Vehicle nitrogen oxide emission model training method and system for virtual calibration

Abstract

The present invention belongs to the technical field of vehicle nitrogen oxide emission prediction. In order to solve the problem that the current vehicle nitrogen oxide emission model cannot guide the actual control of minimizing nitrogen oxide emissions of light vehicles, a vehicle nitrogen oxide emission model training method and system for virtual calibration are provided. Among them, the vehicle nitrogen oxide emission model training method for virtual calibration includes determining the correction coefficient corresponding to different driving behavior style types according to the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient corresponding to the driving behavior style type grid, so as to be used for the virtual calibration of vehicle nitrogen oxide emissions. It can improve the prediction result accuracy of the vehicle nitrogen oxide emission model and provide a guiding scheme for the actual control of minimizing nitrogen oxide emissions of light vehicles.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle nitrogen oxide emission prediction, and in particular relates to a vehicle nitrogen oxide emission model training method and system for virtual calibration. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In order to meet the rapid development needs of automotive products under the current situation, virtual calibration technology, as an advanced calibration method and technology, can complete the calibration process without relying on real vehicles and actual test conditions, thereby effectively reducing costs and shortening the development cycle. Among them, high-precision models are the basis and prerequisite for the successful construction and application of virtual calibration systems.

[0004] NOx emission factor is an important parameter used to evaluate and quantify the nitrogen oxide emission level of a vehicle under specific conditions, and is of great significance for building a nitrogen oxide emission model. The prior art provides a method for calculating the instantaneous NOx emission factor, which is based on the NOx volume concentration in the exhaust gas, the engine intake rate and the fuel flow data collected second by second. However, in the process of calculating the instantaneous NOx emission factor, the air-fuel ratio of the gasoline engine is set to a fixed value, which may ignore the instantaneous fluctuation of the air-fuel ratio of the gasoline engine when the engine output power changes, resulting in a certain deviation in the calculation result; at the same time, when calculating the air-fuel ratio of the gasoline engine, effective engine intake rate data is required. This results in a large deviation between the prediction results of the current vehicle nitrogen oxide emission model and the RDE (Real Drive Emission, actual driving pollutant emissions, refers to the pollutant emissions when the vehicle is driving on the actual road) data, and cannot guide the actual control of minimizing nitrogen oxide emissions of light vehicles. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a vehicle nitrogen oxide emission model training method and system for virtual calibration, which can improve the prediction result accuracy of the vehicle nitrogen oxide emission model and provide guidance for the actual control of minimizing nitrogen oxide emissions of light vehicles.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a vehicle nitrogen oxide emission model training method for virtual calibration.

[0008] In one or more embodiments, a vehicle nitrogen oxide emission model training method for virtual calibration is provided, comprising:

[0009] Acquire actual driving pollutant emission test data of each test vehicle, extract corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters and use them as an initial set of characteristic parameters;

[0010] According to the influence degree of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on the nitrogen oxide emission, the key input characteristic parameters are screened out from the initial set of characteristic parameters;

[0011] The extracted driving characteristic parameters are clustered to determine the driving behavior style type, and the vehicle nitrogen oxide emission model corresponding to the driving behavior style type is trained based on the data sample set consisting of the key input characteristic parameters corresponding to the driving behavior style type and the measured value of the vehicle's instantaneous NOx emission factor;

[0012] Based on the vehicle nitrogen oxide emission model corresponding to the driving behavior style type and the known key input characteristic parameters, the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type is obtained;

[0013] According to the relationship between the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type grid, the engine exhaust rate, the engine fuel rate and the correction coefficient, the correction coefficient corresponding to different driving behavior style types is determined for virtual calibration of vehicle nitrogen oxide emissions.

[0014] As an embodiment of the first aspect of the present invention, the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is expressed as follows:

[0015]

[0016] in, is the instantaneous NOx emission factor of the vehicle; is the NOx concentration downstream of the selective catalytic reduction system; is the engine exhaust velocity; is the engine fuel rate; is the engine fuel flow; is the molecular weight of NOx; is the density of gasoline; is the exhaust density; is the correction factor.

[0017] As an implementation mode of the first aspect of the present invention, the engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed and intake temperature; the driving characteristic parameters include VSP (Vehicle Specific Power), vehicle speed and acceleration; the external characteristic parameters include ambient temperature, ambient pressure and slope.

[0018] As an implementation of the first aspect of the present invention, the BorutaShap algorithm is used to screen out key input feature parameters from an initial set of feature parameters.

[0019] As an implementation manner of the first aspect of the present invention, the screened key input feature parameters include engine speed and vehicle speed.

[0020] As an implementation manner of the first aspect of the present invention, the driving behavior style types include calm, normal and aggressive driving styles.

[0021] A second aspect of the present invention provides a vehicle nitrogen oxide emission model training system for virtual calibration.

[0022] In one or more embodiments, a vehicle nitrogen oxide emission model training system for virtual calibration includes:

[0023] A characteristic parameter initial set construction module is used to obtain the actual driving pollutant emission test data of each test vehicle, extract the corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters and use them as the characteristic parameter initial set;

[0024] A key input characteristic parameter screening module, which is used to screen out key input characteristic parameters from the initial set of characteristic parameters according to the degree of influence of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on the nitrogen oxide emission;

[0025] A vehicle nitrogen oxide emission model training module is used to cluster the extracted driving characteristic parameters to determine the driving behavior style type, and train the vehicle nitrogen oxide emission model corresponding to the driving behavior style type based on a data sample set consisting of key input characteristic parameters corresponding to the driving behavior style type and the vehicle's instantaneous NOx emission factor measurement value;

[0026] An instantaneous NOx emission factor calculation module is used to obtain a calculated value of a vehicle instantaneous NOx emission factor corresponding to a driving behavior style type based on a vehicle nitrogen oxide emission model corresponding to the driving behavior style type and known key input characteristic parameters;

[0027] The correction coefficient determination module is used to determine the correction coefficient corresponding to different driving behavior style types based on the relationship between the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type grid, the engine exhaust rate, the engine fuel rate and the correction coefficient, so as to be used for the virtual calibration of the vehicle nitrogen oxide emissions.

[0028] As an implementation of the second aspect of the present invention, in the correction coefficient determination module, the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is expressed as follows:

[0029]

[0030] in, is the instantaneous NOx emission factor of the vehicle; is the NOx concentration downstream of the selective catalytic reduction system; is the engine exhaust velocity; is the engine fuel rate; is the engine fuel flow; is the molecular weight of NOx; is the density of gasoline; is the exhaust density; is the correction factor.

[0031] As an implementation mode of the second aspect of the present invention, the engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed and intake temperature; the driving characteristic parameters include VSP (Vehicle Specific Power), vehicle speed and acceleration; the external characteristic parameters include ambient temperature, ambient pressure and slope.

[0032] As an implementation manner of the second aspect of the present invention, the driving behavior style types include calm, normal and aggressive driving styles.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention utilizes the vehicle nitrogen oxide emission model to associate the correction coefficient (1-losses) with the characteristic variables vehicle speed and engine speed, thereby solving the problem that the operating conditions of RDE tests of different vehicles change dynamically and it is difficult to obtain accurate calculation results using the same correction coefficient. Based on the data sample set formed by the key input characteristic parameters corresponding to different driving behavior style types and the vehicle's instantaneous NOx emission factor measurement values, the vehicle nitrogen oxide emission model corresponding to different driving behavior style types is trained accordingly, thereby improving the accuracy of the emission model so that it can be effectively applied to virtual calibration, providing a scientific and intelligent technical means for building a high-precision virtual calibration emission model in a wide range of scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0036] Figure 1 It is a flow chart of a vehicle nitrogen oxide emission model training method for virtual calibration according to an embodiment of the present invention;

[0037] Figure 2 This is the data preprocessing process of an embodiment of the present invention;

[0038] Figure 3 is the test result of the first test vehicle of the embodiment of the present invention;

[0039] Figure 4 is the test result of the second test vehicle of the embodiment of the present invention;

[0040] Figure 5 is the feature selection result of the second test vehicle in the embodiment of the present invention;

[0041] Figure 6 is the feature selection result of the third test vehicle in the embodiment of the present invention;

[0042] Figure 7 is the feature selection result of the sixth test vehicle of the embodiment of the present invention;

[0043] Figure 8 is the feature selection result of the eighth test vehicle in the embodiment of the present invention;

[0044] Fig. 9 is the feature selection result of the ninth test vehicle in the embodiment of the present invention;

[0045] Fig.10 is the feature selection result of the tenth test vehicle in the embodiment of the present invention;

[0046] Fig.11 It is the performance evaluation of the ASWC-based clustering algorithm of the embodiment of the present invention;

[0047] Fig.12 It is the NOx emission rate evaluation of different driving behavior styles according to an embodiment of the present invention;

[0048] Fig.13 It is a schematic diagram of the structure of a vehicle nitrogen oxide emission model training system for virtual calibration according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0050] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0052] The current method for calculating the instantaneous NOx emission factor is based on the NOx volume concentration in the exhaust gas, the engine intake speed and the fuel flow data collected second by second. The basic calculation formula is shown in formula (1). When one of the engine speed or fuel flow data is invalid, the air-fuel ratio can be used instead, as shown in formulas (2) and (3):

[0053] When the engine intake speed Q aM and engine fuel flow Q fv When both are valid:

[0054]

[0055] In the formula, EF NOx is the instantaneous NOx emission factor of the vehicle (g / s); C NOx（down） is the NOx concentration downstream of the SCR (ppm); Q aM is the engine air intake rate (kg / h); Q fv is the engine fuel flow rate (L / h); ρ fuel is the density of gasoline, set to 0.73 kg / L; M is the molecular weight of NOx. Since a high percentage of NO is converted to NO2 at a faster rate under ambient conditions, its molecular weight is considered to be the same as that of NO2, which is 46 g. 1-losses is a constant coefficient obtained by validating the calculation results using RDE data. ρ exhaust is the exhaust density, which is assumed to be approximately the same as the density of air at 29 g / mol.

[0056] When the engine intake rate Q aM Effective engine fuel flow Q fv When invalid:

[0057]

[0058] In the formula, αis the air-fuel ratio of a gasoline engine.

[0059] When the engine fuel flow Q fv Effective but engine intake rate Q aM When invalid:

[0060]

[0061] According to the currently available RDE test data, the above formula has some disadvantages: on the one hand, α Setting this to a fixed value may ignore α The instantaneous fluctuations when the engine output power changes will cause a certain deviation in the calculation results. α Valid engine air intake velocity data is required.

[0062] Therefore, from the perspective of the working principle of the engine, the embodiment of the present invention updates and calculates EF by using the difference between the engine exhaust rate and the engine fuel rate instead of the intake rate. NOx The formula is shown in formula (4):

[0063] (4)

[0064] In the formula, Q exhaust is the engine exhaust rate (kg / h); Q fuel is the engine fuel rate (kg / h).

[0065] To verify EF NOx To improve the accuracy and effectiveness of the calculation formula, the present invention uses the EF calculated by formula (4) NOx Values ​​and EF collected through RDE testing NOx The measured values ​​are compared. In this process, the correction coefficient (1-losses) in the calculation formula is preset to 1, which means that no calibration is performed at this time. The least squares method is used for fitting, and the coefficient of determination R 2 and the slope of the fitted curve equation to assess EF NOx The overall correlation between the calculated and measured values ​​of . Taking V1 (the first test vehicle) and V2 (the second test vehicle) as examples, Figure 3 shows the corresponding fitting results. It can be seen that the EF of the two test vehicles NOx There is an obvious linear relationship between the calculated and measured values. For V1, the slope of the fitting curve and R 2 are close to the ideal value; however, for the fitting result of V2, R 2 It is only 0.897, which indicates that EF NOxThe accuracy of the calculated values ​​is not always maintained at a high level.

[0066] The parameter K is defined as the EF of a single vehicle NOx Calculated Value and the values ​​measured in the RDE test The slope of the fitting curve between is as follows:

[0067]

[0068] Table 1 summarizes the E FNOx Correlation parameters between calculated and measured values, without considering the correction factor (1-losses), for the corresponding parameters K, R 2 The data in Table 1 show that, except for V1, there are obvious deviations between the calculated and measured values ​​of the remaining 12 test vehicles. Therefore, it is necessary to further calibrate by introducing (1-losses) to ensure that K and R 2 The value of is close to the ideal value of 1. Since there is a correction coefficient (1-losses) in equation (4), its specific value needs to be verified based on the corresponding variables in the RDE data to further improve the accuracy of the calculation model.

[0069] Table 1 Parameters K, R 2 and RMSE

[0070]

[0071] Figure 1 is a flow chart of a method for training a vehicle nitrogen oxide emission model for virtual calibration in an embodiment of the present invention, such as Figure 1 The vehicle nitrogen oxide emission model training method for virtual calibration in the present embodiment shown may include:

[0072] S101, obtaining actual driving pollutant emission test data of each test vehicle, extracting corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters as an initial set of characteristic parameters;

[0073] S102, selecting key input characteristic parameters from the initial set of characteristic parameters according to the degree of influence of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on nitrogen oxide emissions;

[0074] S103, clustering the extracted driving characteristic parameters to determine the driving behavior style type, and training the vehicle nitrogen oxide emission model corresponding to the driving behavior style type according to the data sample set consisting of the key input characteristic parameters corresponding to the driving behavior style type and the measured value of the vehicle's instantaneous NOx emission factor;

[0075] S104. Based on the vehicle nitrogen oxide emission model corresponding to the driving behavior style type and the known key input characteristic parameters, obtain the calculated value of the instantaneous NOx emission factor of the vehicle corresponding to the driving behavior style type;

[0076] S105. According to the relationship between the calculated value of the instantaneous NOx emission factor of the vehicle corresponding to the driving behavior style type, the engine exhaust rate, the engine fuel rate, and the correction coefficient, determine the correction coefficient corresponding to different driving behavior style types for virtual calibration of vehicle nitrogen oxide emissions.

[0077] In this embodiment, the vehicle nitrogen oxide emission model is used to associate the correction coefficient with the characteristic variables vehicle speed and engine speed, solving the problem that it is difficult to obtain accurate calculation results using the same correction coefficient due to the dynamic changes in the operating conditions of different vehicle RDE tests. Based on the data sample set formed by the key input characteristic parameters corresponding to different driving behavior style types and the measured values of the instantaneous NOx emission factor of the vehicle, the vehicle nitrogen oxide emission models corresponding to different driving behavior style types are trained respectively, improving the accuracy of the emission model and enabling it to be effectively applied to virtual calibration, providing a scientific and intelligent technical means for constructing a high-precision virtual calibration emission model in a wide range of scenarios.

[0078] This embodiment is based on the RDE test data of 13 light-duty vehicles and has a total of four test routes. In step S101, the data includes vehicle information (test ID and vehicle ID), location information (longitude and latitude), vehicle speed, acquisition time, and corresponding engine operation and emission data. This data set covers three operating conditions: urban (v ≤ 60 km / h), suburban (60 km / h < v ≤ 90 km / h), and highway (v > 90 km / h).

[0079] The preliminary data processing results show that some important characteristic parameters in the RDE test data are invalid, such as the fuel consumption rate and the nitrogen oxide concentration downstream of the SCR (Selective Catalytic Reduction System). Therefore, to ensure the accuracy of the results, the data should be preprocessed to check its validity before applying the data. The data preprocessing process is as Figure 2 shown.

[0080] In this embodiment, the engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed, and intake temperature; the driving characteristic parameters include VSP (Vehicle Specific Power), vehicle speed, and acceleration; the external characteristic parameters include ambient temperature, ambient pressure, and slope. The initial set of characteristic parameters is shown in Table 2.

[0081] Table 2 Initial set of feature parameters

[0082]

[0083] In step S102, the key input feature parameters are screened out from the initial set of feature parameters using the BorutaShap algorithm, which is a feature selection method that combines the Boruta algorithm and the SHAP value, and is intended to improve the performance and interpretability of the machine learning model.

[0084] The BorutaShap algorithm takes the input feature parameters in Table 2 as input and the importance of each feature parameter as the algorithm output. The algorithm selects the key input feature parameters after 100 iterations. Some of the output results are shown in Figure 2. Figure 3-Figure 10 shown.

[0085] In such Figure 3 - Fig.10 In Figure 1, the feature importance analysis results of 6 of the test vehicles with relatively large RMSE (root mean square error) among the 13 vehicles (as shown in Table 1) are shown: V2, V3, V6, V8, V9, and V10. Figure 3-Figure 10 In , the x-axis is the input feature parameter, and the y-axis is the corresponding Z_score, where Z_score is calculated by dividing the average loss Z by its standard deviation σ. In addition, the boxes of different colors in the figure represent different screening states of the features: the green box represents the accepted variable, the yellow box represents the tentative variable, and the red box represents the rejected variable. Figure 3-Figure 10 It can be observed that the feature importance ranking and the rejected variables are not always the same for different test vehicles. It is worth noting that although VSP is often regarded as a variable closely related to vehicle emissions, the BorutaShap algorithm only selects it as an important feature in four vehicles (V6, V9, V10 and V11). In contrast, acceleration is not widely accepted in the BorutaShap algorithm analysis (all except V1, V4, V7, V12 and V13 are rejected). This is because vehicle speed has been identified as an important determinant of vehicle emissions, making the contribution of acceleration less significant. Observing the results of the BorutaShap algorithm analysis, engine speed, vehicle speed and ambient pressure are widely accepted by the BorutaShap algorithm. These variables will be used as important input feature parameters to correct equation (4) to further optimize the calibration effect of (1-losses).

[0086] In other embodiments, other feature screening algorithms may be selected to screen key input feature parameters according to actual conditions.

[0087] Since the ambient pressure is not a human-controllable variable among the three input characteristic parameters of engine speed, vehicle speed and ambient pressure, the key input characteristic parameters screened out in this embodiment include engine speed and vehicle speed.

[0088] In step S103 , the driving behavior style types include calm, normal, and aggressive driving styles.

[0089] Each data point in the dataset corresponds to a specific driving event, which can be divided into different clusters. To ensure the comparability of classification features, the data was standardized before clustering. Subsequently, ASWC (average silhouette width coefficient) was used to evaluate and compare the clustering performance of three clustering methods: K-means (k-means clustering algorithm), GMM (Gaussian mixture model) and KCM (K-Centroids Method clustering is a centroid-based clustering algorithm) and determine the optimal number of clusters. Fig.11 The ASWC values ​​of various clustering methods are shown when the number of clusters ranges from 2 to 8. Fig.11 As shown in Figure 3, when the number of clusters is 3, the GMM algorithm obtains the highest ASWC value, indicating that this method has better clustering effect than other methods with different clustering configurations. Therefore, GMM is selected for clustering analysis, and the number of clusters is set to 3.

[0090] The input of the GMM clustering algorithm includes the instantaneous speed (m / s) and acceleration (m / s 2 ), the output assigns a number from 1 to 3 to each event, indicating the cluster to which the driving event belongs. The driving behavior style of each cluster is determined by the amplitude of the volatility index within the cluster. The present invention uses six volatility indexes to characterize driving behavior, including speed standard deviation (Speed ​​SD), speed mean absolute deviation (Speed ​​MAD), deceleration standard deviation (Dec SD), deceleration mean absolute deviation (Dec MAD), acceleration standard deviation (Acc SD) and acceleration mean absolute deviation (Acc MAD). Table 3 lists the volatility indexes of the three clusters determined by the GMM algorithm.

[0091] Table 3 Volatility index values ​​of calm, normal and aggressive driving behavior styles

[0092]

[0093] like Fig.12 As shown in the figure, the nitrogen oxide emission level of the aggressive driving cluster is higher than that of the normal and calm driving clusters. This shows that aggressively driven vehicles generally produce higher nitrogen oxide emissions, further indicating that more erratic driving behavior is associated with increased emission rates, and driving behavior style can be used as one of the bases for improving the accuracy of emission models.

[0094] It should be noted here that the clustering method can be implemented using an existing clustering algorithm, which will not be described in detail here.

[0095] The vehicle nitrogen oxide emission model proposes a method of integrating multiple basic machine learning algorithms to build a more powerful prediction model. Specifically, the method is divided into two levels. In the first level, a group of basic machine learning algorithms are first selected, including linear regression, elastic network, support vector regression, decision tree regression, K nearest neighbor regression, adaptive boosting regression, bagging regression, random forest regression, extreme random tree regression, etc. Each basic algorithm is trained independently on the data set and cross-validated to evaluate the performance of these basic learning algorithms in data analysis. In the second level, the method optimizes the combination of the prediction results of these basic learning algorithms. The task of the second level is to determine the combination weights of each basic learning algorithm, that is, to determine the relative importance of different learning algorithms in the final prediction. The second level uses a linear regression method to minimize the overall prediction error of the weighted combination model by weighting the output of each learning algorithm. In this way, this method combines the advantages of multiple machine learning algorithms and can be better than any single basic machine learning algorithm in prediction accuracy.

[0096] In step S105, the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is:

[0097]

[0098] in, is the instantaneous NOx emission factor of the vehicle, g / s; is the NOx concentration downstream of the selective catalytic reduction system, ppm; is the engine exhaust rate, kg / h; is the engine fuel rate, kg / h; is the engine fuel flow rate, L / h; is the molecular weight of NOx, which is considered to be the same as that of NO2, 46g; is the exhaust density, which is assumed to be approximately the same as the air density at 29 g / mol; 1-losses is the correction factor; is the density of gasoline.

[0099] This embodiment uses the RDE test data of the test vehicle to construct a refined correction coefficient (1-losses) model, aiming to optimize the prediction accuracy of the vehicle nitrogen oxide emission model. 2 _improved evaluates and aims to optimize EF NOxThe correction coefficient value. K_improved represents the value of parameter K after calibration by (1-losses), which is defined as follows:

[0100]

[0101] Among them, EF_calculated_improved represents the calculated value of the nitrogen oxide emission factor after (1-losses) calibration, and EF_measured represents the measured value of the nitrogen oxide emission factor.

[0102] The present invention uses multi-stage, multi-objective iterations to adaptively adjust the learning efficiency for different characteristic parameters to correct the NOx emission calculation model for light vehicles under multiple operating conditions and a wide speed range. In addition, a clustering and recognition method for driving behavior style is introduced to further improve the accuracy of the NOx emission model and ensure that the model can meet the high-precision requirements of virtual calibration.

[0103] The present invention divides the initial data set into three sub-data sets based on the clustering results of driving behavior styles, corresponding to calm, normal and intense driving behavior styles. This classification not only improves the applicability of the model under different driving styles, but also provides more accurate data input for the optimization of the correction coefficient in the next step. Secondly, for each sub-data set, the machine learning integrated algorithm is used to solve the correction coefficient (1-losses) to further improve the correction effect of (1-losses). Finally, the calculated value of the corrected nitrogen oxide emission factor is compared with the actual measured value to intuitively reflect the effectiveness and correction effect of the correction coefficient (1-losses).

[0104] According to K_improved and R 2 _improved (as shown in Table 4), the comprehensive modeling method improves EF NOx Compared with the uncorrected model, the EF constructed using the above method NOx The calculation model and the accuracy of the calculated values ​​have been significantly improved. The K_improved of the 13 test vehicles are all stable within the range of 1.00±0.03. Except for V12, the instantaneous EF NOx The R between the calculated and measured values 2 The values ​​are all 0.999, indicating that there is a strong correlation between the two sets of data. It can be concluded that the correction coefficient (1-losses) model constructed by the above method has a good correction effect on formula (4), significantly improving the accuracy of the model and making it further used in virtual calibration.

[0105] Table 4 Parameters K and R between the calculated and measured values ​​of the corrected NOx emission rate2 value

[0106]

[0107] This embodiment also combines the development of driving behavior style clustering and recognition methods. Based on the sub-data sets corresponding to different driving behavior style types, the vehicle nitrogen oxide emission models corresponding to different driving behavior style types are trained accordingly, thereby improving the accuracy of the emission model so that it can be effectively applied to virtual calibration, providing a scientific and intelligent technical means for building high-precision virtual calibration emission models in a wide range of scenarios.

[0108] Fig.13 is a schematic diagram of a vehicle nitrogen oxide emission model training system for virtual calibration in an embodiment of the present invention. Figure 1 The vehicle nitrogen oxide emission model training method for virtual calibration corresponds to Fig.13 As shown, the vehicle nitrogen oxide emission model training system for virtual calibration in this embodiment may include:

[0109] The characteristic parameter initial set construction module 1301 is used to obtain the actual driving pollutant emission test data of each test vehicle, extract the corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters and use them as the characteristic parameter initial set;

[0110] A key input characteristic parameter screening module 1302 is used to screen out key input characteristic parameters from the initial set of characteristic parameters according to the degree of influence of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on the nitrogen oxide emission;

[0111] The vehicle nitrogen oxide emission model training module 1303 is used to cluster the extracted driving characteristic parameters to determine the driving behavior style type, and train the vehicle nitrogen oxide emission model corresponding to the driving behavior style type according to the data sample set consisting of the key input characteristic parameters corresponding to the driving behavior style type and the vehicle's instantaneous NOx emission factor measurement value;

[0112] The instantaneous NOx emission factor calculation module 1304 is used to obtain a calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type based on the vehicle nitrogen oxide emission model corresponding to the driving behavior style type and known key input characteristic parameters;

[0113] The correction coefficient determination module 1305 is used to determine the correction coefficient corresponding to different driving behavior style types based on the relationship between the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type grid, the engine exhaust rate, the engine fuel rate and the correction coefficient, so as to be used for the virtual calibration of the vehicle nitrogen oxide emissions.

[0114] In the characteristic parameter initial set construction module 1301, the engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed and intake temperature; the driving characteristic parameters include VSP (Vehicle Specific Power), vehicle speed and acceleration; the external characteristic parameters include ambient temperature, ambient pressure and slope.

[0115] In the vehicle nitrogen oxide emission model training module 1303, the driving behavior style types include calm, normal and aggressive driving styles.

[0116] In the correction coefficient determination module 1305, the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is expressed as follows:

[0117]

[0118] in, is the instantaneous NOx emission factor of the vehicle; is the NOx concentration downstream of the selective catalytic reduction system; is the engine exhaust velocity; is the engine fuel rate; is the engine fuel flow; is the molecular weight of NOx; is the density of gasoline; is the exhaust density; is the correction factor.

[0119] It should be noted here that Fig.13 The various modules in the vehicle nitrogen oxide emission model training system for virtual calibration are Figure 1 The various steps in the vehicle nitrogen oxide emission model training method for virtual calibration correspond one to one, and the specific implementation process is the same, which will not be repeated here.

[0120] It should be noted that the electronic device given below is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0121] The electronic device includes a central processing unit (CPU), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) or the program loaded from the storage part to the random access memory (RAM). In the RAM, various programs and data required for system operation are also stored. The central processing unit, ROM and RAM are connected to each other through a bus. The input / output (I / O) interface is also connected to the bus.

[0122] The following components are connected to the I / O interface: an input section including a keyboard, a mouse, etc.; an output section including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive as needed so that a computer program read therefrom is installed into the storage section as needed.

[0123] When the central processing unit in the electronic device of this embodiment executes the program, the following is achieved: Figure 1 The steps in the vehicle nitrogen oxide emission model training method for virtual calibration are shown.

[0124] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, the computer program including a computer program for executing Figure 1 In such an embodiment, the computer program can be downloaded and installed from a network through a communication part, and / or installed from a removable medium. When the computer program is executed by a central processing unit, various functions defined in the device of the present application are executed.

[0125] in, Figure 1 The computer program instructions corresponding to the method shown 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 a product 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.

[0126] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A vehicle nitrogen oxide emission model training method for virtual calibration, characterized in that: include: Acquire actual driving pollutant emission test data of each test vehicle, extract corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters and use them as an initial set of characteristic parameters; According to the influence degree of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on the nitrogen oxide emission, the key input characteristic parameters are screened out from the initial set of characteristic parameters; The extracted driving characteristic parameters are clustered to determine the driving behavior style type, and the vehicle nitrogen oxide emission model corresponding to the driving behavior style type is trained based on the data sample set consisting of the key input characteristic parameters corresponding to the driving behavior style type and the measured value of the vehicle's instantaneous NOx emission factor; Based on the vehicle nitrogen oxide emission model corresponding to the driving behavior style type and the known key input characteristic parameters, the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type is obtained; According to the relationship between the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type grid, the engine exhaust rate, the engine fuel rate and the correction coefficient, the correction coefficient corresponding to different driving behavior style types is determined for virtual calibration of vehicle nitrogen oxide emissions; The relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is expressed as follows: in, is the instantaneous NOx emission factor of the vehicle; is the NOx concentration downstream of the selective catalytic reduction system; is the engine exhaust velocity; is the engine fuel rate; is the engine fuel flow; is the molecular weight of NOx; is the density of gasoline; is the exhaust density; is the correction factor.

2. The vehicle nitrogen oxide emission model training method for virtual calibration according to claim 1, characterized in that: The engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed and intake temperature; the driving characteristic parameters include vehicle specific power, vehicle speed and acceleration; the external characteristic parameters include ambient temperature, ambient pressure and slope.

3. The vehicle nitrogen oxide emission model training method for virtual calibration according to claim 1, characterized in that: The BorutaShap algorithm is used to screen out key input feature parameters from the initial set of feature parameters.

4. The vehicle nitrogen oxide emission model training method for virtual calibration according to claim 1 or 3, characterized in that: The key input feature parameters screened out include engine speed and vehicle speed.

5. The vehicle nitrogen oxide emission model training method for virtual calibration according to claim 1, characterized in that: The driving behavior style types include calm, normal and aggressive driving styles.

6. A vehicle nitrogen oxide emission model training system for virtual calibration, characterized in that: include: A characteristic parameter initial set construction module is used to obtain the actual driving pollutant emission test data of each test vehicle, extract the corresponding engine characteristic parameters, driving characteristic parameters and external characteristic parameters and use them as the characteristic parameter initial set; A key input characteristic parameter screening module, which is used to screen out key input characteristic parameters from the initial set of characteristic parameters according to the degree of influence of each characteristic parameter in the initial set of characteristic parameters of each test vehicle on the nitrogen oxide emission; A vehicle nitrogen oxide emission model training module is used to cluster the extracted driving characteristic parameters to determine the driving behavior style type, and train the vehicle nitrogen oxide emission model corresponding to the driving behavior style type based on a data sample set consisting of key input characteristic parameters corresponding to the driving behavior style type and the vehicle's instantaneous NOx emission factor measurement value; An instantaneous NOx emission factor calculation module is used to obtain a calculated value of a vehicle instantaneous NOx emission factor corresponding to a driving behavior style type based on a vehicle nitrogen oxide emission model corresponding to the driving behavior style type and known key input characteristic parameters; A correction coefficient determination module, which is used to determine the correction coefficients corresponding to different driving behavior style types according to the relationship between the calculated value of the vehicle instantaneous NOx emission factor corresponding to the driving behavior style type grid, the engine exhaust rate, the engine fuel rate and the correction coefficient, so as to be used for virtual calibration of vehicle nitrogen oxide emissions; In the correction coefficient determination module, the relationship between the calculated value of the vehicle instantaneous NOx emission factor, the engine exhaust rate, the engine fuel rate and the correction coefficient is expressed as follows: in, is the instantaneous NOx emission factor of the vehicle; is the NOx concentration downstream of the selective catalytic reduction system; is the engine exhaust velocity; is the engine fuel rate; is the engine fuel flow; is the molecular weight of NOx; is the density of gasoline; is the exhaust density; is the correction factor.

7. The vehicle nitrogen oxide emission model training system for virtual calibration according to claim 6, characterized in that: The engine characteristic parameters include four variables: engine coolant temperature, exhaust temperature, engine speed and intake temperature; the driving characteristic parameters include vehicle specific power, vehicle speed and acceleration; the external characteristic parameters include ambient temperature, ambient pressure and slope.

8. The vehicle nitrogen oxide emission model training system for virtual calibration according to claim 6, characterized in that: The driving behavior style types include calm, normal and aggressive driving styles.

Images