Model training method, thermal balance capability evaluation method and electronic device

Abstract

The present invention discloses a model training method, a thermal balance capability evaluation method and an electronic device, belonging to the field of automobile technology, comprising: obtaining a training sample, the training sample comprising a vehicle operating condition parameter and a vehicle fixed parameter of a historical thermal balance test and a corresponding thermal balance capability test result; using the vehicle operating condition parameter and the vehicle fixed parameter as input parameters, and using the thermal balance capability test result as an output parameter to train a preset model until convergence, and obtain a thermal balance capability evaluation model. The present invention does not require thermal balance road or indoor tests, greatly shortens the test cycle and cost, has higher prediction accuracy, and a shorter cycle.

Description

Technical Field

[0001] The present invention relates to the field of automobile technology, and in particular to a model training method, a thermal balance capability evaluation method and electronic equipment. Background Art

[0002] With the rapid progress and development of my country's automobile industry, the requirements for various automobile production indicators have become increasingly stringent, which has led to automobile manufacturers investing more costs to test various indicators of their vehicles.

[0003] In order to meet the heat dissipation requirements of automobile engines under various working conditions, a thermal balance release test is usually carried out before the automobile leaves the factory to evaluate the heat dissipation performance of its cooling system. Vehicle thermal balance tests can be conducted on the road or indoors. Due to the long development cycle of test samples and high test costs, evaluating the thermal balance capacity of vehicles consumes a lot of time and financial costs.

[0004] Therefore, the technical problem of how to ensure the accuracy of vehicle thermal balance capacity assessment while improving efficiency needs to be solved urgently. Summary of the invention

[0005] In order to solve the technical problem of how to ensure the accuracy of vehicle thermal balance capability evaluation while improving efficiency as described in the above background technology, the present invention provides a model training method, a thermal balance capability evaluation method and an electronic device.

[0006] The first object of the present invention is to propose a model training method. The thermal balance capacity evaluation model trained by the method can achieve a rapid evaluation of the thermal balance capacity of an automobile while ensuring a certain accuracy, can ensure that the prediction result complies with the regulations of the automobile thermal balance capacity road test method, and saves the time and financial cost of automobile thermal balance evaluation.

[0007] A second objective of the present invention is to provide a method for evaluating the thermal balance capability of a vehicle.

[0008] The third object of the present invention is to provide an electronic device. According to the first aspect, the embodiment of the present application provides a model training method, comprising: obtaining a training sample, the training sample including vehicle operating parameters and vehicle fixed parameters of a historical thermal balance test and a corresponding thermal balance capability test result; using the vehicle operating parameters and vehicle fixed parameters as input parameters, and using the thermal balance capability test result as an output parameter to train a preset model until convergence, and obtaining a thermal balance capability evaluation model.

[0009] According to the model training method of the embodiment of the present invention, the trained model can evaluate the data to be evaluated, and there is no need to conduct road and indoor tests, thereby achieving the purpose of reducing costs and increasing efficiency.

[0010] Optionally, the model training method further includes calibrating the prediction results in the model training process using the change information of the liquid-temperature difference curve of the working condition corresponding to the training sample to obtain a calibrated thermal balance capacity evaluation model.

[0011] Optionally, the use of the liquid-air temperature difference curve change information to calibrate the prediction results in the model training process includes: obtaining multiple liquid-air temperature difference first sampling values ​​in the prediction results; calculating the liquid-air temperature difference average value in the prediction results; calculating the difference between the first sampling value and the liquid-air temperature difference average value; correcting the first sampling value based on the difference and the liquid-air temperature difference curve change information to obtain a calibrated prediction result; and correcting the thermal balance capability assessment model based on the calibrated prediction result to obtain a calibrated thermal balance capability assessment model.

[0012] Optionally, the use of the liquid-air temperature difference curve change information to calibrate the prediction results in the model training process includes: obtaining multiple liquid-air temperature difference second sampling values ​​in the prediction results; dividing the second sampling values ​​into multiple groups, wherein each group includes at least two second sampling values; calculating the change rate of the second sampling values ​​in each group; based on the time sequence, calibrating the change rate of the latter group using the previous group in sequence, so that the change rate of the latter group changes toward the change rate of the previous group, to obtain a calibrated prediction result; based on the calibrated prediction result, correcting the thermal balance capability assessment model to obtain a calibrated thermal balance capability assessment model.

[0013] Optionally, the vehicle fixed parameters include: ambient temperature, radiator, intercooler, fan, test bench ambient temperature, coolant flow, coolant heat removal, intake air flow, intercooler heat removal and heat generation parameters; the vehicle operating parameters include: liquid-temperature difference, engine speed, torque and power.

[0014] Optionally, the prediction result is a dynamic liquid-temperature difference parameter and / or a temperature difference constant.

[0015] Optionally, the training samples are normalized.

[0016] Optionally, all combinations of model hyperparameters are traversed in turn, and model hyperparameters are screened based on cross-validation.

[0017] According to the second aspect, an embodiment of the present application provides an application method of a thermal balance capability assessment model, including: obtaining vehicle data to be evaluated, the data of the vehicle to be evaluated including vehicle operating condition parameters and vehicle fixed parameters; inputting the data of the vehicle to be evaluated into the thermal balance capability assessment model to obtain an assessment result of the thermal balance capability of the vehicle to be evaluated, the thermal balance capability assessment model being based on the vehicle operating condition parameters and the vehicle fixed parameters as input parameters, and being trained with the thermal balance capability test results as output parameters.

[0018] According to the vehicle thermal balance capability evaluation method of an embodiment of the present invention, after the thermal balance capability evaluation model is trained, the data to be evaluated is input into the model to obtain the evaluation result, and there is no need to conduct road and indoor tests, thereby achieving the purpose of reducing costs and increasing efficiency. The thermal balance capability evaluation model is based on vehicle operating parameters and vehicle fixed parameters as input parameters, and the thermal balance capability test results are trained as output parameters.

[0019] According to a third aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the model training method as described in any one of the first aspect above and / or the vehicle thermal balance capability assessment method as described in the second aspect above.

[0020] The thermal balance capability assessment model trained to convergence solves the problem that the thermal balance test requires that the ambient temperature should not be lower than 30°C, and the thermal balance test has large errors in low-temperature winter environments. It no longer requires long test cycle steps such as test sample production, sensor measurement point processing, and vehicle modification, and also reduces manpower and material consumption. At the same time, it is no longer limited by the problem that the calculation accuracy cannot be guaranteed due to the limited accuracy of the cooling system and engine bench test data and the vehicle cabin three-dimensional model provided by the manufacturer. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A flowchart of a model training method in an embodiment of the present application;

[0023] Figure 2 Schematic diagram of the XGBoost algorithm principle in the embodiment of the present application;

[0024] Figure 3 This is a schematic diagram of a liquid-air temperature difference curve with an upward trend in an embodiment of the present application;

[0025] Figure 4 A schematic diagram of the model training results in an embodiment of the present application;

[0026] Figure 5A flow chart of a method for evaluating thermal balance capability of a vehicle in an embodiment of the present application;

[0027] Figure 6 This is a schematic diagram comparing the liquid-air temperature difference change curves under high torque point conditions in another embodiment of the present application. DETAILED DESCRIPTION

[0028] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0030] The technical terms appearing in this application are explained as follows:

[0031] Common model parameters: configuration variables within the machine learning model, which are automatically calculated by the computer based on the machine learning algorithm and given data, and are referred to as common parameters below;

[0032] Model hyperparameters: Configuration variables external to the machine learning model, which require manual experience and debugging to confirm the optimal value, hereinafter referred to as hyperparameters.

[0033] The present invention provides a model training method, see Figure 1 As shown, including:

[0034] S11. Obtain training samples, where the training samples include vehicle operating parameters and vehicle fixed parameters of historical thermal balance tests and corresponding thermal balance capability test results.

[0035] The training samples include vehicle operating parameters and vehicle fixed parameters of historical thermal balance tests and corresponding thermal balance capability test results. The training samples are directly obtained from the database, and the inventor does not need to conduct relevant tests to obtain them, thus saving time.

[0036] The training samples may include ambient temperature, liquid-air temperature difference, engine speed, vehicle cooling module configuration parameters, fan speed ratio, vehicle model, engine performance data measured by the test bench, etc. Among them, the vehicle cooling configuration module includes radiator, intercooler, and fan, all of which are represented by numbers for different cooling module configurations; the engine performance data measured by the test bench include bench ambient temperature, coolant flow, coolant heat removal, intake flow, intercooler heat removal, heat generation, torque, power, and the vehicle cooling configuration module and vehicle model are both represented by numbers.

[0037] S12. Using the vehicle operating parameters and vehicle fixed parameters as input parameters and the thermal balance capability test results as output parameters to train the preset model until convergence, thereby obtaining a thermal balance capability evaluation model.

[0038] After the input parameters and output parameters are determined, the model is continuously trained by relying on a large number of training samples, and the common parameters and hyperparameters of the model are gradually improved, thereby ensuring the accuracy of the thermal balance capacity evaluation model.

[0039] The thermal balance capability assessment model trained to convergence solves the problem that the thermal balance test requires that the ambient temperature should not be lower than 30°C, and the thermal balance test has large errors in low-temperature winter environments. It no longer requires long test cycle steps such as test sample production, sensor measurement point processing, and vehicle modification, and also reduces manpower and material consumption. At the same time, it is no longer limited by the problem that the calculation accuracy cannot be guaranteed due to the limited accuracy of the cooling system and engine bench test data and the vehicle cabin three-dimensional model provided by the manufacturer.

[0040] In this embodiment, according to the provisions of the automobile thermal balance capacity road test method, to determine whether the vehicle has reached a thermal balance state, it is necessary to consider the dynamic changes of the relevant parameters in the vehicle operating parameters within 4 consecutive minutes, by introducing the cumulative heat taken away by the coolant, the cumulative heat taken away by the intercooler, and the cumulative heat generation, and processing the liquid temperature difference according to formula (1), to characterize the dynamic influence of the time variable on the prediction result. Among them, ΔT new represents the dynamic liquid-air temperature difference, ΔT represents the liquid-air temperature difference, and t represents time.

[0041]

[0042] The dynamic liquid-air temperature difference is used as the output parameter, and the vehicle operating condition parameters and vehicle fixed parameters are used as the input parameters to form the training samples of the thermal balance capability evaluation model. As shown in Table 1, the thermal balance capability evaluation model is trained based on the training samples. During the training process, the common parameters and hyperparameters of the model are gradually screened until the evaluation accuracy of the model reaches the specified standard of the automobile thermal balance capability road test method.

[0043] In this embodiment, the training samples are divided, and the thermal balance test results of six vehicle configuration schemes are used to generate a sample data set. One sample data of each scheme is used as a test sample, and the remaining data is used for debugging and training of the machine learning model.

[0044] Table 1 Characteristic variables

[0045]

[0046]

[0047] In this embodiment, see Figure 2 As shown, the thermal balance evaluation model is built based on the Extreme Gradient Boosting (XGBoost) tree, which deeply explores the intrinsic relationship between the thermal balance capacity of the automobile and various influencing factors. XGBoost is an ensemble learning algorithm based on decision trees. It builds multiple decision trees through multiple rounds of iterations to improve the accuracy of prediction. Each tree in XGBoost fits the prediction residual of the previous tree model. The prediction results of the tree model obtained from each round of training are summed to obtain the final XGBoost model prediction result, which is the additive model.

[0048] Each base learner corresponds to a decision tree model. The objective function of each decision tree model is shown in formula (2): The first term represents the error sum of each fitting sample, and the second term is the regularization term, which restricts the complexity of the tree to avoid overfitting. The objective function is minimized by performing a second-order Taylor expansion on the loss function and solving the optimal leaf node weight. Among them, y i represents the true value; represents the predicted value of the first t-1 trees; l represents the loss function, which is used to measure the error between the predicted value and the true value; L (t) represents the predicted value of the tth tree; Ω(f t ) represents the regularization term of the tth tree; f t (x i ) represents the prediction result of the t-th tree; T represents the number of leaf nodes of the t-th tree; γ represents the leaf node penalty regularization term of the t-th tree; λ represents the leaf weight penalty regularization term of the t-th tree; w represents the leaf node weight value of the t-th tree.

[0049]

[0050] When selecting the split point, we traverse all the characteristic variable values ​​of the sample data, subtract the sum of the objective function values ​​of the two child nodes after the split from the objective function value before the node split, and calculate the split gain according to formula (3) to obtain the optimal split point of the tree model. The split operation is performed recursively until the conditions for stopping the split are met, and the tree building process of this round of the XGBoost model is completed.

[0051]

[0052] Among them, L split Indicates the node splitting gain value; g i represents the first-order derivative of the loss function of the previous tree; h i represents the second-order derivative of the loss function of the previous tree; I represents the total sample set before the node splits; I R Represents the right subset after the node splits; I L Represents the left subset after the node splits.

[0053] It should be noted that the algorithm mentioned in the present invention is not limited to the XGBoost algorithm, and other algorithms such as neural networks, support vector machines, etc. can also be used to build a prediction model to achieve the prediction of the vehicle's thermal balance capability. In addition, the prediction results in this embodiment can not only represent the dynamic liquid temperature difference curve, but also represent the temperature difference constant, which is the temperature difference of each system, assembly, and component when the vehicle is in thermal balance.

[0054] As an exemplary embodiment, the model training method further includes: calibrating the prediction results in the model training process using the liquid-air temperature difference curve change information of the working conditions corresponding to the training samples to obtain a calibrated thermal balance capacity evaluation model.

[0055] In this embodiment, error calibration is specifically as follows: the thermal balance evaluation model built by machine learning method is actually an empirical statistical model. The change constraint of the temperature difference curve is not considered in the prediction process of this model. Therefore, the liquid temperature difference curve may rise significantly at a certain moment or the prediction result may exceed the limit value, so error correction is required.

[0056] As an exemplary embodiment, the use of the liquid-air temperature difference curve change information to calibrate the prediction results in the model training process includes: obtaining multiple liquid-air temperature difference first sampling values ​​in the prediction results; calculating the liquid-air temperature difference average value in the prediction results; calculating the difference between the first sampling value and the liquid-air temperature difference average value; correcting the first sampling value based on the difference and the liquid-air temperature difference curve change information to obtain a calibrated prediction result; correcting the thermal balance capability evaluation model based on the calibrated prediction result to obtain a calibrated thermal balance capability evaluation model

[0057] In this embodiment, the road test method for automobile thermal balance capability stipulates that the liquid temperature difference changes within ±2°C within 4 consecutive minutes. In order to ensure that the liquid temperature difference curve predicted by the machine learning model meets the regulations, the prediction results are processed and the liquid temperature difference value ΔT at each moment and the average value within 4 minutes of the working condition are calculated according to formula (4). If Δt is greater than 2°C, then If Δt is less than -2°C, then

[0058]

[0059] The specific explanation of this embodiment is that during the error calibration process, among the liquid-air temperature difference values ​​predicted within 4 consecutive minutes, the liquid-air temperature difference values ​​that are 2°C higher or lower than the average liquid-air temperature difference within 4 minutes are reassigned:

[0060] When the liquid temperature difference value at a certain moment is higher than the average value of the liquid temperature difference within 4 minutes by 2°C, the liquid temperature difference value at that moment is reassigned as the sum of the average value of the liquid temperature difference and 2°C;

[0061] When the liquid temperature difference value at a certain moment is lower than the average liquid temperature difference of 2°C within 4 minutes, the liquid temperature difference value at that moment will be reassigned as: the difference between the average liquid temperature difference and 2°C.

[0062] It should be noted that the specific numerical values ​​in this embodiment are only used to help explain the embodiment and do not limit the protection scope of the present invention.

[0063] As an exemplary embodiment, the use of the liquid-air temperature difference curve change information to calibrate the prediction results in the model training process includes: obtaining multiple liquid-air temperature difference second sampling values ​​in the prediction results; dividing the second sampling values ​​into multiple groups, wherein each group includes at least two second sampling values; calculating the change rate of the second sampling values ​​in each group; calibrating the change rate of the latter group using the previous group in sequence based on the time sequence, so that the change rate of the latter group changes toward the change rate of the previous group, to obtain a calibrated prediction result; based on the calibrated prediction result, the thermal balance capability evaluation model is corrected to obtain a calibrated thermal balance capability evaluation model.

[0064] In this embodiment, see Figure 3 The road test method for automobile thermal balance capability stipulates that within 4 consecutive minutes, within the same time period, the rising trend of the liquid-air-temperature difference curve in the next time period cannot exceed the rising trend of the liquid-air-temperature difference curve in the previous time period. In order to ensure that the liquid-air-temperature difference curve predicted by the machine learning model meets the regulations, the prediction results are processed. Figure 4 In the two cases shown, the liquid-temperature difference curves rise significantly and require calibration.

[0065] For one working condition, data is recorded once per second, with a total of 240 data points. Starting from the initial moment, the mean values ​​a1, a2, ..., a are calculated for each group of 10 data points. 24 , taking a1, a2, and a3 as examples, if the liquid-temperature difference curve of the a2-a3 segment has an upward trend compared to the a1-a2 segment, then the value of a3 is calibrated to obtain a 3′ , so that it satisfies formula (5), and then the calibration value △T of a set of liquid-air temperature difference data points corresponding to a3 can be calculated according to formula (6): n′ (n=21, ..., 30), and then traverse backward to correct the liquid-temperature difference curve. Based on the error calibration results, the common parameters of the multi-model are screened and optimized again.

[0066]

[0067]

[0068] It should be noted that in addition to screening and tuning common parameters before training the thermal balance evaluation model, the error calibration process can also be placed after the model training is completed. After the model has completed the prediction of the data to be evaluated, the error calibration mechanism is added to perform error calibration on the results predicted by the model.

[0069] As an exemplary embodiment, the model training method further includes normalizing the training samples.

[0070] In this embodiment, for the same sample, the orders of magnitude of each feature parameter are different. In the process of the model performing gradient descent solution based on the training samples, some feature parameters may be ignored due to the different orders of magnitude between the parameters. Therefore, the training samples must be normalized to make the gradient descent solution of the algorithm model accurate and the convergence speed faster than before the feature parameters are normalized, making it easier to find the optimal solution.

[0071] This embodiment uses the min-max method to normalize the data, as shown in formula (7). Where x* represents the normalized data, x represents the original data, and x min Represents the minimum value of all data, x max Indicates the maximum value of all data. Table 2 and Table 3 list the data distribution of some sample data before and after normalization. It can be seen that the distribution range of each feature parameter is consistent after normalization, which improves the data quality.

[0072]

[0073] Table 2 Part of the original sample data

[0074]

[0075] Table 3 Partial normalized sample data

[0076]

[0077] As an exemplary embodiment, the thermal balance capacity assessment model training method further includes: sequentially traversing all combinations of model hyperparameters, and screening the model hyperparameters based on cross-validation.

[0078] In this embodiment, the XGBoost algorithm model contains many hyperparameters, and the optimal solution of the model is achieved by experimental optimization of different hyperparameter values. This embodiment is based on the grid search method, which traverses all combinations of parameters in turn and uses the cross-validation method to screen out the best hyperparameter values. The results are shown in Table 4. 2) is used as the evaluation function of cross-validation to characterize the fitting effect of the model on the training samples and the validation samples, as shown in formula (8), i represents the model prediction value; y represents the average value of all sample dependent variables; y i represents the true value of the dependent variable; n represents the number of samples.

[0079]

[0080] Table 4 Hyperparameter values

[0081]

[0082] In this embodiment, see Figure 4 As shown in the figure, after the general parameters and hyperparameters are initially screened and tuned, the thermal balance capacity evaluation model is trained. Based on the debugged hyperparameters, the sample data other than the test set are fitted to establish a prediction model. The root mean square error (RMSE) is calculated according to formula (9) to evaluate the fitting effect of the model on the training set. It can be seen that the model converges quickly after 100 iterations, with a mean square error of 0.00317, and then tends to be stable. After 388 iterations, the mean square error is 0.00086. In formula (6), f i Represents the model prediction value; y i represents the true value of the dependent variable; n represents the number of samples.

[0083]

[0084] The present invention also provides a method for evaluating the thermal balance capability of a vehicle, see Figure 5 As shown, including:

[0085] S21. Acquire data of the vehicle to be evaluated, wherein the data of the vehicle to be evaluated includes vehicle operating parameters and vehicle fixed parameters.

[0086] S22. Inputting the data of the vehicle to be evaluated into the thermal balance capability evaluation model to obtain an evaluation result of the thermal balance capability of the vehicle to be evaluated.

[0087] In this embodiment, after obtaining the trained evaluation model, the acquired data to be evaluated is input into the evaluation model to obtain the evaluation result. The data to be evaluated includes vehicle operating parameters and vehicle fixed parameters, and the thermal balance capability evaluation model is obtained by training based on the test sample.

[0088] In another embodiment of the present invention, after the test sample is input into the evaluation model trained to convergence to obtain the evaluation result of the test sample, the output result is processed according to formula (10) to solve the true value of the liquid-air temperature difference ΔT, ΔT new* represents the predicted value of the dynamic liquid temperature difference model, ΔT newminIndicates the minimum value of the dynamic liquid temperature difference in the sample data, ΔT newmax It represents the maximum value of the dynamic liquid temperature difference in the sample data, and t represents the current time.

[0089]

[0090] Calculate the allowable ambient temperature T according to formula (11): apr , where T lim Indicates the maximum allowable temperature of the cooling medium, ΔT n represents the true value of the liquid-temperature difference of the nth sample, N represents the number of samples in the test set, and in this embodiment, N is 240.

[0091]

[0092] Taking the maximum torque point condition as an example, the model predicts the vehicle liquid temperature difference change curve within 4 minutes, see Figure 6 It can be seen that the direct prediction results of the model at the initial moment are quite different from the experimental test results, which is greater than 8°C. By introducing the error calibration mechanism, the abnormal values ​​of the direct prediction of the model are corrected, which significantly reduces the prediction difference to less than 4°C, making the overall change of the curve consistent with the experimental test results.

[0093] The prediction results of the allowable ambient temperature are shown in Table 5 below. Compared with the simulation prediction, the prediction model constructed by this method has an accuracy improvement of 16.6%, and the difference with the experimental test results is smaller, less than 2°C; compared with the experimental test, this method does not require test preparation, and the prediction time is shorter, only 0.5h.

[0094] Table 5 Permissible ambient temperature

[0095]

[0096] An embodiment of the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the model training method and / or the vehicle thermal balance capability assessment method in the above-mentioned embodiments.

[0097] The specific numerical values ​​applied in the embodiments of the present invention are only used to explain the specific embodiments and do not constitute a limitation on the protection scope of the present invention.

[0098] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0099] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A model training method suitable for evaluating the thermal balance capability of an automobile, characterized in that: include: Acquire training samples, wherein the training samples include vehicle operating parameters and vehicle fixed parameters of historical thermal balance tests and corresponding thermal balance capability test results; The vehicle operating condition parameters and the vehicle fixed parameters are used as input parameters, and the thermal balance capacity test results are used as output parameters to train the preset model until convergence, so as to obtain a thermal balance capacity evaluation model; The prediction results in the model training process are calibrated using the change information of the liquid-temperature difference curve under the working conditions corresponding to the training samples to obtain a calibrated thermal balance capacity evaluation model. The method of calibrating the prediction results in the model training process by using the liquid-air temperature difference curve change information of the working condition corresponding to the training sample includes: Acquire a plurality of first sample values ​​of liquid-air temperature difference in the prediction result; Calculate the average value of the liquid-temperature difference in the prediction result; Calculate the difference between the first sampling value and the average value of the liquid temperature difference; Based on the difference and the change information of the liquid-temperature-temperature difference curve, the first sampling value is corrected to obtain a calibrated prediction result; Correcting the thermal balance capability evaluation model based on the calibrated prediction result to obtain a calibrated thermal balance capability evaluation model; and / or: The method of calibrating the prediction results in the model training process by using the liquid-air temperature difference curve change information of the working condition corresponding to the training sample includes: Acquire a plurality of second sample values ​​of the liquid-air temperature difference in the prediction result; Dividing the second sampling values ​​into a plurality of groups, wherein each group includes at least two second sampling values; Calculate the rate of change of the second sample value in each group; Based on the time sequence, the change rate of the latter group is calibrated by using the former group in sequence, so that the change rate of the latter group changes toward the change rate of the former group, and a calibrated prediction result is obtained; The thermal balance capability evaluation model is corrected based on the calibrated prediction result to obtain a calibrated thermal balance capability evaluation model.

2. A model training method as claimed in claim 1, characterized in that: The vehicle fixed parameters include: ambient temperature, radiator, intercooler, fan, test bench ambient temperature, coolant flow, coolant heat removal, intake air flow, intercooler heat removal and heat generation parameters; The vehicle operating parameters include: liquid temperature difference, engine speed, torque and power.

3. A model training method as claimed in claim 1, characterized in that: The prediction result is a dynamic liquid temperature difference parameter and / or a temperature difference constant.

4. A model training method as claimed in claim 1, characterized in that: Also includes: The training samples are normalized.

5. A model training method as claimed in claim 4, characterized in that: Also includes: All combinations of model hyperparameters are traversed in turn, and model hyperparameters are screened based on cross-validation.

6. A method for evaluating vehicle thermal balance capability, characterized in that: include: Acquiring vehicle data to be evaluated, wherein the vehicle data to be evaluated includes vehicle operating parameters and vehicle fixed parameters; The data of the vehicle to be evaluated is input into the thermal balance capability evaluation model as described in any one of claims 1 to 5 to obtain an evaluation result of the thermal balance capability of the vehicle to be evaluated, wherein the thermal balance capability evaluation model is based on vehicle operating condition parameters and vehicle fixed parameters as input parameters and is trained using thermal balance capability test results as output parameters.

7. An electronic device, characterized in that: include: at least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the model training method as described in any one of claims 1 to 5 and / or the vehicle thermal balance capability assessment method as described in claim 6.

Images