Control method of engine, control device of engine, and electronic device

Abstract

The application provides a control method of an engine, a control device of the engine and an electronic device. The method comprises: obtaining engine sample data associated with a target scene, classifying the data and calculating corresponding correlation degrees; constructing a plurality of N-dimensional sample feature data comprising A-dimensional independent variables, B-dimensional characteristic creation variables and C-dimensional characteristic interaction variables, wherein A, B and C are selected according to the correlation degrees from high to low, and N=A+B+C; inputting the plurality of N-dimensional sample feature data into a data prediction model to obtain an operation result to determine optimal N-dimensional sample feature data; inputting the optimal N-dimensional real feature data into an ECU, and determining a target real output value of the target scene according to the optimal N-dimensional real feature data, so as to control the engine to operate by using the target real output value. The method reduces the dimension of the data set, captures key information of the data, reduces the memory occupation of the ECU, and realizes more accurate control of the engine.

Description

Technical Field

[0001] This application relates to the field of engine control, and more specifically, to an engine control method, engine control device, and electronic equipment. Background Technology

[0002] When performing ECU integration, it's crucial to ensure that model performance remains largely unchanged. Data processing often involves independent variables, feature-created variables, and non-independent variables derived from feature interactions. However, excessively high dimensionality of input variables increases the algorithm's time and space complexity, generating unnecessary feature overhead and noise, thus reducing model robustness. Furthermore, excessive memory consumption can lead to performance instability in the ECU integration model, impacting the normal operation of the ECU. Summary of the Invention

[0003] The main objective of this application is to provide an engine control method, an engine control device, and an electronic device, so as to at least solve the problem that the excessive number of variables processed by the ECU leads to excessive time and space complexity of the algorithm and excessive memory consumption of the ECU.

[0004] To achieve the above objectives, according to one aspect of this application, a method for controlling an engine is provided, wherein the engine is equipped with an ECU, comprising: acquiring engine sample data associated with a target scene, and classifying the engine sample data into independent variables and non-independent variables, wherein the independent variables are variables directly collected by sensors, and the non-independent variables include feature creation variables and feature interaction variables, wherein the feature creation variables are new feature variables obtained by transforming and / or combining at least one of the independent variables, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables; acquiring the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene; constructing multiple N-dimensional sample feature data, wherein one N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables, and any two N-dimensional sample feature data corresponding to A, B, and C dimensions are defined as follows: At least one of B and C is different, wherein the A-dimensional independent variable is composed of the top A independent variables with the highest to lowest correlation among the multiple independent variables, the B-dimensional feature creation variable is composed of the top B feature creation variables with the highest to lowest correlation among the multiple feature creation variables, and the C-dimensional feature interaction variable is composed of the top C feature interaction variables with the highest to lowest correlation among the multiple feature interaction variables, wherein N = A + B + C, and N has at least one value; multiple N-dimensional sample feature data are input into a data prediction model for calculation to obtain the corresponding mean error and loss function, and the optimal N-dimensional sample feature data is determined at least based on the mean error and the loss function; the optimal N-dimensional true feature data is sent to the ECU so that the ECU determines the target true output value of the target scenario based on the optimal N-dimensional true feature data, and uses the target true output value to control the engine operation; wherein A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C of the optimal N-dimensional sample feature data.

[0005] Optionally, determining the optimal N-dimensional sample feature data based at least on the mean error and the loss function includes: acquiring the type of the data prediction model, wherein the type of the data prediction model is at least one of the following: GBDT model, XGBoost model, BP neural network model, TDN neural network model, LSTM neural network model, GRU neural network model, and SVM model; and determining the optimal N-dimensional sample feature data based on the type of the data prediction model, the mean error, and the loss function.

[0006] Optionally, determining the optimal N-dimensional sample feature data based at least on the mean error and the loss function includes: determining the final integration location of the prediction model, wherein the final integration location is the cloud or the ECU; and determining the optimal N-dimensional sample feature data based on the final integration location of the prediction model, the mean error, and the loss function.

[0007] Optionally, determining the optimal N-dimensional sample feature data based on the final ensemble position of the prediction model, the mean error, and the loss function includes: determining a first weight of the final ensemble position of the prediction model; determining a second weight of the mean error and the loss function, wherein the first weight and the second weight represent the degree of influence on the optimal selection result of the N-dimensional sample feature data; and determining the optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, the mean error, and the loss function.

[0008] Optionally, obtaining the correlation degree between any one of the plurality of independent variables, any one of the plurality of feature creation variables, and any one of the plurality of feature interaction variables and the target sample output value of the target scene includes: using a formula Determine the degree of correlation. Let x be the sample mean of the variable. Let x be the sample mean of variable y, x represent the independent variable, the feature creation variable, or the feature interaction variable, and y represent the target sample output value of the target scene.

[0009] Optionally, after obtaining engine sample data associated with the target scenario, and before classifying the engine sample data into independent variables and dependent variables, the method further includes: performing data cleaning on the engine sample data associated with the target scenario to obtain cleaned engine sample data, wherein the data cleaning includes deleting missing data, deleting abnormal data, and deleting inconsistent data.

[0010] Optionally, the target scenario is a vehicle load estimation scenario. Engine sample data associated with the target scenario is acquired, and the engine sample data is classified into independent variables and non-independent variables. The non-independent variables include feature creation variables and feature interaction variables, including: acquiring engine sample data associated with the vehicle load estimation scenario; classifying the engine sample data into the independent variables, the feature creation variables, and the feature interaction variables; wherein, the independent variables include: throttle, vehicle speed, engine speed, fuel injection quantity, torque, transient intake air volume, coolant temperature, oil pressure, oil temperature, and gear position; the feature creation variables include: instantaneous fuel consumption, gear ratio, acceleration, wheel-side force, calculated vehicle load, and air resistance; the feature interaction variables include: vehicle speed multiplied by engine speed, vehicle speed multiplied by gear position, engine speed multiplied by gear position, vehicle speed multiplied by engine speed multiplied by gear position, engine speed squared, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, coolant temperature after cooling, and oil temperature minus coolant temperature.

[0011] Optionally, the target scenario is a NOx emission prediction scenario. Engine sample data associated with the target scenario is acquired, and the engine sample data is classified into independent variables and dependent variables. The dependent variables include feature creation variables and feature interaction variables. This includes: acquiring engine sample data associated with the NOx emission prediction scenario; classifying the engine sample data into the independent variables, the feature creation variables, and the feature interaction variables; wherein the independent variables include: after-intercooler temperature, after-intercooler pressure, transient exhaust gas flow rate, etc. Throttle, vehicle speed, fuel injection quantity, coolant temperature, oil pressure, and oil temperature; the feature creation variables include: power, differential of cold-cooled pressure, differential of cold-cooled temperature, differential of intake air volume, steady-state NOx emission, fresh intake air volume, air-fuel ratio, and fuel consumption; the feature interaction variables include: cold-cooled pressure multiplied by cold-cooled temperature, square of cold-cooled pressure, square of cold-cooled temperature, cold-cooled temperature multiplied by coolant temperature multiplied by oil temperature, cold-cooled temperature multiplied by coolant temperature, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, and coolant temperature divided by cold-cooled temperature.

[0012] According to another aspect of this application, an engine control device is provided, wherein an ECU is installed in the engine, comprising: a first acquisition unit, configured to acquire engine sample data associated with a target scene, and classify the engine sample data into independent variables and non-independent variables, wherein the non-independent variables include feature creation variables and feature interaction variables, wherein the independent variables are variables directly collected by sensors, the feature creation variables are new variables obtained by transforming and / or combining at least one of the independent variables, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables; a second acquisition unit, configured to acquire the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene; and a construction unit, configured to construct multiple N-dimensional sample feature data, wherein one N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables, and any two N-dimensional sample feature data correspond to A, B, and C-dimensional feature interaction variables. At least one of B and C is different, wherein the A-dimensional independent variable is composed of the top A independent variables with the highest to lowest correlation among the multiple independent variables, the B-dimensional feature creation variable is composed of the top B feature creation variables with the highest to lowest correlation among the multiple feature creation variables, and the C-dimensional feature interaction variable is composed of the top C feature interaction variables with the highest to lowest correlation among the multiple feature interaction variables, wherein N = A + B + C, and N has at least one value; the calculation unit is used to input the multiple N-dimensional sample feature data into the data prediction model for calculation to obtain the corresponding error mean and loss function, and to determine the optimal N-dimensional sample feature data based at least on the error mean and the loss function; the sending unit is used to send the optimal N-dimensional true feature data to the ECU, so that the ECU determines the target true output value of the target scenario based on the optimal N-dimensional true feature data, and uses the target true output value to control the engine operation; wherein A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C of the optimal N-dimensional sample feature data.

[0013] According to another aspect of this application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including a control method for performing any of the aforementioned engines.

[0014] By applying the technical solution of this application, the impact of independent variables, feature creation variables, and feature interaction variables on the performance of the data prediction model is comprehensively considered. The correlation between the independent variables, feature creation variables, and feature interaction variables and the target scene is calculated respectively. Multiple N-dimensional feature data combination schemes are selected and input into the data prediction model. The N-dimensional feature data consists of independent variables selected from high to low correlation in dimension A, feature creation variables selected from high to low correlation in dimension B, and feature interaction variables selected from high to low correlation in dimension C. The data prediction model calculates and outputs the mean error and loss function corresponding to the multiple N-dimensional feature data combination schemes. The optimal N-dimensional feature data combination scheme is selected based on the magnitude of the mean error and loss function. By inputting the optimal N-dimensional feature data combination scheme into the ECU for calculation, engine control can be achieved. This reduces the dimensionality of the dataset, captures key data information, improves model accuracy, and reduces ECU memory usage. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0016] Figure 1 A hardware structure block diagram of a mobile terminal for executing an engine control method according to an embodiment of this application is shown;

[0017] Figure 2 A schematic flowchart of an engine control method according to an embodiment of this application is shown;

[0018] Figure 3 A flowchart illustrating the process of determining optimal N-dimensional sample feature data according to an embodiment of this application is shown.

[0019] Figure 4 A flowchart illustrating a feature input dimensionality reduction selection method provided according to an embodiment of this application is shown.

[0020] Figure 5 A schematic diagram showing the output results of a data prediction model in a vehicle load estimation scenario provided according to an embodiment of this application is illustrated.

[0021] Figure 6 A structural block diagram of an engine control device according to an embodiment of this application is shown. Detailed Implementation

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] For ease of description, the following explains some of the nouns or terms used in the embodiments of this application:

[0026] ECU (Engine Control Unit): The primary application of the ECU is to control the operation and performance of the engine. It monitors and adjusts parameters such as fuel supply, ignition timing, and exhaust treatment to ensure efficient and reliable engine operation under various conditions. The ECU can also monitor and diagnose engine faults and provide relevant warnings to the driver. In modern automobiles, the ECU also controls and coordinates other systems and components, such as the transmission, braking system, and air conditioning system. Therefore, the ECU holds a very important position in automotive engineering and is one of the key control components for vehicle performance and safety.

[0027] Physical principles refer to the fundamental laws and principles existing in nature, mainly including mechanics, thermodynamics, electromagnetism, optics, and quantum physics. These physical principles are obtained through experiments and observations, and are described and expressed using mathematical language. The study and application of physical principles are of great significance for understanding natural phenomena and developing science and technology.

[0028] Feature creation variables refer to generating new feature variables based on existing feature data through certain transformations or combinations. This can be achieved through mathematical operations, logical operations, or other methods to improve the model's predictive ability or understanding of the data.

[0029] Feature interaction variables: These refer to the interaction of two or more features to generate new feature variables. This method can be used to capture the relationships between features and improve the predictive ability of the model. For example, two features can be multiplied or divided, or other forms of interaction can be performed to generate new feature variables.

[0030] As described in the background section, in the prior art, when integrating ECU models, excessively large input variables can lead to increased time and space complexity of the algorithm, unnecessary feature overhead and noise, and reduced model robustness. To address the problem that excessively large input variables can lead to increased time and space complexity of the algorithm, unnecessary feature overhead and noise, and reduced model robustness, embodiments of this application provide an engine control method, an engine control device, and an electronic device.

[0031] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0032] The methods and embodiments provided in this application can be executed in a mobile terminal, computer terminal, or similar computing device; the aforementioned mobile terminal, computer terminal, or similar computing device communicates with an ECU. Taking operation on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for an engine control method according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0033] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the engine control method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0034] This embodiment provides a control method for an engine running on a mobile terminal, computer terminal, or similar computing device. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0035] Figure 2 This is a flowchart of an engine control method according to an embodiment of this application. Figure 2 As shown, the method includes the following steps:

[0036] Step S201: Obtain engine sample data associated with the target scene, and classify the engine sample data into independent variables and non-independent variables. The independent variables are variables directly collected by sensors, and the non-independent variables include feature creation variables and feature interaction variables. The feature creation variables are new feature variables obtained by transforming and / or combining at least one independent variable, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables.

[0037] Among them, the target scenario can be other target scenarios such as vehicle load estimation scenario, original NOx prediction scenario, etc.

[0038] Feature data determines the upper limit of the model. Machine learning algorithms model and predict through data features. Therefore, it is necessary to further increase the expressive power of features through feature creation, especially by designing new features in combination with vehicle load estimation, to increase the interpretability and versatility of the model.

[0039] For example: Based on the correlation between load and fuel consumption, a feature creation variable is obtained: instantaneous fuel consumption; calculated using the formula: The instantaneous fuel consumption is obtained from the formula. Here, n represents instantaneous fuel consumption, and n represents engine speed. This refers to the amount of fuel injected into the engine during the cycle.

[0040] Based on engine speed and tire speed, a variable speed ratio can be created, representing the product of the gearbox gear ratio and the rear axle speed ratio. This ratio serves to reduce speed and increase torque. Load capacity is correlated with the speed ratio; therefore, the formula... The speed ratio can be obtained; where Engine speed, For vehicle speed, Where is the tire radius, and ispeed is the speed ratio.

[0041] Specifically, to fully extract the information contained in the feature data, this scheme uses polynomial feature interaction to generate new features, with the polynomial degree being 2, and only features of the same degree interacting. For feature interactions of cooled temperature, water temperature, and engine oil temperature, division and subtraction interactions are added in addition.

[0042] For example: 1) Interaction between vehicle speed, engine speed, and gear.

[0043] 2) Interaction between oil pressure, oil temperature, and water temperature.

[0044] 3) The interaction between pressure and temperature after cooling.

[0045] 4) The temperature after cooling interacts with the water temperature and engine oil temperature.

[0046] Step S202: Obtain the correlation degree between any one of multiple independent variables, any one of multiple feature creation variables, and any one of multiple feature interaction variables and the target sample output value of the target scene;

[0047] Among them, correlation is used to measure the degree of correlation or mutual influence between variables;

[0048] Specifically, by calculating the correlation between the independent variables, feature creation variables, and feature interaction variables and the target sample output value of the target scene, the degree of mutual influence between each variable and the target sample output value of the target scene can be obtained.

[0049] Step S203: Construct multiple N-dimensional sample feature data. Each N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. The A-dimensional independent variables are composed of the top A independent variables with the highest correlation from the lowest. The B-dimensional feature creation variables are composed of the top B feature creation variables with the highest correlation from the lowest. The C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest correlation from the lowest. N = A + B + C, and N has at least one value.

[0050] In practice, the value of N is generally between 5 and 15 dimensions. If the selected dimension is too small, the accuracy of the data cannot be guaranteed. If the selected dimension is too large, the purpose of dimensionality reduction cannot be achieved. Selecting a dataset between 5 and 15 dimensions can ensure both a good dimensionality reduction effect and the accuracy of the data.

[0051] Step S204: Input multiple N-dimensional sample feature data into the data prediction model for calculation to obtain the corresponding mean error and loss function, and determine the optimal N-dimensional sample feature data based at least on the mean error and loss function.

[0052] Step S205: The optimal N-dimensional true feature data is sent to the ECU so that the ECU can determine the target true output value of the target scenario based on the optimal N-dimensional true feature data, and control the engine operation using the target true output value; wherein, A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C in the optimal N-dimensional sample feature data.

[0053] This embodiment comprehensively considers the impact of independent variables, feature creation variables, and feature interaction variables on the performance of the data prediction model. It calculates the correlation between the independent variables, feature creation variables, and feature interaction variables and the target scene. Multiple N-dimensional feature data combination schemes are selected and input into the data prediction model. These N-dimensional feature data consist of independent variables selected from high to low correlation in dimension A, feature creation variables selected from high to low correlation in dimension B, and feature interaction variables selected from high to low correlation in dimension C. The data prediction model calculates and outputs the mean error and loss function corresponding to the multiple N-dimensional feature data combination schemes. The optimal N-dimensional feature data combination scheme is selected based on the magnitude of the mean error and loss function. By inputting the optimal N-dimensional feature data combination scheme into the ECU for calculation, engine control can be achieved. This reduces the dimensionality of the dataset while capturing key data information, improving model accuracy, and reducing ECU memory usage.

[0054] In the specific implementation process, step S204 above determines the optimal N-dimensional sample feature data based at least on the mean error and the loss function, including: obtaining the type of data prediction model, wherein the type of data prediction model is at least one of the following: GBDT model, XGBoost model, BP neural network model, TDN neural network model, LSTM neural network model, GRU neural network model and SVM model; and determining the optimal N-dimensional sample feature data based on the type of data prediction model, the mean error and the loss function.

[0055] Among them, the GBDT model: Gradient Boosting Decision Tree Model, improves prediction performance by iteratively training multiple decision tree models.

[0056] XGBoost model: Extreme Gradient Boosting model, is an improved version of GBDT, which uses some optimization strategies to improve model performance.

[0057] BP Neural Network Model: Backpropagation Neural Network Model is a basic feedforward neural network model that optimizes network parameters through the backpropagation algorithm.

[0058] TDN Neural Network Model: Temporal Difference Neural Network Model, a neural network model specifically designed for processing time-series data.

[0059] LSTM Neural Network Model: Long Short-Term Memory Neural Network Model, a special type of recurrent neural network model that can better handle long sequence data.

[0060] GRU Neural Network Model: Gated Recurrent Unit Neural Network Model, also a type of recurrent neural network model used to process time-series data.

[0061] SVM model: Support Vector Machine model, is a supervised learning model used for classification and regression analysis.

[0062] Among them, GBDT is a tree-based regression model with high prediction accuracy and good generalization performance. BP (BackPropagation Network) is a commonly used machine learning modeling method with fast computation speed and minimal memory usage. TDN (Time Delay Network) is the simplest time series neural network, which can evaluate whether the load estimation has time series properties and then perform time series data prediction. SVM model's final result only depends on support vectors, has moderate memory usage, and good robustness.

[0063] The optimal N-dimensional sample feature data is determined based on the type of data prediction model, the mean error, and the loss function, including the following implementation methods:

[0064] When the data prediction model is GBDT, the mean error and loss function are the first values; when the data prediction model is BP, the mean error and loss function are the second values. The second value is greater than the first value. Considering memory, even if the second value is greater than the first value, the optimal N-dimensional sample feature data when the data prediction model is BP is ultimately selected.

[0065] When the data prediction model is GBDT, the mean error and the loss function are the first values; when the data prediction model is BP, the mean error and the loss function are the second values. Since the first value is greater than the second value, and accuracy is given priority, even if the first value is greater than the second value, the optimal N-dimensional sample feature data of the data prediction model GBDT is ultimately selected.

[0066] Specifically, the optimal N-dimensional sample feature data should be determined at least based on the mean error and the loss function, such as... Figure 3 The steps shown are as follows:

[0067] Step S301: Determine the final integration location of the prediction model, which may be either the cloud or the ECU.

[0068] Step S302: Determine the optimal N-dimensional sample feature data based on the final ensemble location, mean error, and loss function of the prediction model.

[0069] The solution requires integrating two suitable models for estimating the vehicle load in the target scenario. The model must be integrated into both the cloud (high accuracy requirement, average error less than 0.1t, no requirements for computing power and memory) and the ECU (moderate accuracy, average error less than 1t, requirements for computing power and memory). Specifically, the model selection depends on whether it's integrated in the cloud or on the ECU. The integration location depends on the amount of memory required, and the selection of the data prediction model determines the size of N in the N-dimensional sample feature data.

[0070] More specifically, the optimal N-dimensional sample feature data is determined based on the final ensemble position of the prediction model, the mean error, and the loss function. This includes: determining a first weight for the final ensemble position of the prediction model; determining a second weight for the mean error and the loss function, where the first and second weights represent the degree of influence on the optimal selection result of the N-dimensional sample feature data; and determining the optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, and the mean error and the loss function.

[0071] The determination of optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, and the mean error and loss function includes the following implementation methods:

[0072] For example, when integrated into the ECU, the first weight is 0.8 and the second weight is 0.2, which means that the memory size has a relatively large impact on the result. Therefore, N-dimensional sample feature data with larger mean error and loss function but lower dimensionality are preferred.

[0073] For example, when integrated in the cloud, the first weight is 0.2 and the second weight is 0.8, meaning that the impact of memory size on the result is relatively small. Therefore, N-dimensional sample feature data with smaller mean error and loss function but higher dimensionality are preferred.

[0074] In some embodiments, step S202 above, obtaining the correlation degree between any one of multiple independent variables, any one of multiple feature creation variables, and any one of multiple feature interaction variables and the target sample output value of the target scene, includes: using a formula Determine the degree of correlation. For variables The sample mean, For variables The sample mean, This refers to independent variables, feature creation variables, or feature interaction variables. This represents the target sample output value for the target scene.

[0075] This method uses the Pearson coefficient method to calculate the correlation between feature data and the output target value of the target scene. This correlation serves only as a reference before being input into the model. The model is based on multiple N-dimensional sample feature data, and the results of the prediction model determine the dominant role of the feature data in the target scene. Dimensionality reduction can be achieved simply by calculating the dominant feature data combination.

[0076] Furthermore, after obtaining engine sample data associated with the target scenario, and before classifying the engine sample data into independent and dependent variables, the method further includes: performing data cleaning on the engine sample data associated with the target scenario to obtain cleaned engine sample data, wherein data cleaning includes deleting missing data, deleting abnormal data, and deleting inconsistent data.

[0077] Specifically, after obtaining engine sample data associated with the target scenario, data analysis and processing cannot be performed directly because the obtained data may not be completely accurate. Directly using this data for model building may introduce significant biases. The purpose of feature data cleaning is to remove missing, abnormal, and inconsistent data sample points. Only in this way can the validity and accuracy of the data be determined.

[0078] Specifically, the target scenario is a vehicle load estimation scenario. Engine sample data associated with the target scenario is acquired and classified into independent variables and non-independent variables. Non-independent variables include feature creation variables and feature interaction variables, including: acquiring engine sample data associated with the vehicle load estimation scenario; classifying engine sample data into independent variables, feature creation variables, and feature interaction variables; where independent variables include: throttle, vehicle speed, engine speed, fuel injection quantity, torque, transient intake air volume, coolant temperature, oil pressure, oil temperature, and gear position; feature creation variables include: instantaneous fuel consumption, gear ratio, acceleration, wheel-side force, calculated vehicle load, and air resistance; feature interaction variables include: vehicle speed multiplied by engine speed, vehicle speed multiplied by gear position, engine speed multiplied by gear position, vehicle speed multiplied by engine speed multiplied by gear position, engine speed squared, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, coolant temperature after cooling, and oil temperature minus coolant temperature.

[0079] The target scenario is vehicle load estimation. The specific selection of independent variables, feature creation variables, and feature interaction variables is shown in Table 1:

[0080] Table 1 Input characteristics for vehicle load estimation

[0081]

[0082]

[0083] Specifically, the target scenario is the original NOx emission prediction scenario. Engine sample data associated with the target scenario is acquired and classified into independent variables and non-independent variables. Non-independent variables include feature creation variables and feature interaction variables, including: acquiring engine sample data associated with the original NOx emission prediction scenario; classifying engine sample data into independent variables, feature creation variables, and feature interaction variables; where independent variables include: intercooler temperature, intercooler pressure, transient exhaust gas flow, throttle, vehicle speed, fuel injection quantity, coolant temperature, oil pressure, and oil temperature; feature creation variables include: power, differential of intercooler pressure, differential of intercooler temperature, differential of intake air volume, steady-state original NOx emission, fresh intake air volume, air-fuel ratio, and fuel consumption; feature interaction variables include: intercooler pressure multiplied by intercooler temperature, intercooler pressure squared, intercooler temperature squared, intercooler temperature multiplied by coolant temperature multiplied by oil temperature, intercooler temperature multiplied by coolant temperature, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, and coolant temperature divided by intercooler temperature.

[0084] The target scenario is the original NOx prediction scenario. The specific selection of independent variables, feature creation variables, and feature interaction variables is shown in Table 2:

[0085] Table 2. Input characteristics for original NOx prediction

[0086]

[0087] Regarding the selection method for feature input dimensionality reduction, such as... Figure 4 As shown, the relevance of independent variables, feature creation, and feature interaction is calculated and arranged in descending order to obtain the independent feature variable library, the feature creation variable library, and the feature interaction variable library. The total number of features is equal to the sum of the feature variables in the independent feature variable library, the feature creation variable library, and the feature interaction variable library. Based on the independent feature variable library, the feature creation variable library, and the feature interaction variable library, different schemes are formed for feature extraction to obtain the optimal combination scheme of independent variables, feature creation, and feature interaction.

[0088] To enable those skilled in the art to better understand the technical solution of this application, the implementation process of the engine control method of this application will be described in detail below with reference to specific embodiments.

[0089] This embodiment relates to a method for selecting data dimensionality reduction in a vehicle load scenario:

[0090] ECU integration requires data dimensionality reduction while maintaining basic model performance. This involves considering the impact of independent variables, feature creation, and feature interactions on model performance under vehicle load scenarios. To study the effect of dimensionality reduction on model performance, 10-dimensional and 5-dimensional data are extracted from the filtered data. Highly relevant features are then selected from the three categories of features: independent variables, feature creation, and feature interactions. The specific approach is as follows:

[0091] 1) Retain 10-dimensional feature data and select 10 independent variables.

[0092] 2) Retain 10-dimensional feature data, with 6 independent variables, 3 dimensions for feature creation, and 1 dimension for feature interaction.

[0093] 3) Retain 10-dimensional feature data, with 5 independent variables, 3 dimensions for feature creation, and 2 dimensions for feature interaction.

[0094] 4) Retain 10-dimensional feature data, with 4 independent variables, 3 dimensions for feature creation, and 3 dimensions for feature interaction.

[0095] 5) Retain 5-dimensional feature data, and take 5 independent variables.

[0096] 6) Retain 5-dimensional feature data, with 3 independent variables and 2-dimensional feature creation.

[0097] 7) Retain 5-dimensional feature data, with 2 independent variables and 3-dimensional feature creation.

[0098] The above seven schemes each have a model. Taking the GBDT model as an example, the results are as follows: Figure 5As shown in the figure, independent variables play a dominant role; with the same total number of variables, the more independent variables, the higher the model accuracy. Feature creation and feature interaction also have an effect, but their impact is not commensurate with the relevance and can only be used as a reference. By inputting the actual independent variables that play a dominant role into the ECU, an accurate vehicle load calculation value can be obtained. This reduces the dimensionality of the dataset, captures key data information, improves model accuracy, reduces ECU memory usage, and achieves more precise engine control.

[0099] This application also provides an engine control device. It should be noted that the engine control device of this application embodiment can be used to execute the engine control method provided in this application embodiment. This device is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0100] The following describes the engine control device provided in the embodiments of this application.

[0101] Figure 6 This is a schematic diagram of an engine control device according to an embodiment of this application. Figure 6 As shown, the device includes:

[0102] The first acquisition unit 61 is used to acquire engine sample data associated with the target scene and classify the engine sample data into independent variables and non-independent variables. The non-independent variables include feature creation variables and feature interaction variables. The independent variables are variables directly collected by sensors, the feature creation variables are new variables obtained by transforming and / or combining at least one independent variable, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables.

[0103] The second acquisition unit 62 is used to acquire the correlation degree between any one of multiple independent variables, any one of multiple feature creation variables, and any one of multiple feature interaction variables and the target sample output value of the target scene;

[0104] Construction unit 63 is used to construct multiple N-dimensional sample feature data. An N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. The A-dimensional independent variables are composed of the top A independent variables with the highest correlation from the lowest, the B-dimensional feature creation variables are composed of the top B feature creation variables with the highest correlation from the lowest, and the C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest correlation from the lowest. N = A + B + C, and N has at least one value.

[0105] The operation unit 64 is used to input multiple N-dimensional sample feature data into the data prediction model for operation, and obtain the corresponding error mean and loss function, and at least determine the optimal N-dimensional sample feature data based on the error mean and loss function.

[0106] The sending unit 65 is used to send the optimal N-dimensional true feature data to the ECU, so that the ECU can determine the target true output value of the target scenario based on the optimal N-dimensional true feature data, and use the target true output value to control the engine operation; wherein, A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C in the optimal N-dimensional sample feature data.

[0107] As an optional approach, the processing unit includes a first acquisition module and a first determination module;

[0108] The first acquisition module is used to acquire the types of data prediction models, wherein the types of data prediction models are at least one of the following: GBDT model, XGBoost model, BP neural network model, TDN neural network model, LSTM neural network model, GRU neural network model and SVM model; the first determination module is used to determine the optimal N-dimensional sample feature data based on the types of data prediction models, the mean error and the loss function.

[0109] In one alternative scheme, the computing unit further includes a second determining module and a third determining module;

[0110] The second determining module is used to determine the final integration location of the prediction model, which is either in the cloud or at the ECU. The third determining module is used to determine the optimal N-dimensional sample feature data based on the final integration location of the prediction model, the mean error, and the loss function.

[0111] In one alternative, the third determining module further includes a first determining submodule, a second determining submodule, and a third determining submodule;

[0112] The first determining submodule is used to determine the first weight of the final ensemble position of the prediction model; the second determining submodule is used to determine the second weight of the mean error and the loss function, where the first weight and the second weight represent the degree of influence on the optimal selection result of the N-dimensional sample feature data; the third determining submodule is used to determine the optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, and the mean error and the loss function.

[0113] In one alternative, the second acquisition unit includes a fourth determining module;

[0114] The fourth determination module is used to apply formulas. Determine the degree of correlation. For variables The sample mean, For variables The sample mean, This refers to independent variables, feature creation variables, or feature interaction variables. This represents the target sample output value for the target scene.

[0115] In one alternative embodiment, the device further includes a cleaning unit;

[0116] The cleaning unit is used to clean the engine sample data associated with the target scenario after acquiring the engine sample data and before classifying the engine sample data into independent variables and dependent variables, so as to obtain cleaned engine sample data. The data cleaning includes deleting missing data, deleting abnormal data, and deleting inconsistent data.

[0117] In one alternative, the third acquisition unit further includes a second acquisition module and a first classification module;

[0118] The second acquisition module is used to acquire engine sample data associated with the vehicle load estimation scenario; the first classification module is used to classify the engine sample data into independent variables, feature creation variables, and feature interaction variables; among them, independent variables include: throttle, vehicle speed, engine speed, fuel injection quantity, torque, transient intake air volume, water temperature, oil pressure, oil temperature, and gear; feature creation variables include: instantaneous fuel consumption, gear ratio, acceleration, wheel-side force, calculated vehicle load, and air resistance; feature interaction variables include: vehicle speed multiplied by engine speed, vehicle speed multiplied by gear, engine speed multiplied by gear, vehicle speed multiplied by engine speed multiplied by gear, engine speed squared, oil pressure multiplied by oil temperature, oil pressure multiplied by water temperature, water temperature multiplied by oil temperature, water temperature after cooling, and oil temperature minus water temperature.

[0119] In one alternative, the fourth acquisition unit further includes a third acquisition module and a second classification module;

[0120] The third acquisition module is used to acquire engine sample data associated with the original NOx prediction scenario; the second classification module is used to classify the engine sample data into independent variables, feature creation variables, and feature interaction variables; among them, independent variables include: intercooler temperature, intercooler pressure, transient exhaust gas flow, throttle, vehicle speed, fuel injection quantity, water temperature, oil pressure, and oil temperature; feature creation variables include: power, differential of intercooler pressure, differential of intercooler temperature, differential of intake air volume, steady-state NOx original emission, fresh intake air volume, air-fuel ratio, and fuel consumption; feature interaction variables include: intercooler pressure multiplied by intercooler temperature, intercooler pressure squared, intercooler temperature squared, intercooler temperature multiplied by water temperature multiplied by oil temperature, intercooler temperature multiplied by water temperature, oil pressure multiplied by oil temperature, oil pressure multiplied by water temperature, water temperature multiplied by oil temperature, and water temperature divided by intercooler temperature.

[0121] In this embodiment, the impact of independent variables, feature creation variables, and feature interaction variables on the performance of the data prediction model is comprehensively considered. The correlation between the independent variables, feature creation variables, and feature interaction variables and the target scene is calculated respectively. Multiple N-dimensional feature data combination schemes are selected and input into the data prediction model. The N-dimensional feature data consists of independent variables selected from high to low correlation in dimension A, feature creation variables selected from high to low correlation in dimension B, and feature interaction variables selected from high to low correlation in dimension C. The data prediction model calculates and outputs the mean error and loss function corresponding to the multiple N-dimensional feature data combination schemes. The optimal N-dimensional feature data combination scheme is selected based on the magnitude of the mean error and loss function. By inputting the optimal N-dimensional feature data combination scheme into the ECU for calculation, engine control can be achieved. This reduces the dimensionality of the dataset, captures key data information, improves model accuracy, and reduces ECU memory usage.

[0122] The engine control unit includes a processor and a memory. The aforementioned building units, arithmetic units, and transmitting units are all stored as program units in the memory. The processor executes these program units stored in the memory to achieve the corresponding functions. All of the above modules reside in the same processor; alternatively, the modules may be located in different processors in any combination.

[0123] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured; adjusting kernel parameters can reduce dataset dimensionality, improve model accuracy, and reduce ECU memory usage.

[0124] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0125] This invention provides a computer-readable storage medium including a stored program, wherein the program, when executed, controls the device containing the computer-readable storage medium to perform a control method for an engine.

[0126] Specifically, the engine control methods include:

[0127] Step S201: Obtain engine sample data associated with the target scene, and classify the engine sample data into independent variables and non-independent variables. The independent variables are variables directly collected by sensors, and the non-independent variables include feature creation variables and feature interaction variables. The feature creation variables are new feature variables obtained by transforming and / or combining at least one of the independent variables, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables.

[0128] Step S202: Obtain the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene;

[0129] Step S203: Construct multiple N-dimensional sample feature data. Each N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. The A-dimensional independent variables are composed of the top A independent variables with the highest to lowest correlation among the multiple independent variables. The B-dimensional feature creation variables are composed of the top B feature creation variables with the highest to lowest correlation among the multiple feature creation variables. The C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest to lowest correlation among the multiple feature interaction variables. N = A + B + C, and N has at least one value.

[0130] Step S204: Input the multiple N-dimensional sample feature data into the data prediction model for calculation to obtain the corresponding mean error and loss function, and determine the optimal N-dimensional sample feature data based at least on the mean error and the loss function;

[0131] Step S205: The optimal N-dimensional true feature data is sent to the ECU so that the ECU can determine the target true output value of the target scenario based on the optimal N-dimensional true feature data, and use the target true output value to control the engine operation; wherein, A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C of the optimal N-dimensional sample feature data.

[0132] This invention provides a processor for running a program, wherein the program executes a control method for the engine during runtime.

[0133] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements at least the steps of an engine control method. The device described herein may be a server, PC, PAD, mobile phone, etc.

[0134] This application also provides a computer program product that, when executed on a data processing device, is adapted to perform a program that initializes a control method step having at least an engine.

[0135] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0136] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied 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.

[0137] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0139] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0140] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0141] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0142] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0143] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0144] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0145] 1) A method for controlling an engine according to this application, wherein an ECU is installed in the engine, comprising: acquiring engine sample data associated with a target scene, and classifying the engine sample data into independent variables and non-independent variables, wherein the independent variables are variables directly collected by sensors, and the non-independent variables include feature creation variables and feature interaction variables, wherein the feature creation variables are new feature variables obtained by transforming and / or combining at least one independent variable, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables; acquiring the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene; constructing multiple N-dimensional sample feature data, wherein an N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables, and at least one of A, B, and C corresponding to any two N-dimensional sample feature data is different, wherein the A-dimensional independent variables are multiple The N-dimensional feature data consists of the top A independent variables with the highest correlation from the lowest, the B-dimensional feature creation variables consist of the top B feature creation variables with the highest correlation from the lowest, and the C-dimensional feature interaction variables consist of the top C feature interaction variables with the highest correlation from the lowest. Here, N = A + B + C, and N has at least one value. Multiple N-dimensional sample feature data are input into the data prediction model for computation, yielding the corresponding mean error and loss function. The optimal N-dimensional sample feature data is determined based at least on the mean error and loss function. The optimal N-dimensional true feature data is then sent to the ECU, enabling the ECU to determine the target true output value for the target scenario and control engine operation using this target true output value. The A, B, and C values ​​of the optimal N-dimensional true feature data are identical to those of the optimal N-dimensional sample feature data. This approach reduces the dimensionality of the dataset while capturing key data information, improving model accuracy, and reducing ECU memory usage.

[0146] 2) An engine control device according to this application includes a first acquisition unit for acquiring engine sample data associated with a target scene, and classifying the engine sample data into independent variables and non-independent variables. Non-independent variables include feature creation variables and feature interaction variables. Independent variables are variables directly acquired by sensors, feature creation variables are new variables obtained by transforming and / or combining at least one independent variable, and feature interaction variables are new feature variables generated by interacting with multiple identical or different independent variables. A second acquisition unit is used to acquire the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene. A construction unit is used to construct multiple N-dimensional sample feature data, where each N-dimensional sample feature data is constructed using A-dimensional independent variables and B-dimensional feature creation variables. The N-dimensional sample feature data is constructed from variables and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. Here, A-dimensional independent variables are composed of the top A independent variables with the highest to lowest correlation among multiple independent variables; B-dimensional feature creation variables are composed of the top B feature creation variables with the highest to lowest correlation among multiple feature creation variables; and C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest to lowest correlation among multiple feature interaction variables. N = A + B + C, and N has at least one value. The computation unit is used to input multiple N-dimensional sample feature data into the data prediction model for computation, obtaining the corresponding mean error and loss function. The optimal N-dimensional sample feature data is determined at least based on the mean error and loss function. This achieves the effect of reducing the dimensionality of the dataset, capturing key data information, improving model accuracy, and reducing ECU memory usage.

[0147] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for controlling an engine, characterized in that, The engine is equipped with an ECU, including: The engine sample data associated with the target scenario is acquired, and the engine sample data is classified into independent variables and non-independent variables. The independent variables are variables directly collected by sensors, and the non-independent variables include feature creation variables and feature interaction variables. The feature creation variables are new feature variables obtained by transforming and / or combining at least one of the independent variables, and the feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables. Obtain the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene; Multiple N-dimensional sample feature data are constructed. Each N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. The A-dimensional independent variables are composed of the top A independent variables with the highest to lowest correlation among the multiple independent variables. The B-dimensional feature creation variables are composed of the top B feature creation variables with the highest to lowest correlation among the multiple feature creation variables. The C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest to lowest correlation among the multiple feature interaction variables. N = A + B + C, and N has at least one value. Multiple N-dimensional sample feature data are input into a data prediction model for calculation to obtain the corresponding mean error and loss function. The optimal N-dimensional sample feature data are determined based at least on the mean error and the loss function. The optimal N-dimensional true feature data is sent to the ECU so that the ECU can determine the target true output value of the target scenario based on the optimal N-dimensional true feature data, and use the target true output value to control the engine operation; wherein, A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C of the optimal N-dimensional sample feature data; Determining optimal N-dimensional sample feature data based at least on the mean error and the loss function includes: determining the final integration location of the prediction model, wherein the final integration location is either in the cloud or at the ECU; and determining optimal N-dimensional sample feature data based on the final integration location of the prediction model, the mean error, and the loss function. Determining optimal N-dimensional sample feature data based on the final ensemble position of the prediction model, the mean error, and the loss function includes: determining a first weight of the final ensemble position of the prediction model; determining a second weight of the mean error and the loss function, wherein the first weight and the second weight represent the degree of influence on the optimal selection result of the N-dimensional sample feature data; and determining the optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, the mean error, and the loss function. Obtaining the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene includes: using the formula Determine the degree of correlation. For variables The sample mean, For variables The sample mean, This refers to the independent variable, the feature creation variable, or the feature interaction variable. This represents the target sample output value of the target scene.

2. The method according to claim 1, characterized in that, Determining the optimal N-dimensional sample feature data based at least on the mean error and the loss function includes: The types of data prediction models are determined, wherein the types of data prediction models are at least one of the following: GBDT model, XGBoost model, BP neural network model, TDN neural network model, LSTM neural network model, GRU neural network model, and SVM model; The optimal N-dimensional sample feature data is determined based on the type of data prediction model, the mean error, and the loss function.

3. The method according to claim 1 or 2, characterized in that, After acquiring engine sample data associated with the target scenario, and before classifying the engine sample data into independent and dependent variables, the method further includes: The engine sample data associated with the target scenario is cleaned to obtain cleaned engine sample data. The data cleaning includes deleting missing data, deleting abnormal data, and deleting inconsistent data.

4. The method according to claim 1 or 2, characterized in that, The target scenario is a vehicle load estimation scenario. Engine sample data associated with the target scenario is obtained and classified into independent variables and dependent variables. The dependent variables include feature creation variables and feature interaction variables, including: Obtain engine sample data associated with the vehicle load estimation scenario; The engine sample data is categorized into independent variables, feature creation variables, and feature interaction variables. The independent variables include: throttle, vehicle speed, engine speed, fuel injection quantity, torque, transient intake air volume, coolant temperature, oil pressure, oil temperature, and gear position. The feature creation variables include: instantaneous fuel consumption, gear ratio, acceleration, wheel-side force, calculated vehicle load, and air resistance. The feature interaction variables include: vehicle speed multiplied by engine speed, vehicle speed multiplied by gear position, engine speed multiplied by gear position, vehicle speed multiplied by engine speed multiplied by gear position, engine speed squared, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, coolant temperature after cooling, and oil temperature minus coolant temperature.

5. The method according to claim 1 or 2, characterized in that, The target scenario is the original NOx emission prediction scenario. Engine sample data associated with the target scenario is obtained, and the engine sample data is classified into independent variables and dependent variables. The dependent variables include feature creation variables and feature interaction variables, including: Obtain engine sample data associated with the original NOx emission prediction scenario; The engine sample data is categorized into independent variables, feature creation variables, and feature interaction variables. The independent variables include: intercooler temperature, intercooler pressure, transient exhaust gas flow, throttle, vehicle speed, fuel injection quantity, coolant temperature, oil pressure, and oil temperature. The feature creation variables include: power, differential of intercooler pressure, differential of intercooler temperature, differential of intake air volume, steady-state NOx emission, fresh intake air volume, air-fuel ratio, and fuel consumption. The feature interaction variables include: intercooler pressure multiplied by intercooler temperature, intercooler pressure squared, intercooler temperature squared, intercooler temperature multiplied by coolant temperature multiplied by oil temperature, intercooler temperature multiplied by coolant temperature, oil pressure multiplied by oil temperature, oil pressure multiplied by coolant temperature, coolant temperature multiplied by oil temperature, and coolant temperature excluding intercooler temperature.

6. A control device for an engine, characterized in that, The engine is equipped with an ECU, including: The first acquisition unit is used to acquire engine sample data associated with the target scene, and classify the engine sample data into independent variables and non-independent variables. The non-independent variables include feature creation variables and feature interaction variables. The independent variables are variables directly collected by sensors. The feature creation variables are new variables obtained by transforming and / or combining at least one of the independent variables. The feature interaction variables are new feature variables generated by interacting multiple identical or different independent variables. The second acquisition unit is used to acquire the correlation degree between any one of the multiple independent variables, any one of the multiple feature creation variables, and any one of the multiple feature interaction variables and the target sample output value of the target scene; A construction unit is used to construct multiple N-dimensional sample feature data. Each N-dimensional sample feature data is constructed using A-dimensional independent variables, B-dimensional feature creation variables, and C-dimensional feature interaction variables. At least one of A, B, and C is different for any two N-dimensional sample feature data. The A-dimensional independent variables are composed of the top A independent variables with the highest to lowest correlation among the multiple independent variables. The B-dimensional feature creation variables are composed of the top B feature creation variables with the highest to lowest correlation among the multiple feature creation variables. The C-dimensional feature interaction variables are composed of the top C feature interaction variables with the highest to lowest correlation among the multiple feature interaction variables. N = A + B + C, and N has at least one value. The computation unit is used to input multiple N-dimensional sample feature data into the data prediction model for computation, to obtain the corresponding mean error and loss function, and to determine the optimal N-dimensional sample feature data based at least on the mean error and the loss function. A sending unit is used to send the optimal N-dimensional true feature data to the ECU, so that the ECU can determine the target true output value of the target scenario based on the optimal N-dimensional true feature data, and use the target true output value to control the engine operation; wherein, A, B, and C of the optimal N-dimensional true feature data are the same as A, B, and C in the optimal N-dimensional sample feature data; The computing unit also includes a second determining module and a third determining module; the second determining module is used to determine the final integration location of the prediction model, which is either the cloud or the ECU; the third determining module is used to determine the optimal N-dimensional sample feature data based on the final integration location of the prediction model, the mean error, and the loss function. The third determining module further includes a first determining submodule, a second determining submodule, and a third determining submodule; the first determining submodule is used to determine the first weight of the final ensemble position of the prediction model; the second determining submodule is used to determine the second weight of the mean error and the loss function, where the first weight and the second weight represent the degree of influence on the optimal selection result of the N-dimensional sample feature data; the third determining submodule is used to determine the optimal N-dimensional sample feature data based on the first weight, the second weight, the final ensemble position of the prediction model, and the mean error and the loss function. The second acquisition unit includes a fourth determination module; the fourth determination module is used to employ a formula Determine the degree of correlation. For variables The sample mean, For variables The sample mean, This refers to independent variables, feature creation variables, or feature interaction variables. This represents the target sample output value for the target scene.

7. An electronic device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including a control method for performing the engine according to any one of claims 1 to 5.

Images