Vehicle thermal management method and device, electronic equipment and vehicle

Abstract

The application discloses a vehicle thermal management method and device, electronic equipment and vehicle, and belongs to the technical field of vehicle thermal management control. The method comprises the following steps: obtaining a glass area image and thermal management related information of a vehicle in a preset time period, predicting the heat change trend of different areas of the vehicle in a future time period by comprehensively utilizing images and information from multiple sources, and obtaining the heat change trend of the heat generating area in advance to facilitate early response. Based on the predicted heat change trend of different areas of the vehicle in the future time period and in combination with the thermal management related information of the vehicle at the current time point, a strategy that matches the heat change trend of different areas and the thermal management related information at the current time point, i.e. a control instruction for thermal management of different areas, is determined, and the vehicle is subjected to thermal management based on the control instruction to adapt to the current heat change, thereby realizing rapid response of thermal management.

Description

Technical Field

[0001] This application belongs to the field of vehicle thermal management control technology, specifically relating to a vehicle thermal management method, device, electronic equipment, and vehicle. Background Technology

[0002] In high-temperature environments, vehicles are exposed to intense solar radiation for extended periods, with the glass area becoming a primary heat intrusion channel. This causes a rapid rise in interior temperature and a surge in battery heat load, severely impacting passenger comfort and powertrain safety. Existing thermal management technologies primarily employ a separate control strategy, where the intelligent glass dimming and battery cooling systems operate independently. This approach fails to consider the interplay between multiple heat-generating areas within the vehicle or to implement proactive countermeasures, resulting in delayed thermal management responses and an inability to adapt to rapidly changing environmental conditions. Summary of the Invention

[0003] In view of the above problems, embodiments of this application provide a vehicle thermal management method, apparatus, electronic device, and vehicle that overcomes or at least partially solves the above problems. The technical solution is as follows: In a first aspect, embodiments of this application provide a vehicle thermal management method, the method comprising: acquiring images of the glass area of ​​a vehicle over a preset time period and thermal management-related information; wherein the thermal management-related information includes at least one of the following: environmental state information of the vehicle, operating state information of the vehicle, and battery state information of the vehicle; predicting the thermal change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area images and the thermal management-related information; determining control commands for thermal management of the different areas based on the thermal change trends of the vehicle in different areas over a future time period and the thermal management-related information of the vehicle at the current time point; and performing thermal management on the vehicle based on the control commands.

[0004] Optionally, after acquiring images of the vehicle's glass area over a preset time period, the method further includes: processing the glass area images using a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass area images. Through this embodiment, the ViT model is used to analyze thermal features such as heat distribution and accumulation paths in the glass area images, providing accurate analytical basis for subsequent prediction of thermal trend changes.

[0005] Optionally, the glass region image is processed using a pre-trained Vision Transformer (ViT) model to obtain the heat features corresponding to the glass region image. This includes: aligning and enhancing the glass region image in both time and spatial dimensions; dividing the aligned and enhanced glass region image into a predetermined number of image blocks; encoding the image blocks and their corresponding positions to obtain an embedding matrix corresponding to the glass region image; and extracting features from the embedding matrix based on a multi-head self-attention mechanism to obtain the heat features corresponding to each element in the embedding matrix. Through this embodiment, the glass region image is first preprocessed, and then the embedding matrix obtained by combining image blocks with position encoding is used for feature extraction via a multi-head self-attention mechanism to finely analyze the heat features corresponding to the glass region image, providing an effective predictive basis for subsequent heat trend prediction.

[0006] Optionally, based on the thermal features corresponding to the glass area image and the thermal management related information, the thermal change trend of the vehicle in different areas over a future time period is predicted. This includes: constructing a preset number of multi-dimensional time series based on the time of the thermal features and the thermal management related information according to a preset time length; wherein the multi-dimensional time series includes the heat of the vehicle in different areas; inputting the preset number of multi-dimensional time series into a pre-trained gated recurrent unit (GRU) model, extracting the thermal differences between adjacent multi-dimensional time series, and predicting the thermal change trend of the vehicle in different areas over a future time period based on the thermal differences. Through the embodiments of this application, the thermal differences between multiple time series are analyzed, and then the thermal change trend of the vehicle in different areas over a future time period is predicted based on the thermal differences, accurately capturing the thermal change trend of the vehicle in different areas, providing decision-making prior support for timely thermal management response, and ensuring the timeliness of thermal management.

[0007] Optionally, based on the heat change trend of the vehicle in different areas over a future time period and the vehicle's thermal management information at the current time, control instructions for thermal management in the different areas are determined. This includes: inputting the heat change trend and the vehicle's thermal management information at the current time into a pre-trained Extreme Gradient Boosting (XGBoost) model; extracting the correlation between the vehicle's thermal management information at the current time and the heat change trend; determining the importance of the vehicle's thermal management information at the current time based on the correlation; and determining control instructions for thermal management in the different areas based on the importance of the vehicle's thermal management information at the current time and a preset control behavior pair. The preset control behavior pair includes the thermal management information and the control instructions corresponding to the thermal management information. Through this embodiment, the obtained heat change trend and real-time thermal management information are converted into control instructions, and then thermal management of the vehicle is performed based on these control instructions, achieving complete thermal management from image recognition and environmental perception to the determination and execution of control instructions.

[0008] Optionally, the control command includes setting operating parameters for thermal management of the vehicle; the method further includes: performing thermal management of the vehicle based on a preset strategy when the operating parameters exceed a preset threshold. Through this embodiment, based on determined operating parameters, if the operating parameters exceed a preset threshold, it indicates that the currently determined control command is abnormal and unsuitable for the current situation. A preset strategy can then be used to perform thermal management of the vehicle, avoiding complete reliance on the aforementioned strategy and providing an alternative strategy for vehicle thermal management.

[0009] Secondly, embodiments of this application provide a vehicle thermal management device, comprising: an acquisition module for acquiring images of the glass area of ​​a vehicle and thermal management-related information over a preset time period; wherein the thermal management-related information includes at least one of the following: environmental state information of the vehicle, operating state information of the vehicle, and battery state information of the vehicle; a prediction module for predicting the thermal change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area images and the thermal management-related information; a determination module for determining control commands for thermal management of the different areas based on the thermal change trends of the vehicle in different areas over a future time period and the thermal management-related information of the vehicle at the current time point; and a thermal management module for performing thermal management on the vehicle based on the control commands.

[0010] Thirdly, embodiments of this application provide an electronic device, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the vehicle thermal management method described in the first aspect.

[0011] Fourthly, embodiments of this application provide a vehicle, the vehicle including at least one processor, which, when executed, implements the vehicle thermal management method described in the first aspect above.

[0012] Fifthly, embodiments of this application provide a computer-readable storage medium storing a program or instructions that, when executed by a processor, implement the vehicle thermal management method described in the first aspect.

[0013] By employing the above technical solution, this application provides a vehicle thermal management method that acquires images of the vehicle's glass area and thermal management-related information over a preset time period. This thermal management-related information includes at least one of the following: the vehicle's environmental state information, the vehicle's operating state information, and the vehicle's battery state information. Based on the thermal characteristics corresponding to the glass area images and the thermal management-related information, the method analyzes the thermal change trends of the vehicle in different areas over a future time period. Based on these future thermal change trends and the vehicle's thermal management-related information at the current time point, it determines control commands for thermal management of the different areas and performs thermal management based on these control commands. This method integrates images and information from multiple sources to predict the thermal change trends of the vehicle in different areas over a future time period, allowing for advance awareness of the thermal change trends of heat-generating areas and facilitating proactive responses. The thermal characteristics and thermal management-related information from multiple sources provide a more comprehensive analytical basis for prediction, avoiding the inability to capture the thermal influence between different heat-generating areas due to single-source information, which would lead to inaccurate predictions. Based on the predicted heat change trends of vehicles in different regions over future time periods and combined with the vehicle's thermal management information at the current time point, a strategy matching the heat change trends of different regions and the thermal management information at the current time point is determined. This strategy is then used to implement thermal management control commands for different regions. Based on these control commands, the vehicle's thermal management is implemented to adapt to the current heat changes and achieve rapid response in thermal management.

[0014] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic flowchart of a vehicle thermal management method provided in an embodiment of this application is shown; Figure 2 A schematic flowchart of another vehicle thermal management method provided in an embodiment of this application is shown; Figure 3 This illustration shows a structural schematic diagram of a vehicle thermal management device according to an embodiment of this application; Figure 4 This illustration shows a structural schematic diagram of an electronic device provided in an embodiment of this application; Figure 5 A schematic diagram of the structure of a vehicle provided in an embodiment of this application is shown. Detailed Implementation

[0016] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0017] In existing technologies, vehicles experience rapid temperature increases in multiple areas due to prolonged exposure to strong solar radiation, such as the glass area, the interior environment, and the battery. To address the adverse effects of high temperatures on occupant comfort and powertrain safety, current thermal management technologies include smart glass dimming and battery cooling. Smart glass dimming involves processing vehicle driving data using a neural network model to obtain adjustment parameters for the windows, and then adjusting the windows based on these parameters. Battery cooling involves generating a set of control strategies based on the acquired ambient temperature inside and outside the vehicle and the battery charge to control the thermal management loop, thereby achieving thermal management. However, these methods primarily employ a separate control strategy, meaning the smart glass dimming and battery cooling systems operate independently. This lack of consideration for the interrelationships between multiple heat-generating areas of the vehicle and the absence of proactive strategies leads to delayed thermal management response, making it difficult to adapt to rapidly changing environmental conditions.

[0018] Based on the above application scenarios, in order to solve the technical problem that the independent operation of the smart glass dimming and battery cooling systems, without considering the influence between multiple heat-generating areas of the vehicle and the lack of proactive strategies, leads to a delayed thermal management response and difficulty in adapting to rapidly changing environmental demands, this application provides a vehicle thermal management method, device, electronic device, and vehicle. Figure 1 As shown, Figure 1 This illustration shows a flowchart of a vehicle thermal management method according to an embodiment of this application. The method can be executed by an electronic device, which may include a vehicle. See also... Figure 1 The method may include the following steps.

[0019] S110, acquire images of the vehicle's glass area and thermal management-related information during a preset time period; wherein, the thermal management-related information includes at least one of the following: the vehicle's environmental status information, the vehicle's operating status information, and the vehicle's battery status information.

[0020] The vehicle is equipped with a sensor system to comprehensively collect data on changes that occur when the vehicle is in a multi-dimensional thermal environment during actual driving. The data collection scope can include: the composition of the vehicle's thermal load, heat conduction paths, control response behavior, and the impact of occupant usage behavior on the heat distribution inside the vehicle. The preset time period is a historical period prior to the current time point.

[0021] The aforementioned images of the glass areas can be acquired using an image-level thermal sensing system. This system consists of a three-primary-color (Red, Green, Blue, RGB) camera and an infrared thermal imager, installed on the roof of the vehicle, the inside of the A-pillar, and the front edge of the dashboard. It primarily acquires thermal radiation images and thermal distribution imaging results of glass areas such as the windshield, side windows, and sunroof under different solar angles. The acquisition frequency of the glass area images can be set to 10 frames per second, with image resolutions of 1080p RGB and 640x480 infrared images, and timestamps are added for synchronized recording.

[0022] The aforementioned thermal management-related information can be collected through an environmental-grade thermal load input system and an in-cabin thermal status acquisition system. The environmental-grade thermal load input system includes a solar radiation intensity sensor (spectral range covering 280nm to 2500nm), an ambient temperature or humidity sensor, a wind speed sensor, and a Global Positioning System (GPS) positioning unit. This unit records external radiation levels, convective environmental parameters, and the vehicle's altitude and latitude, providing regional climatic characteristics. The in-cabin thermal status acquisition system collects data through temperature and humidity sensor arrays located in the driver's seat, front passenger seat, and rear center, with temperature patches placed inside the glass area to assist in collecting information on heat transfer generated after heat absorption on the glass surface. This part of the system samples five times per minute and records the heat conduction gradient as the vehicle's speed changes, used to analyze the mapping relationship between glass insulation performance and the increase in perceived temperature for occupants.

[0023] Among the aforementioned thermal management-related information, the vehicle's battery status information can be collected through a thermocouple matrix deployed in the vehicle's battery area and integrated with the Battery Management System (BMS) bus data integration module. Each cell in the battery area has multiple thermocouples, with a sampling frequency of 1Hz, used to record the temperature difference distribution in the battery's heat concentration areas and along the cooling path. This data, combined with real-time operating status information reported by the BMS, such as cooling pump flow rate, average cell temperature, and coolant inlet / outlet temperature difference, forms the battery area's thermal response curve. Furthermore, to further capture the thermal coupling relationship between the glass area and the battery area, key cooling system operating parameters, such as refrigerant system refrigeration pressure, expansion valve opening, and evaporator outlet temperature, are also collected.

[0024] In addition, in order to understand the impact of occupant usage behavior on the heat distribution inside the vehicle, this application embodiment also collects relevant information on occupant usage behavior, using the vehicle's inertial measurement unit (IMU) and controller to obtain information such as vehicle acceleration, steering wheel angle, brake or accelerator opening, and using the vehicle's cameras to obtain personalized information such as the number of occupants, occupant seating position, and occupant posture.

[0025] It should be noted that the aforementioned thermal management-related information needs to be synchronized in chronological order and timestamped to construct a complete information set of vehicle overheating events. Furthermore, this thermal management-related information needs to be collected multiple times under different solar azimuth angles, vehicle orientations, window opening states, and glass tint control voltages to provide a rich information set for subsequent processing.

[0026] In an optional embodiment, after acquiring images of the vehicle's glass area and thermal management-related information over a preset time period, the method may further include: data preprocessing. This includes: checking the glass area images for anomalies such as defocusing, exposure, and reflection, and filtering out erroneous frames and abnormal specular reflection frames. Drift, power outages, distortions, and abrupt changes in the data measured by the sensor system are repaired using a combination of bilateral filtering and low-pass filtering. The acquired glass area images and thermal management-related information may also be classified and encoded according to different classification standards, such as based on different vehicle models, different control conditions, and different solar irradiance.

[0027] S120, based on the thermal characteristics corresponding to the glass area image and the thermal management related information, predict the thermal change trend of the vehicle in different areas over a future time period.

[0028] Specifically, the heat characteristics and thermal management-related information mentioned above are normalized and outlier truncation and cleaning processes are performed to ensure the quality of the data used for prediction. Prediction methods can include using trend prediction models, which can be neural network models with time-series memory capabilities to accurately capture the heat change trends in different areas of the vehicle, such as Gated Recurrent Unit (GRU) models and Recurrent Neural Network (RNN) models.

[0029] The aforementioned trends in vehicle temperature changes in different areas over a future time period may include: trends in temperature changes in glass areas such as the windshield, side windows, and sunroof over the next 10 minutes; trends in battery electrolyte temperature; and trends in the highest temperature of the battery cells.

[0030] S130, based on the heat change trend of the vehicle in different areas in the future time period and the thermal management information of the vehicle at the current time point, determine the control command for thermal management of the different areas.

[0031] The aforementioned control commands for thermal management of different areas may include: setting the light transmittance of the slewable glass of the Suspended Particle Device (SPD), setting the rotation speed of the battery cooling pump, setting the power of the cooling fan, and setting the flow rate of the evaporator in the air conditioning system. For example, the SPD voltage is set to 0.1V to control the optical properties of the glass, the cooling pump adjusts the flow output at a speed of 200 rpm, and the fan power is mapped to the 0-100% output range using a pulse-width modulation (PWM) signal.

[0032] S140, thermal management of the vehicle is performed based on the control command.

[0033] In summary, in this embodiment, by acquiring images of the vehicle's glass area and thermal management-related information over a preset time period, including at least one of the following: the vehicle's environmental state information, the vehicle's operating state information, and the vehicle's battery state information, the thermal characteristics corresponding to the glass area images and the thermal management-related information are used to predict the thermal change trend of the vehicle in different areas over a future time period. Based on the thermal change trend of the vehicle in different areas over a future time period and the thermal management-related information of the vehicle at the current time point, control commands for thermal management of the different areas are determined, and thermal management of the vehicle is performed based on the control commands. This integrates images and information from multiple sources to predict the thermal change trend of the vehicle in different areas over a future time period, allowing for advance awareness of the thermal change trend of heat-generating areas and enabling proactive responses. The thermal characteristics and thermal management-related information from multiple sources provide a more comprehensive analytical basis for prediction, avoiding the inability to capture the thermal influence between different heat-generating areas due to single information, thus preventing inaccurate predictions. Based on the predicted heat change trends of vehicles in different regions over future time periods and combined with the vehicle's thermal management information at the current time point, a strategy matching the heat change trends of different regions and the thermal management information at the current time point is determined. This strategy is then used to implement thermal management control commands for different regions. Based on these control commands, the vehicle's thermal management is implemented to adapt to the current heat changes and achieve rapid response in thermal management.

[0034] When a vehicle is exposed to intense solar radiation in a hot desert environment, the window glass area becomes the main entry point for the vehicle's heat load, especially around noon when direct sunlight on the windshield and side windows can cause a rapid imbalance in heat accumulation inside the cabin. Therefore, an image processing model with strong region perception capabilities is needed to spatially segment the heat radiation received by the glass surface, extract the location, intensity, and diffusion trend of high-heat accumulation areas, and transmit these thermal features in a structured form to the stage predicting changes in heat trend. Traditional convolutional networks tend to fail in high-contrast or light-spot interference areas of the image. Therefore, in this embodiment, a Vision Transformer (ViT) model is used as the image perception backbone network, utilizing its global attention mechanism to process the complex radiation patterns in the acquired glass area images. In a specific embodiment, after acquiring the vehicle's glass area images over a preset time period in S110, the method further includes: S112, processing the glass area images using a pre-trained ViT model to obtain the corresponding thermal features of the glass area images.

[0035] In this embodiment, the glass area image input to the ViT model originates from infrared thermal images and visible light images obtained by cameras at the front and sides of the vehicle. The ViT model is used to preprocess the glass area image before analyzing the thermal characteristics such as heat distribution and accumulation path in the image. Through the thermal characteristics, the heat, heat absorption tendency, and radiation absorption capacity of different spatial locations in the glass area can be obtained. The thermal characteristics of the glass area can be visualized as a heat map to intuitively understand the thermal characteristics of the glass area, detect the heat accumulation and heat insulation performance of the glass area, and provide accurate analytical basis for subsequent prediction of heat trend changes.

[0036] To more clearly illustrate the technical solution provided in the embodiments of this application, the following detailed explanation is given of the above-mentioned process of processing the glass region image through a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass region image, as shown in the following steps.

[0037] S1121, The glass region image is aligned and enhanced in both the time and spatial dimensions.

[0038] The glass region image includes both infrared thermal and visible light images. These images are aligned and enhanced in both temporal and spatial dimensions to perform multidimensional analysis of the same region using both infrared and visible light. Alignment and enhancement methods may include image interpolation resampling, edge region weighting enhancement, and solar reflection area masking. Each image frame is normalized to a resolution of 224 pixels by 224 pixels, expanding the channel dimension of image alignment and enhancement to four dimensions, where infrared thermal... Figure 1 Dimensions, RGB Figure 3 The dimensions form a multi-channel fusion input, thereby synchronously processing the aforementioned glass region image and improving the efficiency of image alignment and enhancement.

[0039] S1122, the aligned and enhanced glass region image is divided to obtain a preset number of image blocks.

[0040] The method for dividing the aligned and enhanced glass region image can be patch division, which divides the entire glass region image into a preset number of image blocks. Each image block includes some features of the glass region image, enabling fine analysis of the thermal features at different locations in the glass region image.

[0041] S1123, Encode the image block and the position corresponding to the image block to obtain the embedding matrix corresponding to the glass region image.

[0042] In this process, image patches are encoded with their corresponding locations to determine the position of each image patch in the original glass region image. This enables the ViT model to understand the spatial relationships between image patches and thus determine the thermal characteristics of different locations in the glass region image by analyzing the thermal characteristics of each image patch.

[0043] S1124, Based on the multi-head self-attention mechanism, feature extraction is performed on the embedding matrix to obtain the heat feature corresponding to each element in the embedding matrix.

[0044] The ViT model comprises a multi-layer Transformer encoder. Each Transformer encoder's multi-head self-attention mechanism focuses on each element in the embedding matrix, with each element corresponding to an image patch. It captures the relationship between the heat of each element and the overall embedding matrix under high-intensity radiation, the heat aggregation trend from edge elements to center elements, the heat aggregation path, and whether each element can adapt to the hotspot shift caused by changes in the solar angle. By combining these analyses, the heat characteristics of each element are determined, constituting the heat characteristics corresponding to the glass region image. These heat characteristics are then stitched together over time to form a time series, serving as the data basis for subsequent trend prediction. This allows subsequent trend prediction to efficiently predict heat trends based on the acquired glass region heat characteristics without processing the original glass region image, reducing prediction complexity and computational resource consumption.

[0045] In an optional embodiment, training the ViT model may include: acquiring images of glass areas under various solar angles, weather types, and electronically controlled transmittance levels, etc., as training samples to enable the ViT model to recognize thermal features under different lighting conditions; using temperature obtained from infrared thermal imager calibration as a supervisory label; fitting the true thermal values ​​of each region of the image through regression; and further combining hybrid enhancement strategies such as illumination perturbation, occlusion simulation, and glass reflection simulation to train the ViT model, thereby expanding its ability to process images in real-world complex environments. In addition, the ViT model also has over-the-air (OTA) upgrade capabilities, enabling it to periodically acquire new training samples, complete incremental training on the backend, and update model parameters via a wireless network.

[0046] In another optional embodiment, the loading mechanism of the ViT model can be set to a hot-start mode, that is, the ViT model is automatically started to acquire the thermal features of the image when the ambient temperature is higher than a preset temperature threshold or the external radiation intensity exceeds the safety limit. Furthermore, when the ambient temperature is not higher than the preset temperature threshold or the external radiation intensity does not exceed the safety limit, the ViT model enters a low-power sleep state, thereby reducing the energy consumption of the ViT model.

[0047] In this embodiment, the thermal features corresponding to the glass region image can be obtained using the ViT model. First, the glass region image is preprocessed by aligning and enhancing it in both the temporal and spatial dimensions. The aligned and enhanced image is then divided into a predetermined number of image blocks. Each image block and its corresponding position is encoded to obtain an embedding matrix corresponding to the glass region image. Next, feature extraction is performed on the embedding matrix based on a multi-head self-attention mechanism to obtain the thermal features corresponding to each element. This yields the thermal features corresponding to the glass region image. These thermal features not only reveal the heat, heat absorption tendency, and radiation absorption capacity at different spatial locations within the glass region but also provide an effective basis for subsequent thermal trend prediction.

[0048] In high-temperature, high-radiation environments, heat is conducted into the cabin through the vehicle's glass areas, further affecting occupant comfort and the thermal balance of the battery system. Since heat propagation within the cabin exhibits distinct temporal characteristics—such as the increase in heat input due to changes in solar angle, the hysteresis caused by thermal inertia, and thermal disturbances caused by occupant dynamic behavior—this embodiment of the application can utilize a neural network model with temporal memory capabilities to accurately analyze the thermal variation trends of the vehicle in different areas. The neural network model can include a GRU model or an RNN model. Compared to an RNN model, a GRU has a stronger ability to retain information flow, maintaining the memory of key features even when facing slow increases or sudden decreases in thermal fluctuations. Therefore, in a specific embodiment, the aforementioned S120, based on the thermal features corresponding to the glass area image and the thermal management-related information, predicts the thermal variation trends of the vehicle in different areas over a future time period, and may include the following steps.

[0049] S1201, based on the time of the heat characteristics and the thermal management related information, construct a preset multidimensional time series of the heat characteristics and the thermal management related information according to a preset time length; wherein, the multidimensional time series includes the heat of the vehicle in different regions.

[0050] The GRU model processes the thermal features corresponding to the glass area image rather than the original image. This allows the GRU model to focus on the thermal features of the glass area without expending computational resources on image processing, thus reducing the input complexity and computational resource consumption of the trend prediction model. The preset time length can be set to 30 minutes, with sampling once per minute, resulting in 30 multi-dimensional time series. Each multi-dimensional time series can include thermal features of glass areas such as the windshield, side windows, and sunroof; environmental and operational status information such as ambient temperature, humidity, solar radiation intensity, wind speed, vehicle speed, window status, and SPD glass voltage setting; and battery status information such as battery fluid temperature difference, cell temperature, and current flow rate. Optionally, before constructing the multi-dimensional time series, normalization and outlier truncation / cleaning are performed on the aforementioned thermal features and thermal management-related information to ensure the quality of the data used for prediction.

[0051] S1202, input the preset multi-dimensional time series into the pre-trained GRU model, extract the heat difference of the multi-dimensional time series of adjacent times, and predict the heat change trend of the vehicle in different regions in the future time period based on the heat difference.

[0052] The GRU model includes a two-layer recursive structure, with each layer containing multiple gating units. Each gating unit captures the heat differences of multi-dimensional time series of adjacent times through reset and update gates and passes them to the next layer. The next layer captures the long-term trend of heat changes based on the received multiple heat differences, thereby predicting the heat change trend of vehicles in different regions in the future.

[0053] In an optional embodiment, training the GRU model may include: collecting historical vehicle data as supervised training samples, where target labels are generated from measured temperature sequences recorded by thermocouples and thermal imagers, ensuring the output results have physical interpretability and engineering usability. The Adam optimizer is used to update model parameters, with a batch size of 128 training samples and more than 300 training epochs. An early stopping mechanism is incorporated to automatically terminate the model when there is no improvement, determining the generalization ability of the GRU model. Furthermore, to improve the responsiveness of the GRU model in highly dynamic environments, a Bayesian optimization strategy is introduced when updating model parameters to automatically select the optimal number of neurons in the hidden layer, the learning rate, and the dropout ratio. Through this training, the GRU model can anticipate increases in radiation and accurately predict future temperature change trends, thereby determining control commands in advance, executing glass dimming and battery cooling, and improving the response speed of thermal management.

[0054] In this embodiment, the prediction of the vehicle's heat change trend in different areas over a future time period is achieved using a GRU model, based on the thermal features corresponding to the glass area image and the thermal management-related information. First, the thermal features and thermal management-related information are constructed into a preset number of multi-dimensional time series according to a preset time length. Then, these preset number of multi-dimensional time series are input into a pre-trained GRU model to extract the heat differences between adjacent multi-dimensional time series. The analysis of these heat differences reveals the long-term change trend of the vehicle in different areas, providing a reliable basis for prediction. Based on these heat differences, the prediction of the vehicle's heat change trend in different areas over a future time period accurately captures the vehicle's heat change trend in different areas, providing prior decision support for timely thermal management responses and ensuring the timeliness of thermal management.

[0055] In high-temperature, high-radiation environments, vehicle thermal management is achieved through the coordinated adjustment of multiple subsystems, including the transmittance control of dimmable glass, the power allocation of cooling pumps and fans in the battery cooling system, and the optimization of refrigerant flow in the air conditioning system. To achieve this multi-objective, multi-variable, and multi-time-sequence response intelligent control process, in this embodiment, an Extreme Gradient Boosting (XGBoost) model based on a gradient boosting decision tree structure is selected as the coordinated control decision engine. The input integrates the heat change trend output by the aforementioned GRU model with real-time acquired thermal management information, and the output can generate executable control commands to realize dynamic adjustment strategies. Therefore, in a specific embodiment, S130, based on the heat change trend of the vehicle in different regions over a future time period and the vehicle's thermal management information at the current time point, determines the control commands for thermal management in the different regions, which may include the following steps.

[0056] S1301, input the heat change trend and the vehicle's thermal management information at the current time point into the pre-trained extreme gradient XGBoost model, extract the correlation between the vehicle's thermal management information at the current time point and the heat change trend, and determine the importance of the vehicle's thermal management information at the current time point based on the correlation.

[0057] The heat change trend can include: the heat change trend of each glass area in the next 10 minutes, the rate of change of battery coolant outlet temperature, and the expected inflection point of the highest cell temperature. Vehicle thermal management information at the current time point can include: current solar radiation intensity, ambient temperature and humidity, wind speed, vehicle speed, vehicle orientation, glass angle, and current SPD glass voltage setting, among other dynamic operating condition signals. Furthermore, to account for the impact of occupant behavior on in-vehicle heat distribution, occupant status and personalized thermal comfort index parameters can be acquired, including the number of occupants identified by the cockpit camera, occupant body characteristics, thermal sensitivity levels, and historical cooling control parameters, as input data for the XGBoost model to assist in determining control commands.

[0058] The XGBoost model uses feature cross-analysis to capture the correlation between the vehicle's thermal management information and the heat change trend at the current time. For example, when the sun angle is low and the glass angle is steep, the heat change trend under the same radiation intensity shows a significant increase. The XGBoost model captures the correlation between the sun angle and the glass angle. Based on the feature importance ranking mechanism of the XGBoost model, it determines the importance of the vehicle's thermal management information at the current time, thereby determining the degree of influence of the thermal management information on the heat at the current time. It prioritizes thermal management of the heat-generating areas corresponding to the thermal management information with higher importance, and shields the influence of noise information on control commands.

[0059] S1302, based on the importance of the vehicle's thermal management-related information at the current time point and the preset control behavior pair, determine the control instructions for thermal management of the different areas; wherein, the preset control behavior pair includes the thermal management-related information and the control instructions corresponding to the thermal management-related information.

[0060] In an optional embodiment, training the XGBoost model may include: acquiring thermal management-related information and corresponding control commands, constructing control behavior pairs as training samples, and achieving multi-objective joint learning through a constructed composite loss function. The training samples also include battery temperature rise suppression effects, glass temperature change rates, occupant comfort index scores, etc., and to enhance the XGBoost model's ability to cope with abnormal scenarios, thermal mutation segment data samples, long-time lag data samples, and extreme control failure data are also introduced. During the model parameter update process, a grid search method is used to optimize core hyperparameters in multiple rounds, such as the number of trees, maximum depth, learning rate, subsampling rate, and column sampling ratio. In the early stages of training, when the maximum tree depth is controlled at 7 layers, the learning rate is set to 0.03, and the subsampling ratio is set to 0.8, the XGBoost model achieves the optimal trade-off between the accuracy of dynamic thermal response, the stability of control output, and the ability to maintain occupant comfort. In the later stages of training, the weights of the training samples are iteratively updated by combining K-fold cross-validation with feedback from actual vehicle simulations, and new training samples are introduced periodically to expand the applicability of the model.

[0061] In this embodiment, based on the heat change trend of the vehicle in different regions over a future time period and the vehicle's thermal management information at the current time, it is determined that the control commands for thermal management in the different regions can be implemented using an XGBoost model. First, the heat change trend and the vehicle's thermal management information at the current time are input into a pre-trained extreme gradient boosting XGBoost model. The correlation between the vehicle's thermal management information at the current time and the heat change trend is extracted, and the importance of the vehicle's thermal management information at the current time is determined based on this correlation. Then, based on the importance of the vehicle's thermal management information at the current time and preset control behavior pairs, control commands for thermal management in the different regions are determined. The obtained heat change trend and real-time thermal management information are converted into control commands, and then thermal management of the vehicle is performed based on these control commands, achieving complete thermal management from image recognition and environmental perception to the determination and execution of control commands.

[0062] To address potential anomalies or safety hazards at each stage, for example, when predicting the heat change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area image and the thermal management related information, prediction errors may occur. Another example is when determining control commands for thermal management of different areas based on the heat change trend of the vehicle in different areas over a future time period and the vehicle's thermal management related information at the current time, the control commands may be inapplicable to the current situation. Yet another example is when the vehicle is in a relatively simple environment, eliminating the need for the aforementioned complex strategies for thermal management. This application combines the strategies for predicting the heat change trend of the vehicle in different areas over a future time period and determining control commands with existing thermal management strategies. Therefore, in one specific embodiment, the above-mentioned vehicle thermal management method may further include: performing thermal management of the vehicle based on a preset strategy when the operating parameters exceed a preset threshold.

[0063] The preset strategy can be either the vehicle's existing or default thermal management strategy.

[0064] In this embodiment, the control command may include setting operating parameters for the vehicle's thermal management, such as setting the light transmittance of the swivel glass of the Suspended Particle Device (SPD), setting the rotation speed of the battery cooling pump, setting the power of the cooling fan, and setting the flow rate of the evaporator in the air conditioning system. Based on the determined operating parameters, if these parameters exceed a preset threshold, it indicates that the currently determined control command is abnormal and unsuitable for the current situation. A preset strategy can then be used for thermal management of the vehicle, avoiding complete reliance on the aforementioned strategy and providing alternative strategies for vehicle thermal management.

[0065] In an optional embodiment, the above method may further include: after performing thermal management on the vehicle, recording the difference between the result of the vehicle thermal management and the actual changes in the environment, and using the difference as a data source for subsequent fine-tuning and training of each model, thereby further improving the model's capabilities.

[0066] Figure 2 This illustration shows a flowchart of another vehicle thermal management method provided in an embodiment of this application. This method can be executed by an electronic device, which can be a vehicle. See also... Figure 2 The method may include the following steps.

[0067] S201, acquire images of the vehicle's glass area and thermal management-related information over a historical time period.

[0068] Images of the glass area can be acquired using an image-level thermal sensing system consisting of an RGB camera and an infrared thermal imager. Thermal management-related information includes vehicle environmental status, vehicle operating status, and vehicle battery status, which can be collected through a sensor system. This system may include solar radiation intensity sensors (spectral range covering 280nm to 2500nm), ambient temperature / humidity sensors, wind speed sensors, and GPS positioning units. It may also include temperature and humidity sensor arrays located in the driver's seat, passenger seat, and rear center. Furthermore, a thermocouple matrix and BMS bus data integration module should be strategically deployed in the battery area.

[0069] S202 uses the ViT model to perceive the glass region image and obtain the corresponding thermal features of the glass region image.

[0070] The glass region image includes infrared thermal images and visible light images acquired by cameras at the front and sides of the vehicle. The glass region image is aligned and enhanced in both time and space dimensions using the ViT model. The aligned and enhanced glass region image is then divided into a preset number of image blocks. The image blocks and their corresponding positions are encoded to obtain the embedding matrix corresponding to the glass region image. Feature extraction is performed on the embedding matrix based on a multi-head self-attention mechanism to obtain the thermal features corresponding to each element in the embedding matrix, thus obtaining the thermal features corresponding to the glass region image.

[0071] S203 uses the GRU model to predict the thermal characteristics and thermal management information corresponding to the glass area image, and obtains the thermal change trend of different areas of the vehicle in the future time period.

[0072] The aforementioned heat change trends of the vehicle in different areas over a future time period may include: heat change trends of glass areas such as the windshield, side windows, and sunroof within the next 10 minutes; heat change trends of battery electrolyte temperature; and heat change trends of the highest cell temperature, providing a basis for determining subsequent control strategies.

[0073] S204, based on the heat change trend predicted by the GRU model and real-time thermal management related information, uses the XGBoost model to determine the control commands for thermal management of different areas of the vehicle.

[0074] This application's embodiments combine the ViT model, GRU model, and XGBoost model to construct a three-layer fusion structure consisting of a visual perception model, a trend prediction model, and a decision-making model. This enables vehicles to perceive the environment, predict thermal change trends, and determine control strategies in high-temperature and high-radiation environments, with each model having a clearly defined role. Furthermore, through specific data formats, information transmission mechanisms, and joint methods, the continuity and consistency of information flow among the multiple models are established. The thermal features corresponding to the glass area image identified by the ViT model are seamlessly transmitted to the GRU model, and the prediction results of the GRU model are then output to the XGBoost model to determine control commands, forming a closed-loop thermal management system encompassing environmental perception, thermal trend prediction, and control command output.

[0075] In the three-layer fusion structure consisting of a visual perception model, a trend prediction model, and a decision model, the visual perception model (ViT model) is responsible for glass region image recognition. Its output is a thermal attention matrix, representing the thermal characteristics of each glass region. This matrix is ​​compressed into a one-dimensional thermal weight vector in the subsequent prediction stage, with the dimension consistent with the granularity of the window region segmentation. To facilitate use by the GRU model, this thermal weight vector is timestamped after generation and stored in a short-term history cache, updated every 30 seconds, maintaining a 30-minute sliding window to form a continuous input sequence. The data in the short-term history cache is read in its entirety before each GRU model prediction, constituting one of the main input data of the GRU model, forming a complete input data set together with thermal management-related data. The fusion mechanism between the ViT model and the GRU model adopts a shared cache interface to complete the alignment of data dimensions and timestamps. This shared cache interface maps the thermal features output by the ViT model into a structured vector consistent with the distribution of the glass regions, based on the fixed topology of the glass regions, thereby ensuring the spatial consistency of the thermal change trend of the glass regions in subsequent analysis. Meanwhile, a time synchronization mechanism is introduced during data transmission to ensure that heat characteristics and other heat management-related data are updated synchronously, so as to avoid delays and status misalignments between data from different sources, and to ensure time consistency and system response accuracy.

[0076] The trend prediction model, GRU, serves as an intermediate layer, outputting the thermal change trend over the next ten minutes, such as the rate of change of heat flux per unit time in the glass area, the battery thermal rise rate, and the direction of coolant temperature difference development. These results are transmitted in real time to the XGBoost model as the main prior input for strategy generation. The fusion between the two is achieved through feature concatenation and key feature extraction mechanisms. The GRU model output first undergoes feature filtering, retaining features relevant to control commands, such as the predicted inflection point time, temperature rise rate slope, and the battery's highest temperature risk range. This is then merged with real-time thermal management data of the vehicle to construct the input feature vector for the XGBoost model. The data interface between the GRU and XGBoost models supports a dynamic weight adjustment mechanism. This means that when environmental fluctuations are severe or occupant states change frequently, the response weights to the GRU model's predicted features can be enhanced. The weights of the GRU model's predicted features are automatically adjusted based on current control stability, prediction deviation magnitude, and energy consumption index deviation, allowing the XGBoost model to prioritize upcoming thermal events in its decision-making, thus improving overall control robustness.

[0077] To further enhance the information fusion capability among the three models, after each round of control command execution, the control commands output by the XGBoost model are recorded and cached together with the input data of the ViT and GRU models for the next time step. In the ViT model, this cached information is used to correct local interference during image perception, such as color shifts in glass areas caused by control command switching, leading to image reflection distortion. Before perception, the ViT model calls the latest cached information for image correction, improving image recognition accuracy from the source. The GRU model uses this information to optimize the state transition path of heat trend changes during prediction, reducing oscillations when predicting heat trends in the future. This three-layer fusion structure adopts a master-slave scheduling framework at runtime, with the control scheduling center uniformly managing image recognition, trend prediction, and control command determination. In trend prediction, the ViT model, acting as the first-level trigger, generates image perception results within 200 milliseconds each time and immediately provides them to the GRU model. The GRU model then predicts the heat trend within 300 milliseconds and simultaneously transmits this trend to the XGBoost model. The XGBoost model determines the control command within 50 milliseconds and outputs the command. The entire chain runs once per minute to ensure that the control commands are forward-looking, responsive, and continuous, and to provide real-time adjustment capabilities for sudden heat events.

[0078] This structured fusion mechanism enables information sharing, semantic alignment, and control synchronization among multiple models, avoiding the risk of performance degradation of a single model structure under complex thermal environments. It realizes a decoupled control architecture for perception, prediction, and control, providing a stable, controllable, and scalable technical framework for the coordinated adjustment of smart glass and battery cooling strategies.

[0079] S205 performs thermal management of the vehicle based on control commands.

[0080] Among them are the light transmittance of the glass area, the flow rate of the battery coolant, the flow rate of the air conditioner evaporator, and the trigger signal of the PTC heater.

[0081] Prior to S201 above, the vehicle thermal management method may also include the following S206 to S208.

[0082] S206, a three-layer converged architecture deployment and thermal management system integration.

[0083] To ensure the vehicle's smart glass heat insulation and battery cooling capabilities are feasible in high-temperature environments, the aforementioned visual perception model (ViT), trend prediction model (GRU), and decision-making model (XGBoost) are deployed in an in-vehicle embedded computing platform. Each model is mapped to different levels of in-vehicle controller hardware based on computing resource requirements, latency constraints, and data path planning, constructing an end-to-end multi-model scheduling framework. Communication with the existing vehicle's Controller Area Network (CAN) / Local Interconnect Network (LIN) bus architecture is also established, ensuring consistency and synchronization throughout the perception, computing, and control processes. The in-vehicle embedded computing platform optimizes computing resource scheduling to ensure lightweight deployment of the entire model without increasing additional hardware costs.

[0084] The ViT model is a high-computation-density image recognition model. Its main tasks are image-level thermal distribution recognition and thermal weight encoding, therefore it needs to be deployed on an edge computing node equipped with a high-performance Neural Processing Unit (NPU). This edge computing node is typically the Artificial Intelligence (AI) processing core within the vehicle's intelligent cockpit domain controller, continuously receiving images from cameras on glass areas such as the windshield and side windows during operation. The ViT model operates in a dynamically triggered manner; that is, when changes in ambient light, sun angle, or vehicle speed exceed preset thresholds, the model immediately starts to perceive images and generate thermal features corresponding to the glass areas. The results obtained by the ViT model are processed, compressed into thermal weight vectors, and transmitted to the central intelligent controller's cache module to provide thermal feature input for the GRU model. The ViT model's operating cycle is controlled by a task scheduler, running once per minute under normal conditions, and increasing to once every 20 seconds in sudden high-temperature scenarios, with a perception latency controlled within 180 milliseconds.

[0085] The GRU model, serving as an intermediate layer for predicting heat map trends, is deployed within the vehicle's Intelligent Central Control Unit (ICCU). This ICCU possesses programmable logic resources and processing units supporting deep neural network inference, capable of meeting the operational requirements of a medium-sized GRU model. The GRU model's runtime cycle is consistent with the ViT model, using a timestamp synchronization mechanism to read the latest heat weight vector and heat management information to generate heat map trends for the next ten minutes. The GRU model's predictions are denoised in a cache and immediately passed to the XGBoost model after computation. The XGBoost model and the GRU model run on the same processing core, sharing a data buffer and utilizing a dual-channel memory read / write mechanism to improve data access efficiency.

[0086] The XGBoost model is integrated into the Thermal Management System Electronic Control Unit (TMS ECU). This controller manages the battery cooling pump, fan, electric heater, air conditioning, and dimmable glass control modules, providing low-latency response and stable control output. The XGBoost model uses GRU predictions as primary input, combined with real-time thermal management information and control history feedback, to generate multi-objective control commands. These commands can include setting the light transmittance of the glass area, the flow rate of the battery coolant, the flow rate of the air conditioning evaporator, and the PTC heater trigger signal. All control commands are encapsulated in standard CAN messages and sent to the corresponding actuators. Parameter writing and execution are achieved through the internal control protocol of the thermal management controller. During operation, the XGBoost model refreshes the control commands every minute and features strategy drift monitoring, enabling real-time evaluation of the current strategy's execution effect. When abnormal response or decreased control efficiency is detected, a strategy rollback mechanism is triggered, automatically loading a preset strategy.

[0087] In terms of thermal management system integration, the information transmission link between the three models is implemented through a middleware message bus. A unified data synchronization scheduling service is established within the thermal management system to manage the consistency of timestamps for model input and output, signal acquisition refresh frequency, and known command issuance times. All intermediate data is encapsulated in a structured format and cached in the main controller's Random Access Memory (RAM) through a cache pool mechanism. Memory allocation priorities are adjusted according to the model's running status to ensure predictable data accessibility and access security for each model's calculation results during multi-task parallel operation. To prevent data misalignment between sensor acquisition and model execution, the thermal management system introduces a soft synchronization mechanism to uniformly align the delay of external sensor data to the model calculation time before input and processing, improving time accuracy control capabilities.

[0088] Furthermore, to adapt to differences in vehicle models, the response characteristics of thermal management components, and the load capacity of controllers, the thermal management system provides configurable deployment templates before deployment. Engineers can select the appropriate model size, input dimension mapping rules, and set control command boundary values ​​based on the vehicle platform. Regarding OTA upgrades, the ViT and GRU models support post-upgrades, updating model weights through the vehicle gateway and featuring breakpoint resume and error rollback capabilities to ensure a safe upgrade process. The XGBoost model adopts a model version solidification approach, performing a complete model replacement with each strategy update. A configuration verification mechanism verifies parameter consistency before control commands are issued, ensuring the logic of the control commands is not compromised.

[0089] S207, Model running test and verification and policy regression learning.

[0090] To verify the functional stability and control command effectiveness of the thermal management system under extreme high-temperature and high-radiation environments, operational tests were conducted on the model after deployment. Test areas were selected in desert cities and high-temperature basins with intense continuous sunlight, covering typical high-heat scenarios such as high radiation, high-speed thermal shock, alternating sunlight and shadow, and urban congestion. The test vehicles included three types of glass and two cooling system architectures. Each vehicle was equipped with the sensor system and strategy recording module described above to record the inputs, outputs, actual control effects, and cabin thermal state of each model in real time after thermal management. Each control cycle of the thermal management system generated a set of strategy feedback data, including a comparison of target and actual temperature changes, energy consumption distribution, and cooling response delay. This strategy feedback data was then sent back to the backend learning module.

[0091] To optimize the control strategy, the returned strategy feedback data is first labeled and classified to form a control feedback sample set. Then, a dedicated lightweight model performs residual learning on the strategy bias. The learning process does not affect the original model's parameter structure; instead, it fine-tunes the later branch nodes of the XGBoost model to achieve response correction with minimal structural perturbation. The thermal management system dynamically selects learnable samples based on driving cycles and actual control effects to avoid abnormal data contaminating model parameters. Newly generated samples are aggregated weekly and uniformly subjected to micro-batch incremental training. Subsequently, an A / B testing mechanism is used to deploy the version to a subset of vehicles to verify stability and control performance. This optimization mechanism effectively enables the model to adaptively evolve its strategy under natural operating conditions, giving it the ability to continuously optimize the synergistic efficiency of glass heat insulation and battery cooling.

[0092] S208, setting up fault and safety tolerance mechanisms.

[0093] In complex vehicle operating environments, a fault and safety tolerance mechanism with fault perception, anomaly identification, and strategy degradation capabilities ensures the stable operation of the thermal management system. This mechanism covers three risk nodes: sensor level, model level, and execution level, ensuring that the thermal management system can quickly identify and isolate risks and switch to a safety strategy when anomalies occur at any level. At the sensor level, when phenomena such as image blurring, infrared saturation, abnormal radiation intensity fluctuations, or missing data are detected, the thermal management system automatically disables the ViT model and initiates a rule-based redundant heat load estimation strategy to ensure uninterrupted heat source estimation. At the model level, the GRU model and XGBoost model have embedded anomaly output detection mechanisms. When predicted values ​​or control commands exceed reasonable boundaries or experience drastic fluctuations, the thermal management system immediately marks the model as unreliable and activates the baseline model. At the execution level, the thermal management system is equipped with a complete control over-limit protection mechanism. If abnormal glass dimming voltage, excessive deviation between battery cooling pump feedback and setpoint, or sudden increase in cooling system loop impedance occur, the thermal management system immediately limits power output and triggers preset strategies to reduce the risk of heat load growth. Furthermore, a data verification and integrity detection mechanism is configured in the data transmission chain between the three models. When information is interrupted or timestamps are misaligned, the task forcibly terminates that round to prevent erroneous information from participating in decision-making. After detecting three consecutive operational anomalies, the thermal management system enters a restricted operating state, simultaneously recording error logs and synchronizing them to the backend cloud platform via the vehicle communication link. This fault and safety tolerance mechanism effectively avoids risks caused by policy mis-triggers, model runaway, or sensor drift in actual operation, ensuring the control safety and policy stability of the strategy under high-temperature and high-radiation operating scenarios.

[0094] In addition, such as Figure 3 As shown, Figure 3 A schematic diagram of a vehicle thermal management device provided in an embodiment of this application is shown. The device 300 includes the following modules.

[0095] The acquisition module 31 is used to acquire images of the glass area of ​​the vehicle and thermal management related information during a preset time period; wherein, the thermal management related information includes at least one of the following: environmental status information of the vehicle, operating status information of the vehicle, and battery status information of the vehicle.

[0096] Prediction module 32 is used to predict the heat change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area image and the thermal management related information.

[0097] The determining module 33 is used to determine control instructions for thermal management of the different areas based on the heat change trend of the vehicle in different areas in the future time period and the thermal management information of the vehicle at the current time.

[0098] Thermal management module 34 is used to perform thermal management on the vehicle based on the control commands.

[0099] In one specific embodiment, the prediction module 32 described above can also be used to process the glass region image through a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass region image.

[0100] In one specific embodiment, the prediction module 32 described above can be used to align and enhance the glass region image in the time and spatial dimensions; divide the aligned and enhanced glass region image into a preset number of image blocks; encode the image blocks and their corresponding positions to obtain an embedding matrix corresponding to the glass region image; and extract features from the embedding matrix based on a multi-head self-attention mechanism to obtain the heat features corresponding to each element in the embedding matrix.

[0101] In one specific embodiment, the prediction module 32 described above can be used to construct a preset number of multidimensional time series based on the time of the heat features and the heat management related information, according to a preset time length; wherein, the multidimensional time series includes the heat of the vehicle in different regions; the preset number of multidimensional time series are input into a pre-trained gated recurrent unit (GRU) model to extract the heat differences of the multidimensional time series of adjacent times, and predict the heat change trend of the vehicle in different regions in the future time period based on the heat differences.

[0102] In one specific embodiment, the determining module 33 described above can be used to input the heat change trend and the vehicle's thermal management information at the current time point into a pre-trained extreme gradient boosting XGBoost model, extract the correlation between the vehicle's thermal management information at the current time point and the heat change trend, determine the importance of the vehicle's thermal management information at the current time point based on the correlation, and determine control instructions for thermal management of the different areas based on the importance of the vehicle's thermal management information at the current time point and preset control behavior pairs; wherein, the preset control behavior pairs include the thermal management information and the control instructions corresponding to the thermal management information.

[0103] In one specific embodiment, the control command includes setting operating parameters for thermal management of the vehicle; the device 300 described above can also be used to perform thermal management of the vehicle based on a preset strategy when the operating parameters exceed a preset threshold.

[0104] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0105] Based on the same technical concept, embodiments of this application also provide an electronic device for performing the above-described vehicle thermal management method. Figure 4 This is a schematic diagram of the structure of an electronic device to implement the various embodiments of this application. The electronic device can vary significantly due to differences in configuration or performance, and may include a processor 401, a communications interface 402, a memory 403, and a communication bus 404. The processor 401, communications interface 402, and memory 403 communicate with each other via the communication bus 404. The processor 401 can call a computer program stored in the memory 403 and executable on the processor 401 to perform the various steps of the above-described vehicle thermal management method embodiments, achieving the same technical effects. To avoid repetition, further details are omitted here.

[0106] It should be noted that the electronic devices in the embodiments of this application include servers, terminals, or other devices besides terminals. For example, automobiles, robots, and handheld devices.

[0107] The above electronic device structure does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or arrange them differently. For example, an input unit may include a Graphics Processing Unit (GPU) and a microphone, and a display unit may use a liquid crystal display (LCD), organic light-emitting diode (OLED), or other similar display panels. User input units include at least one of a touch panel and other input devices. A touch panel is also called a touchscreen. Other input devices may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be elaborated further here.

[0108] Memory can be used to store software programs and various data. Memory can primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area can store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, memory can include volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0109] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly handles operations related to the operating system, user interface, and applications, while the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into the processor.

[0110] Figure 5 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application.

[0111] For example, such as Figure 5 As shown, the vehicle 500 includes a memory 501 and a processor 502. The memory 501 stores executable program code 5011, and the processor 502 is used to call and execute the executable program code 5011 to perform the above-mentioned vehicle thermal management method.

[0112] This embodiment can divide the vehicle into functional modules according to the above method example. For example, each function can be assigned to a separate module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0113] When each functional module is divided according to its corresponding function, the vehicle 500 may include: an acquisition module, a prediction module, a determination module, or a thermal management module, etc. It should be noted that all relevant content of each step involved in the above method embodiments can be referenced from the functional description of the corresponding functional module, and will not be repeated here.

[0114] The vehicle provided in this embodiment is used to execute the vehicle thermal management method described above, and therefore can achieve the same effect as the above implementation method.

[0115] When using integrated units, the vehicle may include a processing module and a storage module. The processing module is used to control and manage the vehicle's actions. The storage module supports the vehicle in executing program code and data.

[0116] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits as disclosed in this application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc., and the storage module may be a memory.

[0117] This embodiment also provides a computer-readable storage medium (including but not limited to disk storage, CD-ROM, optical storage, etc.) storing computer program code. When the computer program code is run on a computer, the computer executes the above-mentioned related method steps to implement a vehicle thermal management method provided in the above embodiment.

[0118] This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to implement a vehicle thermal management method provided in the above embodiment.

[0119] The beneficial effects of the above embodiments can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.

[0120] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0121] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0122] In the description of this application, it should be understood that if the terms "upper", "lower", "front", "rear", "left" and "right" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the position or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0123] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0124] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A vehicle thermal management method, characterized in that, include: Acquire images of the vehicle's glass area and thermal management-related information over a preset time period; wherein the thermal management-related information includes at least one of the following: the vehicle's environmental status information, the vehicle's operating status information, and the vehicle's battery status information; Based on the thermal characteristics corresponding to the glass area image and the thermal management related information, the thermal change trend of the vehicle in different areas over a future time period is predicted; Based on the heat change trend of the vehicle in different areas in the future time period and the thermal management information of the vehicle at the current time, control instructions for thermal management in the different areas are determined. Thermal management of the vehicle is performed based on the control commands; The method further includes, after acquiring the glass area image of the vehicle within a preset time period, processing the glass area image through a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass area image. The step of predicting the heat change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area image and the thermal management related information includes: Based on the time of the heat characteristics and the thermal management related information, the heat characteristics and the thermal management related information are constructed into a preset number of multidimensional time series according to a preset time length; wherein, the multidimensional time series includes the heat of the vehicle in different areas; The preset multidimensional time series are input into a pre-trained gated recurrent unit (GRU) model to extract the heat differences of multidimensional time series between adjacent times. Based on the heat differences, the heat change trend of the vehicle in different regions in the future time period is predicted.

2. The method according to claim 1, characterized in that, The step of processing the glass region image using a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass region image includes: The glass region image is aligned and enhanced in both the temporal and spatial dimensions; The aligned and enhanced glass region image is divided into a preset number of image blocks; Encode the image block and its corresponding position to obtain the embedding matrix corresponding to the glass region image; Feature extraction is performed on the embedding matrix based on a multi-head self-attention mechanism to obtain the popularity feature corresponding to each element in the embedding matrix.

3. The method according to claim 1, characterized in that, The control instructions for thermal management in different areas, based on the heat change trends of the vehicle in different areas over a future time period and the vehicle's thermal management information at the current time, include: The heat change trend and the vehicle's thermal management information at the current time point are input into a pre-trained extreme gradient boosting XGBoost model. The correlation between the vehicle's thermal management information at the current time point and the heat change trend is extracted. The importance of the vehicle's thermal management information at the current time point is determined based on the correlation. Based on the importance of the vehicle's thermal management information at the current point in time and the preset control behavior pairs, control instructions for thermal management of the different areas are determined. The preset control behavior includes the thermal management related information and the control instructions corresponding to the thermal management related information.

4. The method according to claim 3, characterized in that, The control commands include setting the operating parameters for thermal management of the vehicle; the method further includes: If the operating parameters exceed a preset threshold, thermal management of the vehicle is performed based on a preset strategy.

5. A vehicle thermal management device, characterized in that, include: The acquisition module is used to acquire images of the glass area of ​​the vehicle and thermal management-related information during a preset time period; wherein, the thermal management-related information includes at least one of the following: environmental status information of the vehicle, operating status information of the vehicle, and battery status information of the vehicle. The prediction module is used to predict the heat change trend of the vehicle in different areas over a future time period based on the thermal characteristics corresponding to the glass area image and the thermal management related information. The determination module is used to determine control instructions for thermal management of the different areas based on the heat change trend of the vehicle in different areas in the future time period and the thermal management information of the vehicle at the current time point. A thermal management module is used to perform thermal management on the vehicle based on the control commands; The prediction module is further used to process the glass region image through a pre-trained Vision Transformer (ViT) model to obtain the thermal features corresponding to the glass region image. The prediction module is used to construct a preset number of multidimensional time series based on the heat features and the heat management related information according to a preset time length, according to the time of the heat features and the heat management related information. The multidimensional time series includes the heat of the vehicle in different regions. The preset number of multidimensional time series is input into a pre-trained gated recurrent unit (GRU) model to extract the heat differences of the multidimensional time series of adjacent time periods, and predict the heat change trend of the vehicle in different regions in the future time period based on the heat differences.

6. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the vehicle thermal management method as described in any one of claims 1 to 4.

7. A vehicle, characterized in that, It includes at least one processor, which, when executed, implements the vehicle thermal management method as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the vehicle thermal management method as described in any one of claims 1 to 4.

Images