation on the pantograph-catenary system according to the initial simulation parameters.

[0054] A loss function construction module 720, configured to construct an objective loss function for the pantograph-catenary simulation model; the objective loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws.

[0055] A model training module 730, configured to train the pantograph-catenary simulation model according to the training sample set, and obtain the trained pantograph-catenary simulation model by minimizing the objective loss function.

[0056] Figure 8 It is a block diagram of the pantograph-catenary system simulation device shown in an exemplary embodiment of the present application. As Figure 8 shown, the exemplary pantograph-catenary system simulation device 800 includes: A current data acquisition module 810, configured to acquire the current simulation parameters of the pantograph-catenary system.

[0057] It should be noted that the power system composed of a pantograph and a catenary is called a pantograph-catenary system.

[0058] A result determination module 820 is configured to determine a current pantograph-catenary simulation result according to current simulation parameters and a pantograph-catenary simulation model. The pantograph-catenary simulation model is trained based on the method in Figure 1 the embodiment.

[0059] An embodiment of the present application further provides an electronic device, including: one or more processors; a storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the methods provided in the above various embodiments.

[0060] On the other hand, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is caused to execute the methods provided in the above various embodiments. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist separately without being assembled into the electronic device.

[0061] On the other hand, the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to execute the methods provided in the above various embodiments.

[0062] In the embodiments of the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. The terms "comprising" and "including" mentioned throughout the specification and claims are open-ended terms and should be construed as "including but not limited to".

[0063] The above embodiments are only used to exemplarily illustrate the principles and effects of the present application, rather than to limit the present application. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present application should still be covered by the claims of the present application.

Claims

1. A model training method, characterized in that: include: Obtain a training sample set; The training sample set includes multiple training samples, and the training samples include input features and sample labels. The input features include initial simulation parameters and the simulation results of the bow-net system at multiple continuous time steps after finite element simulation of the bow-net system. The sample labels include the simulation results of the bow-net system at one time step after the multiple continuous time steps. Constructing the target loss function of the pantograph-catenary simulation model; the target loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws; The bow-net simulation model is trained according to the training sample set, and the trained bow-net simulation model is obtained by minimizing the target loss function.

2. The model training method according to claim 1, characterized in that: The pantograph-catenary simulation results include the state matrix, external force matrix, mass matrix, damping matrix and stiffness matrix. The physical loss functions include: ； in, is the loss value of the physical loss function, is the total number of training samples, Indicates training samples, is the physical loss term, , According to the pantograph-catenary simulation model and The state matrix obtained by the input features of training samples is According to the pantograph-catenary simulation model and The external force matrix obtained by the input features of training samples is According to the pantograph-catenary simulation model and The quality matrix obtained by the input features of training samples is According to the pantograph-catenary simulation model and The damping matrix obtained by the input features of training samples is According to the pantograph-catenary simulation model and The stiffness matrix is ​​obtained by combining the input features of training samples.

3. The model training method according to claim 1, characterized in that: Construct the objective loss function of the pantograph simulation model, including: The weighted sum of the basic loss function and the physical loss function is determined as the target loss function.

4. The model training method according to claim 1, characterized in that: Construct the objective loss function of the pantograph simulation model, including: The sum of the basic loss function and the physical loss function is determined as the target loss function.

5. The model training method according to claim 3 or 4, characterized in that: The basic loss function is the mean square error loss function.

6. The model training method according to claim 1, characterized in that: The bow network simulation model includes a temporal convolutional network, a long short-term memory network, a multi-head attention mechanism, and a fully connected network that are connected in sequence.

7. A method for simulating a pantograph-catenary system, characterized in that: include: Get the current simulation parameters of the pantograph-catenary system; According to the current simulation parameters and the bow-net simulation model, the current bow-net simulation result is determined; wherein the bow-net simulation model is trained based on the method described in any one of claims 1-6.

8. A model training device, characterized in that: include: A training data acquisition module is used to acquire a training sample set; The training sample set includes a plurality of training samples, and the training samples include initial simulation parameters and a pantograph-catenary simulation result obtained after a finite element simulation of the pantograph-catenary system is performed according to the initial simulation parameters; A loss function construction module is used to construct a target loss function of the pantograph-catenary simulation model; the target loss function includes a physical loss function, and the physical loss function is used to constrain the output of the pantograph-catenary simulation model to conform to physical laws; The model training module is used to train the bow-net simulation model according to the training sample set, and obtain the trained bow-net simulation model by minimizing the target loss function.

9. A pantograph-catenary system simulation device, characterized in that: include: The current data acquisition module is used to obtain the current simulation parameters of the pantograph-catenary system; The result determination module is used to determine the current bow-net simulation result according to the current simulation parameters and the bow-net simulation model; wherein the bow-net simulation model is trained based on the method described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the method according to any one of claims 1 to 6.

Images