Air conditioner air outlet temperature determining method and device, vehicle and storage medium

Abstract

The invention relates to an air conditioner air outlet temperature determining method and device, a vehicle and a storage medium, and the air conditioner air outlet temperature determining method comprises the steps that vehicle actual measurement data of multiple input variables in an input variable set are obtained, and the multiple input variables in the input variable set are determined according to a thermal management system of the vehicle; inputting the vehicle actual measurement data into a preset fusion model, so that the fusion model outputs the air conditioner air outlet temperature corresponding to the vehicle actual measurement data; the fusion model comprises a linear capture model, a non-linear capture model and a fusion layer, the linear capture model is used for outputting a reference prediction value and a residual value based on the vehicle actual measurement data, the non-linear capture model is used for outputting a prediction residual value based on the residual value and the vehicle actual measurement data, and the fusion layer fuses the prediction residual value and the reference prediction value to obtain a fusion result; the air conditioner air outlet temperature is obtained. According to the embodiment of the invention, accurate prediction of the temperature of the air conditioner air outlet can be realized without additionally arranging a sensor at the air conditioner air outlet.

Description

Technical Field

[0001] This application relates to the field of automotive thermal management technology, and in particular to a method, device, vehicle, and storage medium for determining the temperature of an air conditioning outlet. Background Technology

[0002] With the increasing prevalence of intelligent vehicles, the requirements for the comfort and energy efficiency of in-vehicle air conditioning (HVAC) systems are rising. To achieve precise temperature control in the passenger compartment and improve energy efficiency, multiple temperature sensors need to be installed at key airflow nodes in the HVAC system. This ensures thermal comfort for passengers and optimizes system energy consumption.

[0003] Currently, to achieve high-precision closed-loop control of the passenger compartment's zoned temperature, multiple physical air outlet temperature sensors are strategically deployed in key air ducts such as those for face and foot ventilation in the air conditioning system. The data collected by these sensors is the core basis for constructing closed-loop feedback control and realizing real-time dynamic adjustment of the supply air temperature.

[0004] However, relying on physical sensors to measure the temperature of air conditioning vents presents significant technical and engineering challenges: ① Hardware redundancy and integration barriers: Each sensor node requires an independent sensor module, signal conditioning circuit, and dedicated I / O port, increasing system complexity and space occupation; ② Cost rigidity: Sensor modules, wiring harnesses, connectors, and increased I / O resources drive up the bill of materials (BOM) cost; ③ Assembly and space constraints: Precisely deploying sensors and their wiring harnesses within a compact air conditioning duct assembly increases assembly complexity and potential failure points; ④ Long-term reliability risks: Widely used NTC thermistors have inherent resistance and temperature drift characteristics, causing measurement accuracy to decrease over time or with environmental changes, affecting control stability; ⑤ Poor maintainability: Sensors are usually embedded inside the duct assembly, and failure often requires disassembling large interior components such as the dashboard, resulting in lengthy and costly repairs; ⑥ Functional limitations: Single-point physical sensors can only provide local duct temperature, making it difficult to accurately reflect the actual temperature felt by occupants. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this application provides a method, device, vehicle and storage medium for determining the temperature of an air conditioning outlet.

[0006] Firstly, this application provides a method for determining the temperature of an air conditioner outlet, including: Obtain vehicle measured data for multiple input variables from a set of input variables, wherein the multiple input variables in the set of input variables are determined based on the vehicle's thermal management system; The measured vehicle data is input into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the measured vehicle data. The fusion model includes a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the measured vehicle data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the measured vehicle data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

[0007] Optionally, the set of input variables includes multiple input variables such as: outside ambient temperature, inside ambient temperature, vehicle internal / external air circulation setting, blower airflow setting, evaporator temperature, hot / cold air damper opening, and PTC heater surface temperature.

[0008] Secondly, this application provides a training method for a fusion model, including: Get the set of input variables containing multiple input variables and the set of output variables containing output variables; Acquire training data and label data of the output variables collected under multiple vehicle operating conditions for multiple input variables; The fusion model is trained using training data of multiple input variables and label data of the output variables to obtain a trained fusion model.

[0009] Optionally, obtain a set of input variables that contains multiple input variables, including: Based on the vehicle's thermal management system, several initial and output variables are determined. By using correlation analysis, multiple input variables that are most correlated with the output variable are selected from the multiple initial variables, thus obtaining a set of input variables.

[0010] Optionally, a set of input variables is obtained by filtering out the input variables with the highest correlation to the output variable from the multiple initial variables through correlation analysis, including: Obtain the variable parameters of each initial variable and the variable parameters of the output variable; For each initial variable, the linear correlation between the initial variable and the output variable is calculated based on a preset first correlation coefficient formula to obtain first correlation information; The nonlinear correlation between the initial variable and the output variable is calculated based on a preset second correlation coefficient formula to obtain second correlation information; The influence of the initial variable on the output variable is calculated based on time series feature engineering to obtain third correlation information; Based on the first correlation information, the second correlation information, and the third correlation information, a fourth correlation information between the initial variable and the output variable is determined; The initial variables whose fourth correlation information is greater than the preset correlation threshold are determined as input variables, and the input variables are added to the input variable set.

[0011] Optionally, training data on the multiple input variables and label data on the output variables collected under multiple vehicle operating conditions are taken, including: The experimental design DOE model is determined based on multiple input variables and the output variables. Based on the DOE model, the level factors of multiple input variables are configured to obtain the mapping relationship between the measured value and the standardized level of each input variable; A data acquisition condition matrix is ​​constructed based on the mapping relationship between each input variable and its corresponding measured value and the standardization level. Based on the aforementioned data acquisition matrix, measured values ​​are collected on the vehicle for each input and output variable. The measured values ​​of the input variables are mapped according to the mapping relationship to obtain training data for each input variable; The measured value of the output variable is determined as the label data of the output variable.

[0012] Optionally, the fusion model is trained using training data of multiple input variables and label data of the output variables to obtain a trained fusion model, including: Training data of multiple input variables are input into the linear capture model, so that the linear part of the linear capture model outputs a first predicted value, and the nonlinear part of the linear capture model outputs a residual value. The training data of multiple input variables and the residual values ​​are input into the nonlinear capture model so that the nonlinear capture model outputs the predicted residual. The predicted residual and the first predicted value are fused through the fusion layer to obtain the second predicted value; Determine the deviation between the second predicted value and the label data of the output variable; If the deviation value is less than a preset deviation threshold, the fusion model is determined to have completed training.

[0013] Optionally, the method further includes: Obtain real vehicle test data related to multiple input variables in the input variable set; The test training data of multiple input variables are input into the trained fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the real vehicle test data of multiple input variables. Obtain the reference temperature value corresponding to the real vehicle test data of multiple input variables in the input variable set; Determine whether the error between the air conditioner outlet temperature and the reference temperature value exceeds a preset error threshold. If the error exceeds a preset error threshold, the model parameters of most fusion models are updated to obtain the fusion model with updated parameters.

[0014] Thirdly, this application provides an air conditioner outlet temperature determination device, comprising: The acquisition module is used to acquire the vehicle's measured data of multiple input variables in the input variable set, wherein the multiple input variables in the input variable set are determined according to the vehicle's thermal management system; An input module is used to input the measured vehicle data into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the measured vehicle data. The fusion model includes a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the measured vehicle data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the measured vehicle data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

[0015] Fourthly, this application provides a vehicle, including: a thermal management system, a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the air conditioner outlet temperature determination method described in any of the first aspects. The thermal management system includes: an outside temperature sensor, an inside temperature sensor, a circulating damper, a blower, an evaporator, a heating and cooling mixing damper, a PTC heater, and a mode damper.

[0016] Fifthly, this application provides a server, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the training method of the fusion model described in any of the second aspects.

[0017] In a sixth aspect, this application provides a computer-readable storage medium storing a program for a method of determining the temperature of an air conditioner outlet, wherein when the program for determining the temperature of an air conditioner outlet is executed by a processor, it implements the steps of the method for determining the temperature of an air conditioner outlet as described in any of the first aspects, or the steps of the method for training a fusion model as described in any of the second aspects.

[0018] The beneficial effects of this invention are: This application embodiment collects actual vehicle data as input to a fusion model for multiple input variables affecting the air conditioning vent temperature in the thermal management system. This allows the fusion model to output the air conditioning vent temperature more comprehensively, improving the accuracy of air conditioning vent temperature determination. The fusion model integrates the advantages of linear and nonlinear capture models, simultaneously capturing both linear trends and nonlinear characteristics of the data. Compared to a single model, it has higher prediction accuracy and avoids the shortcomings of single models, such as poor generalization ability and large transient condition prediction errors. It achieves accurate prediction of air conditioning vent temperature without the need for additional sensors at the air conditioning vent, effectively solving the problems of hardware redundancy, high cost, and complex assembly caused by relying on physical sensors in the prior art. This reduces vehicle costs and improves system reliability. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A structural diagram of an air conditioning outlet temperature determination system provided in this application embodiment; Figure 2 A flowchart illustrating a method for determining the temperature of an air conditioner outlet, provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the forward principle of a thermal management system for a REEV vehicle model, provided as an embodiment of this application; Figure 4 A schematic diagram illustrating the forward principle of a thermal management system for an EV vehicle provided in this application embodiment; Figure 5 A flowchart illustrating a training method for a fusion model provided in an embodiment of this application; Figure 6 for Figure 5 Flowchart of step S201; Figure 7 for Figure 6 Flowchart of step S302; Figure 8 for Figure 5 Flowchart of step S202; Figure 9 for Figure 5 Flowchart of step S203; Figure 10 Another flowchart of a training method for a fusion model provided in an embodiment of this application; Figure 11 A structural diagram of an air conditioner outlet temperature determination device provided in this application embodiment; Figure 12 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] The current technology for measuring air conditioning vent temperature using physical sensors presents significant technical and engineering challenges: ① Hardware redundancy and integration barriers: Each sensor node requires an independent sensor module, signal conditioning circuit, and dedicated I / O port, increasing system complexity and space requirements; ② Rigid cost: Sensor modules, wiring harnesses, connectors, and additional I / O resources drive up the bill of materials (BOM) cost; ③ Assembly and space constraints: Precisely deploying sensors and their wiring harnesses within a compact air conditioning duct assembly increases assembly complexity and potential failure points; ④ Long-term reliability risks: Widely used NTC thermistors have inherent resistance and temperature drift characteristics, causing measurement accuracy to decrease over time or with environmental changes, affecting control stability; ⑤ Poor maintainability: Sensors are usually embedded inside the duct assembly, and failure often requires disassembling large interior components such as the dashboard, resulting in lengthy and costly repairs; ⑥ Functional limitations: Single-point physical sensors can only provide local duct temperature, making it difficult to accurately reflect the actual perceived temperature of occupants. Therefore, embodiments of this application provide a method, apparatus, vehicle, and storage medium for determining the temperature of an air conditioning outlet.

[0024] like Figure 1As shown, this application provides an air conditioner outlet temperature determination system, which mainly includes four modules: a multi-dimensional input variable filtering module, a DOE global operating condition acquisition module, an LSTM and ARIMA fusion modeling module, and a dynamic feedback correction module.

[0025] The multidimensional input variable screening module, based on the architecture and forward working principle of the vehicle thermal management system, analyzes and identifies all initial variables that may affect the air conditioning vent temperature. It then uses correlation analysis to screen out key linear, nonlinear, and dynamic characteristic variables, ultimately determining the set of input variables (such as environmental parameters, vehicle status, air conditioning system status, etc.) and output variables (vent temperature) for model training and prediction. Next, an experimental matrix is ​​designed using the Design of Experiments (DOE) method based on the determined final variables.

[0026] The DOE full-domain operating condition acquisition module uses the Design of Experiments (DOE) method to scientifically plan the test matrix and collect high-quality training data under the condition of covering the full range of vehicle operating conditions, thus constructing a time series dataset containing various operating condition samples. The LSTM and ARIMA fusion modeling module trains a fusion model using a pre-built dataset. This model combines the advantages of Long Short-Term Memory (LSTM) networks, which are good at capturing complex nonlinear temporal dependencies, and Autoregressive Integrated Moving Average (ARIMA), which handles linear trends and autoregressive integrated moving averages. The optimal parameters are determined through training and optimization. The dynamic feedback correction module deploys the trained fusion model into the vehicle's electronic control unit (ECU) to predict the air vent temperature in real time and continuously monitors the residuals by comparing the predicted results with the actual sensor measurements. Once the error exceeds a preset threshold, the online model update mechanism is triggered—adaptively adjusting the parameters of the ARIMA model and performing fine-tuning transfer learning on the LSTM model based on new data. If the error is within the threshold, the predicted value is output normally, thus forming a closed-loop prediction system that can adapt to changes and continuously optimize.

[0027] This application provides a method for determining the temperature of an air conditioning vent, which can be applied to vehicles, such as... Figure 2 As shown, the method includes: Step S101: Obtain the actual vehicle data of multiple input variables in the input variable set; In this embodiment, the input variable set refers to a set of multiple input variables that affect the temperature of the air conditioning vent. The input variable set is the same as the multiple input variables in the fusion model described later. The multiple input variables in the input variable set are determined based on the forward principle of the vehicle's thermal management system. The vehicle's thermal management system refers to various components and signal transmission systems related to regulating the temperature of the vehicle's air conditioning vent. The multiple input variables in the input variable set include: outside ambient temperature, inside ambient temperature, vehicle internal / external air circulation setting, blower fan speed setting, evaporator temperature, hot / cold air damper opening, and PTC heater surface temperature.

[0028] By clearly defining the specific types of input variables through the forward principle of the thermal management system, interference from irrelevant variables is avoided, ensuring that the data input into the model are all core data that have a critical impact on the air conditioning vent temperature. This helps improve the accuracy and efficiency of model prediction. At the same time, these variables determined based on the forward principle of the vehicle thermal management system can comprehensively cover the main factors affecting the vent temperature, providing comprehensive and effective data support for the fusion model and ensuring the reliability and stability of temperature prediction.

[0029] Vehicle measured data refers to the real data corresponding to the input variables collected during the actual operation of the vehicle. Vehicle measured data includes: external ambient temperature data, internal ambient temperature data, vehicle internal and external air circulation settings data, blower airflow settings data, evaporator temperature data, hot and cold air damper opening data, and PTC heater surface temperature data.

[0030] For range-extended electric vehicles (REEVs), such as Figure 3 As shown, the thermal management system includes: an outside temperature sensor, an inside temperature sensor, internal and external circulation dampers, a blower, an evaporator, a heating / cooling mixing damper, a heater core, a PTC heater, and a mode damper, etc. For pure electric vehicles (EVs), such as... Figure 4 As shown, the thermal management system includes: an outside temperature sensor, an inside temperature sensor, an internal and external circulation damper, a blower, an evaporator, a heating and cooling mixing damper, a PTC heater, and a mode damper, etc.

[0031] Among them, the outside ambient temperature data is collected by the outside temperature sensor, and the outside ambient temperature directly affects the initial temperature of the external air circulation entering the air conditioning unit; The in-vehicle ambient temperature data is collected by the in-vehicle temperature sensor, and the in-vehicle ambient temperature directly affects the initial temperature of the air conditioning unit that the recirculation system enters. The vehicle's internal / external air circulation setting data refers to the setting of the internal / external air circulation damper, which corresponds to the vehicle's internal / external air circulation mode. The vehicle's internal / external air circulation setting determines the source of air (outside / inside the vehicle) and its initial temperature when it enters the air conditioning unit. The blower airflow setting determines the mass flow rate of air passing through the evaporator, heater core, or PTC. The blower airflow is one of the main factors in regulating the outlet temperature. Evaporator temperature is the most critical factor determining the cold air outlet temperature; the evaporator temperature is a characteristic of the cold source. The opening data of the cold and warm air damper is collected at the cold and warm air mixing damper. The cold and warm air mixing damper is used to control the mixing ratio of cold air flowing through the evaporator and hot air flowing through the warm air core / PTC. The opening of the cold and warm air damper directly affects the final air temperature reaching the air outlet. The surface temperature of the PTC heater is the most critical factor reflecting the temperature of the hot air outlet. In vehicle configurations, the surface of the heater core is generally not equipped with a temperature sensor, while the PTC is equipped with a temperature sensor. Based on the example schematic diagram of the thermal management system, the air must first pass through the heater core and then through the PTC surface temperature sensor. The temperature difference is negligible. In summary, the PTC surface temperature is a representation of the heat source.

[0032] The mode damper does not directly affect or determine the outlet temperature, but it does determine which outlet the regulated air mixture will ultimately flow to (e.g., face, feet, defrosting).

[0033] In this step, real data can be collected for each input variable in the input variable set during actual vehicle operation using data acquisition devices such as sensors on the vehicle, which can then be used as the vehicle's actual test data.

[0034] Step S102: Input the measured vehicle data into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the measured vehicle data.

[0035] In this embodiment of the application, the fusion model includes: a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the vehicle's measured data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the vehicle's measured data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

[0036] In other words, a fusion model is a composite model that integrates a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to capture the linear trend of data, the nonlinear capture model is used to capture the nonlinear characteristics of data, and the fusion layer is a module used to fuse the output results of different models. The reference predicted value is the preliminary temperature prediction result obtained by the linear capture model based on the actual vehicle data. The residual value is the difference between the actual temperature and the reference predicted value. The predicted residual value is the result obtained by the nonlinear capture model based on the residual value and the actual data, which is used to correct the reference predicted value.

[0037] In this embodiment, the fusion model is trained on the server and deployed on mass-produced vehicles after reaching the required accuracy. The trained fusion model can accurately predict the air conditioning vent temperature based on the vehicle's measured data on the input variables.

[0038] An example of a linear capture model is an Autoregressive Integrated Moving Average (ARIMA) model, and an example of a nonlinear capture model is a Long Short-Term Memory Network (LSTM) model. The fusion layer is used to add the product of the predicted residual and the first coefficient, and the product of the reference predicted value and the second coefficient, to obtain the air conditioner outlet temperature.

[0039] In this step, the collected vehicle measurement data can be input into a pre-built fusion model. Through the collaborative work of various modules within the fusion model, the air conditioning vent temperature predicted based on the vehicle measurement data is finally output.

[0040] This application embodiment collects actual vehicle data as input to a fusion model for multiple input variables affecting the air conditioning vent temperature in the thermal management system. This allows the fusion model to output the air conditioning vent temperature more comprehensively, improving the accuracy of air conditioning vent temperature determination. The fusion model integrates the advantages of linear and nonlinear capture models, simultaneously capturing both linear trends and nonlinear characteristics of the data. Compared to a single model, it has higher prediction accuracy and avoids the shortcomings of single models, such as poor generalization ability and large transient condition prediction errors. It achieves accurate prediction of air conditioning vent temperature without the need for additional sensors, effectively solving the problems of hardware redundancy, high cost, and complex assembly caused by relying on physical sensors in the prior art. This reduces vehicle costs and improves system reliability.

[0041] In another embodiment of this application, a training method for a fusion model is also provided, which can be applied to a server, such as... Figure 5 As shown, the method includes: Step S201: Obtain the set of input variables containing multiple input variables and the set of output variables containing output variables; In this embodiment of the application, the set of output variables refers to the set containing the target variable to be predicted (i.e., the output variable).

[0042] In this step, the initial variable pool can be obtained by analyzing the forward principle of the thermal management system. The set of input variables that affect the output variables can be determined in the initial variable pool by methods such as correlation analysis. The set containing the output variables is then used as the output variable set.

[0043] Step S202: Obtain training data on multiple input variables and label data of the output variables collected under multiple vehicle operating conditions; In this embodiment of the application, vehicle operating conditions refer to the working conditions of the vehicle under different environments and operating states, training data refers to the measured data of the input variables used for model training, and label data refers to the measured data of the output variables corresponding to multiple training data.

[0044] In this step, training data for each input variable and label data corresponding to multiple training data can be collected under various vehicle operating conditions.

[0045] Step S203: The fusion model is trained using the training data of the multiple input variables and the label data of the output variables to obtain the trained fusion model.

[0046] In this embodiment of the application, training refers to the process of adjusting model parameters by inputting training data and label data so that the model output results are close to the label data.

[0047] In this step, the collected training data and label data can be input into the fusion model, and the model parameters can be adjusted through iterative calculations so that the trained fusion model has the ability to accurately predict the temperature of the air conditioner outlet.

[0048] This application embodiment collects data under multiple vehicle operating conditions, making the training data comprehensive and diverse, covering various scenarios of actual vehicle operation. The fusion model trained based on this data has stronger generalization ability and can accurately predict the air conditioning vent temperature under different operating conditions. This solves the problem that the existing model can only adapt to local operating conditions and has limited prediction accuracy, and at the same time provides reliable model support for the accurate determination of the air conditioning vent temperature in the future.

[0049] In another embodiment of this application, step S201 obtains a set of input variables containing multiple input variables, such as... Figure 6 As shown, it includes: Step S301: Determine multiple initial variables and output variables based on the vehicle's thermal management system; In this embodiment of the application, the initial variable refers to the variable that is initially identified as potentially affecting the output variable. The output variable may refer to the air outlet temperature of the air conditioner. For example, it may be the temperature of the left air outlet or the temperature of the right air outlet of the air conditioner.

[0050] In this step, we can start from the overall architecture and working principle of the vehicle thermal management system to comprehensively identify the input variables that may affect the output variables.

[0051] Step S302: Through correlation analysis, select the input variables that are most correlated with the output variables from the multiple initial variables to obtain the input variable set.

[0052] In this embodiment of the application, correlation analysis refers to an analytical method used to determine the degree of association between two variables, and the set of input variables refers to a set of multiple input variables that are strongly correlated with the output variable after screening.

[0053] In this step, scientific analysis methods can be used to remove initial variables that are weakly correlated with the output variables, retain key variables, avoid irrelevant variables from interfering with model training, and improve model training efficiency and prediction accuracy.

[0054] This application embodiment determines the initial variables based on the vehicle's thermal management system, ensuring the comprehensiveness of input variable identification and avoiding the omission of key influencing factors. By screening input variables through correlation analysis, irrelevant and redundant variables can be eliminated, reducing the amount of data for model training and improving training efficiency. At the same time, all variables input to the model are key variables that are highly correlated with the output variables, which is conducive to improving the accuracy of model prediction.

[0055] In yet another embodiment of this application, as Figure 7 As shown, step S302 uses correlation analysis to select the input variables that are most correlated with the output variable from the multiple initial variables, thus obtaining a set of input variables, including: Step S401: Obtain the variable parameters of each initial variable and the variable parameters of the output variable; In this embodiment of the application, the initial variables X1-X8 include: outside ambient temperature, inside ambient temperature, vehicle internal / external air circulation setting, blower airflow setting, evaporator temperature, hot / cold air damper opening, PTC heater surface temperature, and mode damper. The output variables Y1-Y2 are the air outlet temperature (left / right). The variable parameters refer to the specific values ​​of the initial input or output variables.

[0056] Step S402: For each initial variable, calculate the linear correlation between the initial variable and the output variable based on a preset first correlation coefficient formula to obtain first correlation information; In this embodiment of the application, the preset first correlation coefficient formula refers to a preset formula used to calculate the degree of linear correlation between variables. Linear correlation refers to the degree of association between variables that have a linear relationship. The first correlation information refers to information that reflects the degree of linear correlation between the initial variable and the output variable.

[0057] In one embodiment of this application, the preset first correlation coefficient formula can be the Pearson correlation coefficient formula:

[0058] in, For input variables, For the average value of the input variable, For output variables, To output the average value of the variables, when r=1, it indicates that the two variables are perfectly positively correlated; when r=-1, it indicates that the two variables are perfectly negatively correlated; and when r=0, it indicates that there is no linear relationship between the two variables. The first correlation coefficient formula can be used to screen out variables with strong linear correlation.

[0059] In this step, the variable parameters of each initial variable and the variable parameters of the output variable can be substituted into the first correlation coefficient formula to obtain the first correlation information r, thereby quantifying the degree of linear association between the initial variable and the output variable.

[0060] Step S403: Calculate the nonlinear correlation between the initial variable and the output variable based on the preset second correlation coefficient formula to obtain the second correlation information; In this embodiment of the application, the preset second correlation coefficient formula refers to a preset formula used to calculate the degree of nonlinear correlation between variables. Nonlinear correlation refers to the degree of association between variables that exhibit a nonlinear relationship. The second correlation information refers to information reflecting the degree of nonlinear correlation between the initial variable and the output variable.

[0061] In one embodiment of this application, the preset second correlation coefficient formula can be the Spearman correlation coefficient formula:

[0062] Among them, D i =rank( )-rank( ) represents the order difference between any input variable and any output variable after sorting, and n represents the sample size. Variables with non-linear associations can be screened out by using a preset second correlation coefficient formula.

[0063] In this step, the variable parameters of each initial variable and the variable parameters of the output variable can be substituted into the second correlation coefficient formula to obtain the second correlation information. This is used to quantify the degree of nonlinear correlation between the initial and output variables.

[0064] Step S404: Calculate the influence of the initial variable on the output variable based on time series feature engineering to obtain third correlation information; In this embodiment of the application, time series feature engineering refers to the technical method used to analyze the characteristics and effects of variables changing over time, and the third correlation information refers to the information reflecting the degree of dynamic influence of the initial variable on the output variable.

[0065] In one embodiment of this application, time series characteristic engineering can refer to the multivariate Granger test formula:

[0066] in, , These are output variables, corresponding to the temperatures of the driver and passenger air vents, respectively. For input variables, X1-X7 (evaporator temperature, blower airflow, etc.); , This is a constant term (intercept term); , ' is the lag coefficient of Y itself and its influence on X; , The lag coefficients are those for X itself and the effect of Y on X. , The lag coefficients of input variable Z on Y and X; , For the random error term; p, q, r: respectively, Y, q, r, and r'. X The lag order of Y, X, and Z is given by , and i, j, and k are the initial values ​​of the lag order of Y, X, and Z, respectively.

[0067] Based on time-series feature engineering calculation of dynamic features, the influence of multiple input variables on output variables is obtained. Input variables include compressor speed, PTC temperature, etc., and output variables are the main and auxiliary air outlet temperatures.

[0068] In this step, the variable parameters of each initial variable and the variable parameters of the output variable can be substituted into the multivariate Granger test formula to obtain the third correlation information, and the dynamic influence of the initial variable on the output variable can be analyzed from the time dimension.

[0069] Step S405: Determine the fourth correlation information between the initial variable and the output variable based on the first correlation information, the second correlation information, and the third correlation information; In this embodiment of the application, the fourth correlation information refers to comprehensive information that reflects the degree of linear, nonlinear and dynamic correlation between the initial variable and the output variable.

[0070] In this step, the correlation information of the first three dimensions can be multiplied by their corresponding weights to obtain the overall correlation information between the initial variable and the output variable, which is the fourth correlation information.

[0071] Step S406: Determine the initial variable whose fourth correlation information is greater than the preset correlation threshold as the input variable, and add the input variable to the input variable set.

[0072] The preset relevance threshold refers to the pre-set critical value for judging whether a variable is a key variable. The input variable refers to the initial variable that has a strong correlation with the output variable. The set of input variables refers to the set that includes all input variables.

[0073] In this step, key variables, i.e. input variables, can be selected based on the relationship between the fourth correlation information and the preset correlation threshold, forming a set of input variables.

[0074] Based on the comprehensive analysis of data correlation, the input variable mode damper has little correlation with the output. Therefore, the input variable mode damper is removed, and the final set of input variables is determined (X1-X7 variable factors: outside ambient temperature, inside ambient temperature, vehicle internal and external circulation settings, blower airflow settings, evaporator temperature, hot and cold air damper opening degree, and PTC heater surface temperature).

[0075] This application embodiment comprehensively considers the linear, nonlinear, and dynamic correlation between initial and output variables through multi-dimensional correlation analysis. Compared with single-dimensional correlation analysis, the screening results are more accurate and comprehensive, effectively retaining variables that have a key impact on output variables and eliminating irrelevant and redundant variables. This provides a precise set of input variables for model training, which is beneficial to improving model training efficiency and prediction accuracy. It solves the problem of poor model performance caused by unscientific variable screening in the prior art.

[0076] In yet another embodiment of this application, as Figure 8 As shown, step S202 acquires training data on multiple input variables and label data on the output variables collected under multiple vehicle operating conditions, including: Step S501: Determine the experimental design DOE model based on the multiple input variables and the output variables; In this embodiment of the application, the Design of Experiments (DOE) model refers to a mathematical model used for scientifically planning experiments and collecting data. Furthermore, the DOE model formula is a second-order response surface model (RSM).

[0077] Where Y is the response variable (outlet temperature, which needs to be measured); X i The i-th input factor after standardization (such as X1-X7 external and internal vehicle temperatures, scaled to [-1, 1]). The constant term (intercept) represents the baseline response when all factors are at their median levels; The linear effect coefficient reflects the linear influence of Xi on Y; This is the second-order effect coefficient, reflecting the nonlinear (curvature) influence of Xi; The interaction coefficient reflects the combined effects of Xi and Xj. The random error term is assumed to follow a normal distribution. N(0,σ 2 ), can be any value randomly selected from the normal distribution as .

[0078] In this step, a suitable DOE model can be selected based on the characteristics of the input and output variables, providing a scientific basis for subsequent experimental design and data collection.

[0079] Step S502: Configure the level factors of multiple input variables according to the DOE model to obtain the mapping relationship between the measured value and the standardized level of each input variable; In this embodiment of the application, level factor configuration refers to the process of setting different standardization levels and standard values ​​for each input variable according to the DOE model. The measured value refers to the specific value of the input variable under actual working conditions. The standardization level can be low, medium, or high, and the standard values ​​corresponding to each standardization level are [-1, 0, 1]. The mapping relationship refers to the correspondence between the measured value, the standardization level, and the standard value.

[0080] Furthermore, based on the selected DOE model, the input quantities (X1-X7) can be set with level factors according to the second-order response surface model. Each input variable needs to be set with 3 levels (low, medium, and high) and standardized to the interval [-1, 0, 1]. For example, the opening degree of the hot and cold air damper: low level (-1): 0%; medium level (0): 50%; high level (+1): 100%. Other input variables are set in sequence.

[0081] Step S503: Construct a data acquisition condition matrix based on the mapping relationship between each input variable and its corresponding measured value and standardization level; In this embodiment of the application, the data acquisition condition matrix refers to a set of test plans containing different combinations of all input variables at different levels. The data acquisition condition matrix includes: 1. Cube point (2^k): Full factorial design, generating all high-low combinations, i.e., the input variables X1-X7 are 2^7=128 times.

[0082] 2. Axis point ( ): Extending along the axis of each factor (± ), used to estimate quadratic terms, Usually taken = (For the 7 input variable factors, ≈3.36).

[0083] 3. Center point: Take the median level (0) of all factors and repeat 3-5 times to estimate the pure error.

[0084] In this step, different levels of input variables can be combined to generate a test plan that comprehensively covers various working conditions, ensuring that the collected data can cover all possible combinations of input variable values.

[0085] Step S504: Based on the collected working condition matrix, collect measured values ​​for each input and output variable on the vehicle; In this step, data is collected on the actual vehicle according to the data acquisition condition matrix, and the measured values ​​of each input and output variable are obtained.

[0086] Step S505: The measured values ​​of the input variables are mapped according to the mapping relationship to obtain training data for each input variable; In this embodiment of the application, mapping processing refers to the process of converting the measured values ​​of input variables into corresponding standardized levels.

[0087] In this step, the collected measured values ​​of the input variables are converted into standardized data according to the mapping relationship to obtain the training data for each input variable.

[0088] Step S506: Determine the measured value of the output variable as the label data of the output variable.

[0089] Label data refers to the real data of the output variables corresponding to the training data, which is used for error calculation and parameter adjustment during model training.

[0090] This application's embodiments, through scientific DOE model design and level factor configuration, ensure the rationality and comprehensiveness of the experimental scheme. The collected operating condition matrix can cover all possible combinations of input variables and various extreme and transient operating conditions. The data collected based on this matrix is ​​comprehensive and representative, effectively supporting model training. The standardized training data can improve the stability and efficiency of model training, solving the problems of poor model generalization ability caused by the lack of scientific data collection and incomplete operating condition coverage in the prior art. It provides high-quality training data and label data for the model, ensuring the model's prediction accuracy and reliability.

[0091] In another embodiment of this application, step S203 uses the training data of the multiple input variables and the label data of the output variables to train the fusion model, thereby obtaining a trained fusion model, such as... Figure 9 As shown, it includes: Step S601: Input the training data of the multiple input variables into the linear capture model so that the linear part of the linear capture model outputs a first predicted value and the nonlinear part of the linear capture model outputs a residual value. In this embodiment of the application, the first predicted value is the temperature prediction value under a linear trend obtained by the linear capture model based on the training data, and the residual value is the difference between the label data of the output variable and the first predicted value.

[0092] The linear capture model consists of three parts: autoregression (AR), differencing (I), and moving average (MA). Its basic form is as follows:

[0093] in, The time series represents the observations at time point t, which are the variables that the model wants to explain and predict; L is the lag operator; d This indicates the number of differences performed on the sequence. To express [the opinion / towards] The purpose of performing d-order differencing is to make the sequence stationary; p represents the order of the autoregressive part, i.e., the number of past steps used in the model. p Use the value at each moment to predict the current value; It is the autoregressive coefficient, representing right The degree of influence; q represents the order of the moving average part, that is, the error term of the past q times is used in the model to predict the current value; It is the moving average coefficient; This represents the random error (white noise) at time t, which is usually assumed to follow a normal distribution with a mean of 0 and a constant variance, and that each period is independent of the others.

[0094] The specific principle of the linear capture model is as follows: The linear capture model can first check the stationarity of the data (by the difference order d); then, based on the specified seasonal cycle, establish a seasonal ARIMA model, estimate the model parameters, use the model to make predictions, and calculate the residuals (actual value - predicted value).

[0095] In this step, the linear capture model can be used to mine the linear trends and periodic characteristics in the training data to obtain preliminary prediction results, namely the first predicted value and the residuals that the linear model cannot explain.

[0096] Step S602: Input the training data of the multiple input variables and the residual values ​​into the nonlinear capture model so that the nonlinear capture model outputs the predicted residual value; In this embodiment of the application, the predicted residual is the result obtained by the nonlinear capture model based on the training data and the residual value, which is used to correct the first predicted value.

[0097] Using an LSTM neural network model as a non-linear pattern learner, the basic form of the model is as follows: 1. The Gate of Oblivion:

[0098] in, The output value of the forget gate, ranging from [0,1], controls the cell state at the previous time step. How much information is retained? The Sigmoid activation function maps the input to [0,1] and is used to simulate the "degree of forgetting"; The weight matrix of the forget gate is used to weight the concatenated input. Perform a linear transformation; The bias term of the forget gate is used to adjust the result of the linear transformation; The hidden state from the previous moment With current input The features are concatenated to form the input vector at the current moment; The hidden state from the previous moment, which includes historical information; This is the input vector at the current time.

[0099] 2. Input Gate:

[0100]

[0101] in, The output of the input gate, with a value range of [0, 1], controls the state of the candidate cells. How much information is written to the current cell state? Candidate cell state, value range [ [1, 1] is calculated from the current input and the hidden state at the previous time step, representing the possible new information at the current time step; , The weight matrix and bias terms of the input gate; , Weight matrix and bias terms for candidate cell states; The hyperbolic tangent activation function maps the input to [ [1, 1] is used to generate candidate information.

[0102] 3. Cell state update:

[0103] in, The current state of the cell is the long-term memory carrier of the LSTM; The cell state is shown in the previous time step; ⊙ represents the element-wise multiplication (Hadamard product), used to proportionally fuse information. This formula combines information from the forget gate and the input gate to update the cell state. Decide how much old information to retain. Decide how much new information to add.

[0104] 4. Output gate:

[0105]

[0106] in, The output of the output gate controls the current cell state. How much information is output to the hidden state? ; The hidden state at the current moment can be passed as output to the next moment, or it can be used as the output at the current moment. , The weight matrix and bias terms of the output gate; Map cell states to [ [1, 1], and then combined with the output gate to obtain the final hidden state.

[0107] The specific principle is that the model is set as follows: 1. Input data: The standardized original 7 features plus ARIMA residuals (a total of 8 features) are used as input, with one feature vector at each time step; 2. Input layer: Receives 8-dimensional features; 3. LSTM layer: 64 units, processes sequence data, and outputs the sequence; 4. Dropout layer: Prevents overfitting; 5. Fully connected layer: 32 neurons, performs feature transformation; 6. ReLU activation function: Introduces non-linearity; 7. Output layer: 1 neuron (because the output is a single variable), outputs residual prediction.

[0108] In this step, the nonlinear capture model can be used to mine the nonlinear features in the training data and the nonlinear information contained in the residuals to obtain the prediction residuals used to correct the linear prediction results.

[0109] Step S603: The predicted residual and the first predicted value are fused through the fusion layer to obtain the second predicted value; In this embodiment, the fusion layer is a module used to integrate the output results of different models, and the second predicted value is the final predicted value obtained after fusing the linear prediction result and the nonlinear correction result.

[0110] In this step, the second predicted value can be obtained by adding the product of the predicted residual and the first coefficient, and the product of the first predicted value and the second coefficient, through the fusion layer.

[0111] Step S604: Determine the deviation between the second predicted value and the label data of the output variable; In this embodiment, the deviation value refers to the degree of difference between the second predicted value and the label data, and is used to evaluate the accuracy of the model prediction.

[0112] In this step, the difference between the model prediction result (the second predicted value) and the true value (the label data) can be calculated to obtain the deviation value.

[0113] Step S605: If the deviation value is less than a preset deviation threshold, the training of the fusion model is determined to be complete.

[0114] The preset deviation threshold refers to the pre-set critical value of deviation used to determine whether the model training has met the target.

[0115] In this step, the deviation value can be used to determine whether the model has reached sufficient prediction accuracy. If it does, the training is complete; if it does not, the model parameters need to be adjusted further.

[0116] This application embodiment utilizes the collaborative work of a linear capture model and a nonlinear capture model to mine the linear and nonlinear features of the data respectively. The fusion layer integrates the output results of the two models, achieving complementary advantages. Compared with a single model, it can capture data features more comprehensively and improve prediction accuracy. By comparing the deviation value with a preset threshold, it determines whether the model training is complete, ensuring that the trained model has sufficient prediction accuracy.

[0117] In another embodiment of this application, after step S203, Figure 10 As shown, the method further includes: Step S701: Obtain real vehicle test data for multiple input variables in the input variable set; Step S702: Input the test training data of multiple input variables into the trained fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the real vehicle test data of multiple input variables. Step S703: Obtain the reference temperature value corresponding to the real vehicle test data of multiple input variables in the input variable set; The baseline temperature value refers to the actual temperature value of the air conditioning vent, which is collected by a temperature sensor set at the air vent and corresponds to the training data of each input variable. It is used to verify the accuracy of the model prediction results. Temperature sensors can be set at the air conditioning vent of the test vehicle to collect the baseline temperature value.

[0118] Step S704: Determine whether the error between the air conditioner outlet temperature and the reference temperature value exceeds a preset error threshold. Error refers to the degree of difference between the air conditioner outlet temperature predicted by the model and the reference temperature value. The preset error threshold is a pre-set critical value for judging whether the model prediction result is acceptable.

[0119] This step can determine the accuracy of the model's prediction results. If the error exceeds the threshold, it means that the model's prediction accuracy is insufficient and the parameters need to be adjusted.

[0120] Step S705: If the error exceeds a preset error threshold, update the model parameters of most fusion models to obtain the fusion model with updated parameters.

[0121] Model parameters refer to various adjustable parameters in the fusion model that affect the prediction results. Parameter updates refer to adjusting the model parameters according to new error conditions so that the model prediction results are closer to the true values.

[0122] In this step, when the model's prediction accuracy is insufficient, the model performance can be optimized by adjusting the parameters to ensure that the model can maintain a high prediction accuracy.

[0123] This application embodiment verifies the model prediction results by using a reference temperature value, which can promptly detect changes in the model's prediction accuracy. When the error exceeds a threshold, the model parameters are updated, realizing online optimization of the model. This allows the model to adapt to changes in various factors during vehicle operation, such as sensor baseline shift and attenuation of air conditioning system thermal efficiency, thereby maintaining high prediction accuracy and ensuring the continuous accuracy of air conditioning vent temperature prediction.

[0124] In another embodiment of this application, an air conditioning outlet temperature determination device is also provided, such as... Figure 11 As shown, it includes: The acquisition module 11 is used to acquire vehicle measured data of multiple input variables in the input variable set, wherein the multiple input variables in the input variable set are determined according to the vehicle's thermal management system. Input module 12 is used to input the vehicle measured data into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the vehicle measured data. The fusion model includes: a linear capture model, a nonlinear capture model and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the vehicle measured data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the vehicle measured data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

[0125] In another embodiment of this application, a vehicle is also provided, including: a thermal management system, a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the air conditioner outlet temperature determination method described in any of the foregoing method embodiments. The thermal management system includes: an outside temperature sensor, an inside temperature sensor, a circulating damper, a blower, an evaporator, a heating and cooling mixing damper, a PTC heater, and a mode damper.

[0126] In another embodiment of this application, a server is also provided, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the training method of the fusion model described in any of the foregoing method embodiments.

[0127] The vehicle provided in this embodiment of the invention uses a processor that executes a program stored in memory to collect measured vehicle data as input to a fusion model for multiple input variables affecting the air conditioning vent temperature in the thermal management system. This allows the fusion model to output the air conditioning vent temperature by comprehensively analyzing the input data, thus improving the accuracy of the air conditioning vent temperature determination. The fusion model integrates the advantages of linear and nonlinear capture models, simultaneously capturing both linear trends and nonlinear characteristics of the data. Compared to a single model, it has higher prediction accuracy and avoids the shortcomings of single models, such as poor generalization ability and large transient condition prediction errors. This achieves accurate prediction of the air conditioning vent temperature without the need for additional sensors, effectively solving the problems of hardware redundancy, high cost, and complex assembly caused by relying on physical sensors in the prior art, reducing vehicle costs and improving system reliability.

[0128] The server provided in this embodiment of the invention allows the processor to collect data under multiple vehicle operating conditions by executing programs stored in the memory. This makes the training data comprehensive and diverse, covering various scenarios of actual vehicle operation. The fusion model trained based on this data has stronger generalization ability and can accurately predict the air conditioning vent temperature under different operating conditions. This solves the problem that the existing model can only adapt to local operating conditions and has limited prediction accuracy. At the same time, it provides reliable model support for the accurate determination of the air conditioning vent temperature in the future.

[0129] The communication bus 1140 mentioned in the aforementioned vehicles and servers can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 12 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0130] The communication interface 1120 is used for communication between the above-mentioned electronic device and other devices.

[0131] The memory 1130 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0132] The processor 1110 mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0133] In another embodiment of this application, a computer-readable storage medium is also provided, on which a program for determining the temperature of an air conditioner outlet is stored. When the program for determining the temperature of an air conditioner outlet is executed by a processor, it implements the steps of the method for determining the temperature of an air conditioner outlet as described in any of the foregoing method embodiments, or the steps of the method for training a fusion model as described in any of the foregoing method embodiments.

[0134] It should be noted that, in this document, relational terms such as "first" and "second" are used merely 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. Furthermore, 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 a process, method, article, or apparatus. Without further limitations, 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 said element.

[0135] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for determining the temperature of an air conditioner outlet, characterized in that, include: Obtain vehicle measured data for multiple input variables from a set of input variables, wherein the multiple input variables in the set of input variables are determined based on the vehicle's thermal management system; The measured vehicle data is input into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the measured vehicle data. The fusion model includes a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the vehicle's measured data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the vehicle's measured data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

2. The method for determining the air conditioner outlet temperature according to claim 1, characterized in that, The set of input variables includes multiple input variables such as: outside ambient temperature, inside ambient temperature, vehicle internal / external air circulation setting, blower fan speed setting, evaporator temperature, hot / cold air damper opening, and PTC heater surface temperature.

3. A training method for a fusion model, characterized in that, include: Get the set of input variables containing multiple input variables and the set of output variables containing output variables; Acquire training data and label data of the output variables collected under multiple vehicle operating conditions for multiple input variables; The fusion model is trained using training data of multiple input variables and label data of the output variables to obtain a trained fusion model.

4. The training method for the fusion model according to claim 3, characterized in that, Retrieve a set of input variables that contains multiple input variables, including: Based on the vehicle's thermal management system, several initial and output variables are determined. By using correlation analysis, multiple input variables that are most correlated with the output variable are selected from the multiple initial variables, thus obtaining a set of input variables.

5. The training method for the fusion model according to claim 4, characterized in that, Correlation analysis is used to select the input variables that are most correlated with the output variable from among the initial variables, resulting in a set of input variables, including: Obtain the variable parameters of each initial variable and the variable parameters of the output variable; For each initial variable, the linear correlation between the initial variable and the output variable is calculated based on a preset first correlation coefficient formula to obtain first correlation information; The nonlinear correlation between the initial variable and the output variable is calculated based on a preset second correlation coefficient formula to obtain second correlation information; The influence of the initial variable on the output variable is calculated based on time series feature engineering to obtain third correlation information; Based on the first correlation information, the second correlation information, and the third correlation information, a fourth correlation information between the initial variable and the output variable is determined; The initial variables whose fourth correlation information is greater than the preset correlation threshold are determined as input variables, and the input variables are added to the input variable set.

6. The training method for the fusion model according to claim 3, characterized in that, Acquiring training data and label data of the output variables collected under multiple vehicle operating conditions, including: The experimental design DOE model is determined based on multiple input variables and the output variables. Based on the DOE model, the level factors of multiple input variables are configured to obtain the mapping relationship between the measured value and the standardized level of each input variable; A data acquisition condition matrix is ​​constructed based on the mapping relationship between each input variable and its corresponding measured value and the standardization level. Based on the aforementioned data acquisition matrix, measured values ​​are collected on the vehicle for each input and output variable. The measured values ​​of the input variables are mapped according to the mapping relationship to obtain training data for each input variable; The measured value of the output variable is determined as the label data of the output variable.

7. The training method for the fusion model according to claim 3, characterized in that, The fusion model is trained using training data from multiple input variables and label data from the output variables to obtain a trained fusion model, including: Training data of multiple input variables are input into a linear capture model, so that the linear part of the linear capture model outputs a first predicted value, and the nonlinear part of the linear capture model outputs a residual value. The training data of multiple input variables and the residual values ​​are input into the nonlinear capture model so that the nonlinear capture model outputs the predicted residual. The predicted residual and the first predicted value are fused together by a fusion layer to obtain a second predicted value; Determine the deviation between the second predicted value and the label data of the output variable; If the deviation value is less than a preset deviation threshold, the fusion model is determined to have completed training.

8. The training method for the fusion model according to claim 3, characterized in that, The method further includes: Obtain real vehicle test data related to multiple input variables in the input variable set; The test training data of multiple input variables are input into the trained fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the real vehicle test data of multiple input variables. Obtain the reference temperature value corresponding to the real vehicle test data of multiple input variables in the input variable set; Determine whether the error between the air conditioner outlet temperature and the reference temperature value exceeds a preset error threshold; If the error exceeds a preset error threshold, the model parameters of most fusion models are updated to obtain the fusion model with updated parameters.

9. A device for determining the temperature of an air conditioner outlet, characterized in that, include: The acquisition module is used to acquire the vehicle's measured data of multiple input variables in the input variable set, wherein the multiple input variables in the input variable set are determined according to the vehicle's thermal management system; An input module is used to input the measured vehicle data into a preset fusion model so that the fusion model outputs the air conditioning vent temperature corresponding to the measured vehicle data. The fusion model includes a linear capture model, a nonlinear capture model, and a fusion layer. The linear capture model is used to output a reference predicted value and a residual value based on the measured vehicle data. The nonlinear capture model is used to output a predicted residual value based on the residual value and the measured vehicle data. The fusion layer fuses the predicted residual value and the reference predicted value to obtain the air conditioning vent temperature.

10. A vehicle, characterized in that, include: The system includes a thermal management system, a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. Memory, used to store computer programs; The processor, when executing the program stored in the memory, implements the air conditioner outlet temperature determination method according to any one of claims 1 to 2; The thermal management system includes: an outside temperature sensor, an inside temperature sensor, a circulating damper, a blower, an evaporator, a heating and cooling mixing damper, a PTC heater, and a mode damper.

11. A server, characterized in that, include: The processor, communication interface, memory, and communication bus are connected, with the processor, communication interface, and memory communicating with each other via the communication bus. Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the training method of the fusion model according to any one of claims 3 to 8.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for a method of determining the temperature of an air conditioner outlet. When the program for determining the temperature of an air conditioner outlet is executed by a processor, it implements the steps of the method for determining the temperature of an air conditioner outlet as described in any one of claims 1 to 2, or the steps of the training method for the fusion model as described in any one of claims 3 to 8.

Images