Vehicle cabin temperature field prediction method based on neural network
By integrating finite element simulation with a CNN-LSTM neural network to predict vehicle cabin temperatures, the method addresses computational inefficiencies, achieving precise and efficient energy management in electric vehicles.
Patent Information
- Application Number
- CN202510479492.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-16
- Publication Date
- 2025-07-15
AI Technical Summary
Existing finite element simulation methods for predicting vehicle cabin temperature fields in electric vehicles are computationally expensive and time-consuming, limiting their applicability in real-time energy management systems.
Combining finite element simulation with a CNN-LSTM neural network to predict vehicle cabin temperature fields, incorporating factors like vehicle speed and solar radiation, enhances prediction accuracy and reduces computational burden.
The proposed method provides a high-precision simulation model for vehicle cabin temperature prediction, supporting efficient energy management in electric vehicles.
Smart Images
Figure CN120317064A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the energy management of electric vehicles, and particularly to a method for rapidly predicting the temperature field of a bus based on a neural network. Background Art
[0002] In the energy management of electric vehicles combined with an air-conditioning system, the accuracy of the simulated air-conditioning system model is particularly important. The prediction accuracy of the temperature field inside and outside the vehicle affects the accuracy determination of the air-conditioning system. The finite element simulation technology is applied to the processing of the temperature field model as a powerful tool for dealing with complex problems in the engineering field. However, the computational cost of finite element simulation processing is often high, and the computational time often increases with the complexity of the simulation modeling. Summary of the Invention
[0003] In order to solve the above problems existing in the prior art, the object of the present invention is to propose a method for predicting the temperature field of a vehicle cabin based on a neural network, which combines the finite element simulation results of the vehicle cabin with a deep neural network, establishes an external flow field region on the basis of considering disturbance factors such as driving speed and solar radiation, and makes the simulation model more accurate. And use the CNN-LSTM neural network to extract the spatio-temporal features of the data to make the data prediction more accurate.
[0004] In a first aspect, the present case proposes a method for predicting the temperature field of a vehicle cabin based on a neural network, and the steps include: performing finite element simulation of a three-dimensional model based on the vehicle cabin, setting disturbance factors affecting the temperature of the vehicle cabin and material parameters of the vehicle cabin wall in the simulation, and at the same time setting an external flow field region, and obtaining a temperature value sequence corresponding to n uniformly distributed coordinate points as multi-dimensional data through finite element simulation; training a CNN-LSTM neural network based on the multi-dimensional data; inputting the actual vehicle cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values of each monitoring point in the vehicle cabin temperature field for a period of time in the future.
[0005] In an embodiment of the above technical solution, the disturbance factors include driving speed and solar radiation.
[0006] In an embodiment of the above technical solution, the material parameters include the density, specific heat, and thermal conductivity of four materials: aluminum, tempered glass, and high-density polyethylene, and the material parameters also include the density, specific heat, thermal conductivity, and viscosity of air.
[0007] In an embodiment of the above technical solution, the external flow field region is a spatial region from a set distance outside the outer wall surface of the vehicle cabin to the outer wall of the vehicle cabin.
[0008] In a second aspect, the present case proposes a vehicle cabin temperature field prediction system based on a neural network. The system includes a simulation module, a training module, and a prediction module. The simulation module is configured to perform finite element simulation of a three-dimensional model based on the vehicle cabin, set disturbance factors affecting the vehicle cabin temperature and vehicle cabin wall material parameters during the simulation, and simultaneously set the external flow field region. Through finite element simulation, a temperature value sequence corresponding to n uniformly distributed coordinate points is obtained as multi-dimensional data. The training module is configured to train a CNN-LSTM neural network based on the multi-dimensional data. The prediction module is configured to input the actual vehicle cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values at each monitoring point of the vehicle cabin temperature field for a future period of time.
[0009] In a third aspect, the present case proposes a computer-readable storage medium storing a computer program that can be loaded and executed by a processor to perform any of the above methods.
[0010] The beneficial technical effects of the present case: A high-precision simulation model of the electric vehicle cabin temperature is constructed, providing support for electric vehicle energy management. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings 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.
[0012] Figure 1 、 One Schematic diagram of a bus model in a certain implementation manner.
[0013] Figure 2 、 One Schematic diagram of a temperature field model obtained by finite element simulation in a certain implementation manner.
[0014] Figure 3 、 One Schematic diagram of a convolutional neural network structure in a certain implementation manner.
[0015] Figure 4 、 One Schematic diagram of a long short-term memory neural network structure in a certain implementation manner.
[0016] Figure 5 、 One Schematic diagram of predicting the three-dimensional space temperature inside a bus using a trained neural network in a certain implementation manner. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0017] In simulations related to automotive air conditioners, computational fluid dynamics (CFD) software and finite element software play a significant role in establishing a cabin heat load model. According to the established cabin model and the design of disturbance variables to be considered, the simulation quality can be effectively improved, which is beneficial to later research such as data prediction, impact evaluation, and air-conditioning load optimization. However, for the cabin models established by CFD and finite element simulations, although the models are more precise than empirical models and simple models and can improve prediction accuracy, the complex models have a long running time in simulations and are relatively slow in the prediction process, which will lead to the prediction progress being unable to catch up with the required progress. In the technical solutions for studying the temperature field using neural networks, due to the design of a simple temperature field model or simple data relationships being substituted into the neural network training, and there being energy disturbances inside and outside the cabin, the existing technical solutions for studying the temperature field using neural networks are not applicable to the cabin temperature field.
[0018] Based on this, the method in this case combines neural networks with finite element simulations of the cabin temperature field to establish a characteristic association between the vehicle driving space data and its interior temperature field under the condition of changes in the air-conditioning system power, thereby establishing a higher-precision air-conditioning system simulation model under different temperature field conditions. This solution helps to implement the energy management of electric buses equipped with air conditioners.
[0019] Taking a bus as an example below, the implementation of the technical solution in this case will be clearly and completely described. Obviously, the described implementation manners are only part of the implementation manners of this case, rather than all the implementation manners. Based on the implementation manners in this case, all other implementation manners obtained by those of ordinary skill in the art without creative efforts belong to the scope protected by this application.
[0020] (1) Establishing a bus model
[0021] First, build a reasonable three-dimensional bus model, including determining parameters such as the size of the bus, the specific positions of the windows and doors, the size and quantity of the seats inside the vehicle, and the position and quantity of the air-conditioning ducts, and establish a three-dimensional bus model through 3D modeling software.
[0022] Secondly, determine the disturbance factors and the material parameters of each part. The disturbance factors include driving speed, solar radiation amount, etc. The settings of the solar radiation amount during calculation are shown in Table 1.
[0023] Table 1 Solar radiation calculator settings
[0024] Time zone Eastern Time Zone 8 Longitude 118°33’E Latitude 39°24’N Time 9:00, 13:00, 17:00 on June 21st
[0025] For the material parameters of each part, see Table 2.
[0026] Table 2 Material parameters
[0027]
[0028] (2) Calculation of the temperature field inside the vehicle
[0029] Perform finite element simulation through ANSYS software. Import the three-dimensional model of the bus into ANSYS software, and then set the boundary conditions and material coefficients of the bus model wall. The specific parameters can refer to the parameters shown in Table 1, and set the external flow field region. Table 2 shows the settings of the solar radiation calculator in ANSYS software. According to the settings, ANSYS automatically calculates the solar radiation amount at a specific time. Then, through iterative calculation of this software, a stable temperature value sequence is obtained. The external flow field region is the spatial region from a set distance outside the bus wall to outside the bus wall. Exemplarily, the set distance is 1 meter. Since the bus can be regarded as a cuboid, its external flow field region can also be regarded as a cuboid region on each side of the bus wall.
[0030] The established bus model is as Figure 1 shown. Finally, the temperature field model obtained by finite element simulation is as Figure 2 shown, where the dark part is the temperature field region of the passenger compartment, and the light part is the external flow field region.
[0031] Use the adaptive mesh generation strategy in the Fluent mesh generation module. According to the changes in the physical field, dynamically adjust the mesh density. Use smaller elements in areas with large stress concentration and large gradient changes, and use larger elements in uniform areas to improve the calculation efficiency and accuracy.
[0032] After importing the bus model, control the surface mesh size growth rate to be 1.2, the curvature normal angle to be 18°, and perform self-intersection inspection. Set a 40° separation angle constraint and call quality improvement. Set the skewness limit for quality optimization to 0.8. And automatically remesh to remove the influence of the mesh quality brought by the cluster. In the improvement of the surface mesh, set the surface quality limit to 0.7, the maximum angle for quality optimization to 160°, the number of iterations for quality optimization to 5, and the skewness threshold for improving quality based on the collapse method to 0.8. Set the air inlets, outlets, and walls, update the boundaries and regions, and the growth rate of the regular polyhedron mesh generated is 1.2, the maximum aspect ratio is 25, the minimum aspect ratio is 1, the characteristic angle of the polyhedron mesh is 30°, and the gap factor is 0.25. The improved volume mesh element quality limit is 0.15, the minimum angle for quality optimization is 15°, and the number of iterations for quality optimization is 5. The total number of nodes is 8,046,104, the total number of faces is 9,183,053, and the number of mesh elements is 1,374,227.
[0033] Taking the bus as the reference frame, different driving speeds are equivalent to different external flow field wind speeds. Different vehicle speeds (20 m / s, 40 m / s, 60 m / s) are set to obtain the simulated temperature fields under different vehicle speed conditions. During the driving process of the vehicle, due to the relatively high relative speed with the air, the convective heat transfer amount between the vehicle body and the external air increases, which has a significant impact on the change of the temperature field inside the vehicle.
[0034] Obtain the temperature value sequence {T1, T2... T m} and the spatial coordinates {(x1, y1, z1)... (x n , y n , z n )} from ANSYS, where the temperature value sequence {T1, T2... T m} of each point corresponds to its coordinate values (x n , y n , z n ).
[0035] (III) Prediction of future temperature value sequence
[0036] Use a model that combines a one-dimensional convolutional neural network and a long short-term memory network to predict the temperature field in the vehicle cabin.
[0037] First, use two one-dimensional convolutional layers. The first convolutional layer has 128 filters, the convolutional kernel size is 3, and the activation function is ReLU; the second convolutional layer also has 128 filters, the convolutional kernel size is 3, and the activation function is ReLU. A max-pooling layer is used between the two convolutional layers, and the pooling window size is 2, which is used to reduce the data dimension and computational amount while retaining important features. The structure of the convolutional neural network (CNN) is shown in Figure 3 .
[0038] Then, input the feature sequence extracted by the convolutional layer into the LSTM layer. The first LSTM layer has 512 units, and set return_sequences = True to ensure that the output sequence information can be passed to the next layer; then use the Dropout layer with a dropout ratio of 0.2 to prevent overfitting. The second LSTM layer has 256 units and does not return the sequence anymore. Finally, the predicted value is output through the fully connected layer, and the activation function is sigmoid. The structure diagram of the long short-term memory neural network (LSTM) is shown in Figure 4 .
[0039] Use ANSYS to perform finite element simulation on the temperature field of a bus to obtain the temperature value sequence information of each monitoring point inside the cabin. Under the TensorFlow framework, use the mean squared error loss function (MSE), and define the loss function by calling tf.keras.losses.MeanSquaredError(). Its mathematical expression is:
[0040]
[0041] where L is the loss value, n is the number of samples, y i is the true value of the i-th sample, and y' i is the predicted value of the i-th sample. Through the calculation of the loss function and data comparison, it is proved that the obtained prediction accuracy is relatively high.
[0042] Input the actual cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values of each monitoring point in the cabin temperature field for a period of time in the future (see Figure 5 for illustration).
[0043] (IV) Summary
[0044] To sum up, the data processing method proposed in this solution is based on the ANSYS finite element simulation software, and the Fluent module attached to it is used to simulate the temperature field (fluid analysis). The three-dimensional model of the bus and the external flow field area are built through 3D modeling software. The data set obtained through ANSYS finite element simulation includes the three-dimensional coordinates inside the bus and the original temperature value sequence inside the vehicle. Use Python to build a CNN-LSTM neural network, use the data set obtained from ANSYS as the input end, and fit and predict it. There are two reasons for using the CNN-LSTM neural network: one is that the Convolutional Neural Networks (CNN) has the ability of feature learning and can perform translation-invariant classification on the input information according to its hierarchical structure, that is, convert multi-dimensional data into a one-dimensional array; the other is that the Long Short-Term Memory (LSTM) neural network can capture long-term dependence relationships in time series prediction and effectively process data with long-distance dependence and irregular time intervals in time series. Finally, combine the predicted temperature values with the three-dimensional space coordinates to obtain a new bus temperature field model.
[0045] It should be noted that the method in this case can also be used for the prediction of the temperature field of other fluids with heat sources.
[0046] Through the description of the above embodiments, those skilled in the art can clearly understand that according to the method of the present disclosure, the corresponding system can be implemented. Exemplarily, a vehicle cabin temperature field prediction system based on a neural network, the system includes a simulation module, a training module, and a prediction module; the simulation module is configured to perform finite element simulation of a three-dimensional model based on the vehicle cabin, set disturbance factors affecting the vehicle cabin temperature and vehicle cabin wall material parameters during the simulation, and at the same time set the external flow field area, and obtain a temperature value sequence corresponding to n uniformly distributed coordinate points as multi-dimensional data through finite element simulation; the training module is configured to train a CNN-LSTM neural network based on the multi-dimensional data; the prediction module is configured to input the actual vehicle cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values of each monitoring point in the vehicle cabin temperature field for a period of time in the future.
[0047] Through the description of the above embodiments, those skilled in the art can clearly understand that the method or system of the present disclosure can be implemented by means of software plus necessary general hardware. Of course, it can also be implemented by dedicated hardware including application-specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures used to implement the same function can also be various, such as analog circuits, digital circuits or dedicated circuits. However, in more cases for the present disclosure, software program implementation is a better implementation manner.
[0048] Although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific embodiments and application fields. The above specific embodiments are merely illustrative and guiding, rather than restrictive. Those of ordinary skill in the art can also make many forms under the inspiration of this specification and without departing from the scope protected by the claims of the present invention, and these all belong to the scope of protection of the present invention.
Claims
1. A method for predicting the temperature field in a vehicle cabin based on a neural network, characterized in that the steps Including: Performing finite element simulation of a three-dimensional model based on the vehicle cabin. In the simulation, disturbance factors affecting the cabin temperature and material parameters of the cabin wall are set. At the same time, an external flow field region is set, and a temperature value sequence corresponding to n uniformly distributed coordinate points is obtained through finite element simulation as multi-dimensional data; Training a CNN-LSTM neural network based on the multi-dimensional data; Inputting the actual cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values at each monitoring point in the cabin temperature field for a future period of time.
2. The method according to claim 1, wherein The disturbance factors include driving speed and solar radiation.
3. The method according to claim 1, wherein The material parameters include the density, specific heat, and thermal conductivity of four materials: aluminum, tempered glass, and high-density polyethylene. The material parameters also include the density, specific heat, thermal conductivity, and viscosity of air.
4. The method according to claim 1, wherein The external flow field region is the spatial region from a set distance outside the outer wall of the vehicle cabin to the outer wall of the vehicle cabin.
5. A vehicle cabin temperature field prediction system based on a neural network, characterized in that, The system includes a simulation module, a training module, and a prediction module; The simulation module is configured to perform finite element simulation of a three-dimensional model based on the vehicle cabin. In the simulation, disturbance factors affecting the cabin temperature and material parameters of the cabin wall are set. At the same time, an external flow field region is set, and a temperature value sequence corresponding to n uniformly distributed coordinate points is obtained through finite element simulation as multi-dimensional data; The training module is configured to train a CNN-LSTM neural network based on the multi-dimensional data; The prediction module is configured to input the actual cabin temperature sequence into the trained CNN-LSTM neural network to predict the temperature values at each monitoring point in the cabin temperature field for a future period of time.
6. The system according to claim 1, wherein The disturbance factors include driving speed and solar radiation.
7. The system according to claim 1, characterized in that The material parameters include the density, specific heat, and thermal conductivity of four materials: aluminum, tempered glass, and high-density polyethylene. The material parameters also include the density, specific heat, thermal conductivity, and viscosity of air.
8. The system according to claim 1, wherein The external flow field region is the spatial region from a set distance outside the outer wall of the vehicle cabin to the outer wall of the vehicle cabin.
9. A computer-readable storage medium, characterized in that: There is a computer program stored that can be loaded and executed by a processor to perform any one of the methods as claimed in claims 1 to 4.