Vehicle cabin temperature field rapid simulation method, device and system and storage medium
By combining the finite element simulation results with neural networks and introducing a multi-head attention mechanism, the existing automotive air conditioning system has been solved, and higher accuracy and fast cabin temperature field simulation is achieved.
Patent Information
- Application Number
- CN202510101563.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-22
- Publication Date
- 2025-05-06
AI Technical Summary
The simulation work of existing automotive air conditioning systems relies on empirical formulas, resulting in low accuracy of energy consumption calculation and cannot meet the needs of improving the energy efficiency of the whole vehicle. At the same time, due to the high cost and long calculation time, finite element simulation limits its iterative computing capability in actual vehicle hardware testing.
Combining the finite element simulation results and neural networks, a multi-head attention mechanism is introduced, focusing on the more important parts of the input volume from different angles, establishing a higher-precision air-conditioning system, and improving the speed of simulation.
It realizes rapid prediction of the temperature field in buses equipped with air conditioning systems, helps to study and optimize the energy management strategies of electric buses, and improves simulation accuracy and speed.
Smart Images
Figure CN119940135A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of electric vehicle energy management, and in particular to a method, device, system and storage medium for rapid simulation of vehicle cabin temperature field. Background Art
[0002] In the energy management of automotive air conditioning systems, it is crucial to accurately simulate the temperature distribution in the cabin. Although finite element simulation technology performs well in solving engineering problems, especially in temperature field modeling, most of the current simulation work of automotive air conditioning systems relies on empirical formulas, which leads to low accuracy in energy consumption calculations and cannot meet the needs of improving vehicle energy efficiency. Although finite element simulation has been applied in temperature field modeling, its high cost and long calculation process limit its iterative computing capabilities in actual vehicle hardware testing. Summary of the invention
[0003] In order to solve the above problems, the present application provides a method, device, system and storage medium for rapid simulation of cabin temperature field, so as to combine finite element simulation results with neural network, and introduce a multi-head attention mechanism to focus on the more important parts of the input quantity from different angles, thereby increasing the speed on the basis of establishing a higher precision air-conditioning system.
[0004] In order to achieve the above purpose, the technical solution adopted in this application is as follows:
[0005] In a first aspect, the present application provides a method for rapid simulation of a vehicle cabin temperature field, the method comprising:
[0006] The three-dimensional bus model is subjected to finite element simulation analysis to obtain a series of bus cabin temperature values within a preset time;
[0007] Normalizing the bus cabin temperature value sequence so that all data are limited to a range of 0-1 to obtain normalized data;
[0008] Constructing a future temperature prediction model; wherein the future temperature prediction model includes a time series convolutional network and a bidirectional long short-term memory neural network, the time series convolutional network uses the normalized data as input to extract local features of the input, and the bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates;
[0009] A multi-head attention mechanism is introduced into the future temperature prediction model, so that the future temperature prediction model pays attention to different parts of the sequence data in parallel.
[0010] Furthermore, a DBO sample is obtained, and a finite element simulation analysis is performed on the bus three-dimensional model to obtain a bus cabin temperature value sequence within a preset time, including:
[0011] Build a 3D model of the bus;
[0012] Based on the solid wall boundary conditions, solar radiation, material properties of the body wall and external flow field area of the preset three-dimensional bus model, a finite element simulation analysis is performed on the three-dimensional bus model to calculate the solar radiation at a set time, and through iterative calculation, a sequence of bus cabin temperature values within the preset time is obtained.
[0013] Furthermore, the bus cabin temperature value sequence is normalized by the following formula:
[0014]
[0015] In the formula, x i ' is the data after Min-Max normalization, x i is the original data, min(x) is the minimum value of the sample, and max(x) is the maximum value of the sample.
[0016] Furthermore, the temporal convolutional network is used to extract the changing trend of the temperature value; wherein, the size of the convolution kernel in the temporal convolutional network is determined according to the value of each unit in the temperature value sequence and the size of the target area, the number of convolution kernels in the temporal convolutional network is determined according to the number of features set to be learned in each convolutional layer, and the values in the expansion parameter list of the temporal convolutional network are gradually increased so that in different convolutional layers, the time step covered by the convolution kernel gradually increases, thereby capturing longer-distance dependencies.
[0017] Furthermore, the bidirectional long short-term memory neural network is set to two, wherein the first bidirectional long short-term memory neural network takes the output of the temporal convolutional network as input and returns a complete sequence for each time step. Based on the two bidirectional long short-term memory neural networks, the bidirectional dependency in the sequence is captured by combining the outputs of the forward and backward LSTM networks. A Dropout layer is set after each bidirectional long short-term memory neural network to reduce overfitting of the model during training through regularization, and a first fully connected layer is set after the Dropout layer to output a sequence of future temperature values of each point under the spatial coordinates.
[0018] Furthermore, the multi-head attention mechanism outputs the final prediction result by stacking multiple attention layers in parallel and setting a second fully connected layer after the multiple attention layers; wherein the multiple attention layers are used to perform different line-to-line transformations on the same input, and the input of the attention layer is the output of the bidirectional long short-term memory neural network.
[0019] Furthermore, performing different linear transformations on the same input is to use three vectors Q, K and V to perform m linear projections through a learnable linear projection matrix; where m is the number of attention layers.
[0020] In a second aspect, the present application provides a rapid simulation device for a vehicle cabin temperature field, the device comprising:
[0021] The finite element analysis module is configured to perform finite element simulation analysis on the three-dimensional model of the bus to obtain a sequence of bus cabin temperature values within a preset time;
[0022] A normalization module is configured to perform normalization processing on the bus cabin temperature value sequence so that all data are limited to a range of 0-1 to obtain normalized data;
[0023] A feature extraction module is configured to construct a future temperature prediction model; wherein the future temperature prediction model includes a time series convolutional network and a bidirectional long short-term memory neural network, the time series convolutional network uses the normalized data as input to extract local features of the input, and the bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates;
[0024] The temperature prediction module is configured to introduce a multi-head attention mechanism into the future temperature prediction model so that the future temperature prediction model pays attention to different parts of the sequence data in parallel.
[0025] In a third aspect, the present application provides a system for rapid simulation of vehicle cabin temperature field, the system comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the method described above.
[0026] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium storing instructions, which, when executed by a processor, executes the method described above.
[0027] This application has at least the following beneficial effects:
[0028] This application combines neural network and finite element simulation technology, and incorporates an attention mechanism to focus on monitoring temperature changes in key areas. This method can quickly predict the temperature field inside a bus equipped with an air-conditioning system, which helps to study and optimize the energy management strategy of electric buses. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 The present invention is a flowchart of a method for rapid simulation of vehicle cabin temperature field according to an embodiment of the present application.
[0030] Figure 2 This is a three-dimensional model diagram of a bus according to an embodiment of the present application.
[0031] Figure 3 It is a temperature field model obtained by finite element simulation according to an embodiment of the present application.
[0032] Figure 4 This is a structural diagram of a multi-head attention mechanism according to an embodiment of the present application.
[0033] Figure 5 The three-dimensional spatial temperature map of a bus is predicted by the trained neural network according to an embodiment of the present application.
[0034] Figure 6 It is a structural diagram of a rapid simulation device for vehicle cabin temperature field according to an embodiment of the present application. DETAILED DESCRIPTION
[0035] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.
[0036] The specific implementation of the present application is further described in detail below in conjunction with the drawings and examples.
[0037] See also Figure 1 , is a flow chart of a method for rapid simulation of vehicle cabin temperature field, the method for rapid simulation of vehicle cabin temperature field includes steps S1 to S4, which are described in detail as follows.
[0038] S1, data acquisition: build a three-dimensional model of the bus, substitute it into ANSYS for finite element simulation analysis, and solve the temperature value sequence and spatial coordinates.
[0039] Specifically, firstly, a 3D bus model is established to determine the specific size of the bus. The schematic diagram of the 3D bus model is as follows: Figure 2 As shown in the figure, the established model is substituted into ANSYS for finite element simulation analysis. Through ANSYS software, the three-dimensional geometric data of the bus is imported, and the solid wall boundary conditions, solar radiation, material properties of the vehicle wall and the external flow field area of the model are set. According to the set amount, ANSYS automatically calculates the solar radiation at a specific time. Through iterative calculation, the temperature field model obtained by finite element simulation is as follows Figure 3 As shown, a stable sequence of temperature values in three-dimensional space is obtained.
[0040] S2, data preprocessing, performing Min-Max normalization processing on the obtained data.
[0041] In this embodiment, the data extracted in step S1 is subjected to Min-Max normalization so that all data are limited to the range of 0-1. The minimum and maximum normalization formula is:
[0042]
[0043] In the formula, x i ' is the data after Min-Max normalization, x i is the original data, min(x) is the minimum value of the sample, and max(x) is the maximum value of the sample.
[0044] S3, future temperature value sequence model: Build a TCN-BiLSTM neural network (future temperature prediction model) to predict future temperature value sequences.
[0045] In this embodiment, the TCN-BiLSTM neural network includes a temporal convolutional network TCN and a bidirectional long short-term memory neural network BiLSTM. The temporal convolutional network uses the normalized data as input to extract local features of the input. The bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates. The two characteristics of the temporal convolutional network TCN, causal convolution and dilated convolution, are used to extract local features of the input temperature value sequence and strengthen the mining of feature information. The bidirectional long short-term memory network BiLSTM is used to capture the bidirectional dependency in the temperature value sequence by combining the outputs of the previous and next two LSTM networks, and output the prediction results.
[0046] For example, a TCN network is used to extract local features of the temperature value sequence (the changing trend of the temperature value). The size and number of convolution kernels are set to 8 and 32 respectively. The size of the convolution kernel determines how large the value of each unit in the temperature value sequence is related to the input region. When convolving the input matrix, the convolution kernel extracts information of 8 time steps of the input matrix with each sliding step. The number of convolution kernels determines the number of features that the network can learn in each convolution layer. The network can learn 32 different features in each convolution layer. Set the expansion parameter list to [1,2,4,8,16,32,64,128,256]. In different convolution layers, the time steps that the convolution kernel can cover will gradually increase, thereby capturing longer-distance dependencies.
[0047] Two BiLSTM layers are nested in the model. The BiLSTM layer receives the features extracted by TCN as input. The first BiLSTM layer sets return_sequences = True, which will return the complete sequence of each time step, which is necessary to maintain the integrity of the time series data in the subsequent layers. Through the BiLSTM layer, the bidirectional dependencies in the sequence can be captured by combining the outputs of the forward and backward LSTM networks. Dropout layers are added after the two BiLSTM layers, and regularization can reduce the overfitting of the model during training. Finally, a fully connected layer is used to output the final prediction results.
[0048] S4, introduces the multi-head attention mechanism, adds the multi-head attention mechanism to the TCN-BiLSTM neural network, outputs the prediction results again, and compares and analyzes them with the previous prediction results.
[0049] In this embodiment, a multi-head attention mechanism is introduced into the TCN-BiLSTM network built in step S3. The attention mechanism can identify and strengthen those information points that are crucial to the final prediction results when processing time series data. The multi-head attention mechanism is one of the core features of the Transformer architecture. By stacking m attention layers in parallel, different linear transformations are performed on the same input, that is, Q, K, and V are linearly projected m times through a learnable linear projection matrix. It enables the model to pay attention to different parts of the sequence data in parallel. Under this mechanism, each attention head can capture different features or patterns in the sequence, and then integrate this information. In this way, the model can gain a more comprehensive understanding of the input data. The multi-head attention mechanism is added to the TCN-BiLSTM neural network, and 4 attention heads are set, each with a dimension of 256. The structure of the multi-head attention mechanism is as follows Figure 4 As shown in the figure, F is the data matrix, Q is the query vector, and C is the context vector. Finally, a fully connected layer is used to output the final prediction result.
[0050] The data preprocessed by the time series convolution network (TCN) is used as input. The BiLSTM neural network learns and captures the time series characteristics of the temperature change inside the bus, and introduces a multi-head attention mechanism. In the Python environment, the deep learning framework is used to train the above network so that it can accurately predict the temperature value of each coordinate point in the bus within a specific time in the future. After the training is completed, the neural network can output the prediction results, which are attached. Figure 5 It is displayed in the form of a three-dimensional temperature diagram.
[0051] By combining neural network with finite element simulation, a simulation prediction model of cabin temperature field is built, which solves the problem of long calculation time and inaccurate precision of finite element simulation on complex temperature field, and speeds up the simulation time of cabin temperature field. The introduction of multi-head attention mechanism can make the network focus on the more important part of the input from different angles, improve the prediction accuracy, compare and analyze the two prediction results, obtain the input part with larger loss change, and then analyze which part of the cabin has larger temperature change.
[0052] The present application also provides a rapid simulation device for cabin temperature field. Figure 6 As shown, the device comprises:
[0053] The finite element analysis module 601 is configured to perform finite element simulation analysis on the three-dimensional model of the bus to obtain a sequence of bus cabin temperature values within a preset time;
[0054] A normalization module 602 is configured to perform normalization processing on the bus cabin temperature value sequence so that all data are limited to a range of 0-1 to obtain normalized data;
[0055] The feature extraction module 603 is configured to construct a future temperature prediction model; wherein the future temperature prediction model includes a time series convolutional network and a bidirectional long short-term memory neural network, the time series convolutional network uses the normalized data as input to extract local features of the input, and the bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates;
[0056] The temperature prediction module 604 is configured to introduce a multi-head attention mechanism into the future temperature prediction model, so that the future temperature prediction model pays attention to different parts of the sequence data in parallel.
[0057] In some embodiments, the finite element analysis module is further configured to:
[0058] Build a 3D model of the bus;
[0059] Based on the solid wall boundary conditions, solar radiation, material properties of the body wall and external flow field area of the preset three-dimensional bus model, a finite element simulation analysis is performed on the three-dimensional bus model to calculate the solar radiation at a set time, and through iterative calculation, a sequence of bus cabin temperature values within the preset time is obtained.
[0060] In some embodiments, the normalization module is further configured to normalize the bus cabin temperature value sequence by the following formula:
[0061]
[0062] In the formula, x i ' is the data after Min-Max normalization, x i is the original data, min(x) is the minimum value of the sample, and max(x) is the maximum value of the sample.
[0063] In some embodiments, the temporal convolutional network is used to extract the changing trend of temperature values; wherein, the size of the convolution kernel in the temporal convolutional network is determined according to the value of each unit in the temperature value sequence and the size of the target area, the number of convolution kernels in the temporal convolutional network is determined according to the number of features set to be learned in each convolutional layer, and the values in the expansion parameter list of the temporal convolutional network are gradually increased so that in different convolutional layers, the time step covered by the convolution kernel gradually increases, thereby capturing longer distance dependencies.
[0064] In some embodiments, the bidirectional long short-term memory neural network is set to two, wherein the first bidirectional long short-term memory neural network takes the output of the temporal convolutional network as input and returns a complete sequence for each time step. Based on the two bidirectional long short-term memory neural networks, the bidirectional dependency in the sequence is captured by combining the outputs of the forward and backward LSTM networks. A Dropout layer is set after each bidirectional long short-term memory neural network to reduce overfitting of the model during training through regularization, and a first fully connected layer is set after the Dropout layer to output a sequence of future temperature values of each point under the spatial coordinates.
[0065] In some embodiments, the multi-head attention mechanism outputs a final prediction result by stacking multiple attention layers in parallel and setting a second fully connected layer after the multiple attention layers; wherein the multiple attention layers are used to perform different line-to-line transformations on the same input, and the input of the attention layer is the output of a bidirectional long short-term memory neural network.
[0066] In some embodiments, performing different linear transformations on the same input is to perform m linear projections using three vectors Q, K, and V through a learnable linear projection matrix; where m is the number of attention layers.
[0067] It should be noted that the device described in this embodiment and the method described previously belong to the same technical concept, have the same technical principles, and can achieve the same beneficial effects, so they will not be repeated here.
[0068] An embodiment of the present application further provides a system for rapid simulation of vehicle cabin temperature field, the system comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the method described in any of the above embodiments.
[0069] An embodiment of the present application further provides a non-temporary computer-readable storage medium storing instructions, and when the instructions are executed by a processor, the method described in any of the above embodiments is executed.
[0070] The above implementation modes are only used to illustrate the present application, and are not intended to limit the present application. Ordinary technicians in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present application. Therefore, all equivalent technical solutions also belong to the scope of the present application, and the scope of patent protection of the present application shall be limited by the claims.
Claims
1. A rapid simulation method for cabin temperature field, characterized in that: The method comprises: The three-dimensional bus model is subjected to finite element simulation analysis to obtain a series of bus cabin temperature values within a preset time; Normalizing the bus cabin temperature value sequence so that all data are limited to a range of 0-1 to obtain normalized data; Constructing a future temperature prediction model; wherein the future temperature prediction model includes a time series convolutional network and a bidirectional long short-term memory neural network, the time series convolutional network uses the normalized data as input to extract local features of the input, and the bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates; A multi-head attention mechanism is introduced into the future temperature prediction model, so that the future temperature prediction model pays attention to different parts of the sequence data in parallel.
2. The method for rapid simulation of cabin temperature field according to claim 1, characterized in that: Obtain DBO samples, perform finite element simulation analysis on the bus 3D model, and obtain the bus cabin temperature value sequence within a preset time, including: Build a 3D model of the bus; Based on the solid wall boundary conditions, solar radiation, material properties of the body wall and external flow field area of the preset three-dimensional bus model, a finite element simulation analysis is performed on the three-dimensional bus model to calculate the solar radiation at a set time, and through iterative calculation, a sequence of bus cabin temperature values within the preset time is obtained.
3. The method for rapid simulation of cabin temperature field according to claim 1, characterized in that: The bus cabin temperature value sequence is normalized by the following formula: In the formula, x i ' is the data after Min-Max normalization, x i is the original data, min(x) is the minimum value of the sample, and max(x) is the maximum value of the sample.
4. The method for rapid simulation of cabin temperature field according to claim 1, characterized in that: The temporal convolutional network is used to extract the changing trend of the temperature value; wherein, the size of the convolution kernel in the temporal convolutional network is determined according to the value of each unit in the temperature value sequence and the size of the target area, the number of convolution kernels in the temporal convolutional network is determined according to the number of features set to be learned in each convolutional layer, and the values in the expansion parameter list of the temporal convolutional network are gradually increased so that in different convolutional layers, the time step covered by the convolution kernel gradually increases, thereby capturing longer-distance dependencies.
5. The method for rapid simulation of cabin temperature field according to claim 4, characterized in that: The bidirectional long short-term memory neural network is set to two, wherein the first bidirectional long short-term memory neural network takes the output of the temporal convolutional network as input and returns a complete sequence for each time step. Based on the two bidirectional long short-term memory neural networks, the bidirectional dependency in the sequence is captured by combining the outputs of the forward and backward LSTM networks. A Dropout layer is set after each bidirectional long short-term memory neural network to reduce overfitting of the model during training through regularization, and a first fully connected layer is set after the Dropout layer to output a sequence of future temperature values of each point under the spatial coordinates.
6. The method for rapid simulation of cabin temperature field according to claim 1, characterized in that: The multi-head attention mechanism outputs the final prediction result by stacking multiple attention layers in parallel and setting a second fully connected layer after the multiple attention layers; wherein the multiple attention layers are used to perform different line-to-line transformations on the same input, and the input of the attention layer is the output of the bidirectional long short-term memory neural network.
7. The method for rapid simulation of cabin temperature field according to claim 6, characterized in that: Performing different linear transformations on the same input is to use three vectors Q, K and V to perform m linear projections through a learnable linear projection matrix; where m is the number of attention layers.
8. A rapid simulation device for cabin temperature field, characterized in that: The device comprises: The finite element analysis module is configured to perform finite element simulation analysis on the three-dimensional model of the bus to obtain a sequence of bus cabin temperature values within a preset time; A normalization module is configured to perform normalization processing on the bus cabin temperature value sequence so that all data are limited to a range of 0-1 to obtain normalized data; A feature extraction module is configured to construct a future temperature prediction model; wherein the future temperature prediction model includes a time series convolutional network and a bidirectional long short-term memory neural network, the time series convolutional network uses the normalized data as input to extract local features of the input, and the bidirectional long short-term memory neural network uses the local features of the input as input to capture the time features of the temperature value sequence and output the future temperature value sequence of each point under the spatial coordinates; The temperature prediction module is configured to introduce a multi-head attention mechanism into the future temperature prediction model so that the future temperature prediction model pays attention to different parts of the sequence data in parallel.
9. A rapid simulation system for cabin temperature field, characterized in that: The system comprises: Memory for storing computer programs; A processor, configured to execute the computer program to implement the method according to any one of claims 1 to 7. 10 . A non-transitory computer-readable storage medium storing instructions, which, when executed by a processor, perform the method according to claim 1 .