A data-driven method and system for correcting transient fuel consumption in engines

Abstract

This invention relates to the field of automotive energy management and data-driven technology, providing a data-driven method and system for correcting engine transient fuel consumption. The method includes the following steps: real-time acquisition of multi-source data during vehicle operation; preprocessing of the multi-source data to output effective operating condition data; construction of three transient fuel consumption correction sub-models based on various artificial neural networks and the effective operating condition data; linear weighted fusion of the three transient fuel consumption correction sub-models to obtain a total transient fuel consumption correction model; and correction of engine transient fuel consumption based on the total transient fuel consumption correction model. This invention, by real-time acquisition of multi-source data during vehicle operation and establishing a total transient fuel consumption correction model through data-driven and intelligent modeling methods, achieves accurate prediction and real-time correction of transient fuel consumption, thereby improving the fuel economy and energy management optimization effect of hybrid vehicles under actual operating conditions.

Description

Technical Field

[0001] This invention belongs to the field of automotive energy management and data-driven technology, and particularly relates to a data-driven method and system for correcting transient fuel consumption of an engine. Background Technology

[0002] With increasingly stringent requirements for energy conservation and emission reduction, and the rapid development of new energy technologies, hybrid electric vehicles (HEVs) are gradually becoming the mainstream in the market due to their balance between fuel economy and power performance. Currently, the energy management strategy (EMS) of HEVs typically relies on the engine's steady-state fuel consumption characteristic map (MAP) and a rule-based energy management strategy during the design and optimization process. This type of fuel consumption MAP is usually obtained through engine bench testing and can reflect the engine's fuel consumption level at different speeds and torques. During energy management, the controller adjusts the power distribution between the engine and the electric motor by calling upon the steady-state fuel consumption MAP, thereby improving fuel economy.

[0003] However, engines do not operate in a steady-state condition continuously during actual road use; instead, they frequently experience transient conditions such as rapid acceleration, deceleration, and start-stop. During these processes, engine combustion boundary conditions (such as intake pressure, intake temperature, coolant temperature, and throttle opening) constantly change, causing significant fluctuations in engine torque and its rate of change. Furthermore, there is a dynamic response delay in engine torque output, resulting in a significant deviation between actual fuel consumption and the steady-state fuel consumption map (MAP). Traditional energy management strategies based on the steady-state fuel consumption map are logically simple and easy to implement, but because they only describe the fuel consumption rate at a specific steady-state operating point, they cannot accurately reflect fuel consumption under transient conditions. This leads to inconsistencies between predicted and actual fuel consumption in actual operation, thus reducing the overall vehicle's fuel economy. On the other hand, hybrid vehicles also incorporate an electric motor and battery system. The electric motor provides auxiliary power during acceleration and recovers energy during deceleration. The battery's state of charge (SOC), the motor's torque output characteristics, and the vehicle's energy management commands all significantly impact actual fuel consumption. However, most existing studies only correct the engine's steady-state fuel consumption MAP, failing to fully consider the effects of electric motor assistance and battery status, and lacking a transient fuel consumption correction mechanism for the entire vehicle system.

[0004] In summary, energy management strategies relying on traditional steady-state fuel consumption MAPs suffer from fuel consumption prediction biases under real-world transient conditions, leading to decreased fuel economy and failing to fully realize the energy-saving potential of hybrid vehicles. Therefore, there is an urgent need for a transient fuel consumption correction method that integrates multiple transient factors such as engine combustion boundary conditions, torque and torque change rate, and electric motor assistance. This method would dynamically correct the steady-state fuel consumption MAP, improving the accuracy and stability of hybrid energy management strategies during actual operation and reducing the overall vehicle fuel consumption rate. Summary of the Invention

[0005] The purpose of this invention is to provide a data-driven method for correcting transient fuel consumption in engines, thereby addressing the aforementioned technical problems.

[0006] This invention is implemented as follows: a data-driven method for correcting transient engine fuel consumption includes the following steps:

[0007] Real-time acquisition of multi-source data during vehicle operation; the multi-source data includes engine operation data, motor operation data, battery operation data, and vehicle operation data;

[0008] The multi-source data is preprocessed to output valid operating condition data;

[0009] Based on various artificial neural networks, three transient fuel consumption correction sub-models are constructed according to the effective operating condition data. The transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction model, and an electric motor auxiliary correction model.

[0010] The three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model; the engine transient fuel consumption is corrected according to the total transient fuel consumption correction model; wherein, the weight factors of the linear weighted fusion are dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

[0011] Preferably, the preprocessing method includes data synchronization, data cleaning, and data extraction; the data synchronization is used to interpolate and align all data according to a unified time base; the data cleaning is used to remove abnormal data, missing data, and static constant data segments; and the data extraction is used to extract effective operating points based on fuel efficiency.

[0012] Preferably, the effective operating condition data includes:

[0013] Engine combustion boundary condition parameters; the engine combustion boundary condition parameters include throttle opening, intake pressure, intake temperature, and coolant temperature;

[0014] Actual engine output torque and torque change rate;

[0015] Engine start-stop status, rapid acceleration and rapid deceleration parameters;

[0016] Short-term vehicle speed prediction results;

[0017] Motor-related operating parameters; the motor-related operating parameters include the motor's transient torque, output power and its rate of change, battery SOC, current, voltage, and motor power allocation commands issued by the energy management system.

[0018] Preferably, the method for constructing the engine combustion boundary condition correction sub-model includes the following steps:

[0019] Using engine combustion boundary condition parameters as input, a sub-model for correcting engine combustion boundary conditions is established using a multilayer perceptron (MLP) and / or a backpropagation (BP) neural network, and the first prediction correction factor is output.

[0020] Preferably, the method for constructing the torque and torque change rate correction sub-model includes the following steps:

[0021] Using the engine's actual output torque, torque change rate, engine start-stop status, rapid acceleration and deceleration parameters, and short-term vehicle speed prediction as inputs, a long short-term memory network (LSTM) is used to establish a torque and torque change rate correction sub-model, and outputs a second prediction correction factor. The short-term vehicle speed prediction result is obtained by using a time-series prediction model to predict the future short-term vehicle speed change trend based on historical vehicle speed, throttle opening, and acceleration data.

[0022] Preferably, the method for constructing the motor-assisted correction sub-model includes the following steps:

[0023] Using relevant motor operating parameters as input, a motor-aided correction sub-model is established using a Long Short-Term Memory (LSTM) network, and a third prediction correction factor is output.

[0024] Preferably, the step of linearly weighting and fusing the three transient fuel consumption correction sub-models to obtain the total transient fuel consumption correction model specifically includes:

[0025] The outputs of the three transient fuel consumption correction sub-models are used as inputs, and the actual fuel consumption is used as the output. The parameters are fitted based on the multiple linear regression method to determine the initial weight factors.

[0026] Based on the genetic algorithm, the initial weight factors are optimized to obtain the optimized weight factors;

[0027] In actual vehicle operation, based on the adaptive weighting mechanism, the optimized weighting factors are dynamically adjusted according to the real-time operating status of the vehicle to obtain the final weighting factors.

[0028] Based on the final weighting factor, the three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model.

[0029] Another object of the present invention is to provide a data-driven engine transient fuel consumption correction system for implementing the above-mentioned engine transient fuel consumption correction method, comprising:

[0030] The multi-source data acquisition module is used to collect multi-source data in real time during vehicle operation; the multi-source data includes engine operation data, motor operation data, battery operation data, and vehicle operation data.

[0031] The data preprocessing module is used to preprocess the multi-source data and output valid operating condition data;

[0032] The correction sub-model construction module is used to construct three transient fuel consumption correction sub-models based on various artificial neural networks and the effective operating condition data. The transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction sub-model, and an electric motor-assisted correction model.

[0033] The correction sub-model fusion module is used to linearly weight and fuse three transient fuel consumption correction sub-models to obtain a total transient fuel consumption correction model; the engine transient fuel consumption is corrected according to the total transient fuel consumption correction model; wherein, the weight factors of the linear weighted fusion are dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

[0034] The data-driven engine transient fuel consumption correction method provided by this invention collects multi-source data during vehicle operation in real time, and establishes a total transient fuel consumption correction model through data-driven and intelligent modeling methods. This enables accurate prediction and real-time correction of transient fuel consumption, thereby improving the fuel economy and energy management optimization effect of hybrid vehicles under actual operating conditions. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the data-driven engine transient fuel consumption correction method provided in an embodiment of the present invention.

[0036] Figure 2 A schematic diagram of the structure of a data-driven engine transient fuel consumption correction system provided in an embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0038] Specifically, such as Figure 1 As shown, in one embodiment of the present invention, a data-driven engine transient fuel consumption correction method is provided, comprising the following steps:

[0039] S100. Real-time acquisition of multi-source data during vehicle operation; the multi-source data includes engine operation data, motor operation data, battery operation data, and vehicle operation data;

[0040] S200: Preprocess the multi-source data and output valid operating condition data;

[0041] S300. Based on multiple artificial neural networks, three transient fuel consumption correction sub-models are constructed according to the effective operating condition data. The transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction sub-model, and an electric motor auxiliary correction model.

[0042] S400. The three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model; the engine transient fuel consumption is corrected according to the total transient fuel consumption correction model; wherein, the weight factor of the linear weighted fusion is dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

[0043] In practical applications, a large-scale collection of multi-source data on the transient operation of vehicles is achieved through extensive on-road operation under various working conditions and driving habits. This multi-source data primarily includes: engine speed, torque, torque change rate, throttle opening, intake pressure, intake air temperature, coolant temperature, and engine start-stop status; motor torque and power and their change rate; battery SOC, current, and voltage; and power distribution commands from the vehicle energy management system. The above data undergoes systematic data preprocessing to extract typical operating condition segments, and the actual fuel consumption of the engine is obtained using a fuel flow meter.

[0044] In a preferred embodiment of the present invention, large-scale vehicle transient operating data needs to be preprocessed before being input into the artificial neural network. The preprocessing method includes:

[0045] (1) Data synchronization: All data are interpolated and aligned according to a unified time base.

[0046] (2) Data cleaning: Based on the throttle opening at each operating point, determine the vehicle's real-time power demand and sequentially remove operating conditions that do not require energy input. On this basis, remove abnormal sensor data, interpolation data shortage data, and static constant data segments.

[0047] (3) Data extraction: Extract the effective data operating points after cleaning. The extraction principle is the effective utilization rate of fuel. Among them, the effective utilization rate of fuel The calculation formula is as follows:

[0048] ;

[0049] In the formula, W s Q1 represents the effective work of the engine; Q2 represents the total energy consumed by the fuel.

[0050] Specifically, the effective operating condition data includes:

[0051] Engine combustion boundary condition parameters; the engine combustion boundary condition parameters include throttle opening, intake pressure, intake temperature, and coolant temperature;

[0052] Actual engine output torque and torque change rate;

[0053] Engine start-stop status, rapid acceleration and rapid deceleration parameters;

[0054] Short-term vehicle speed prediction results;

[0055] Motor-related operating parameters; the motor-related operating parameters include the motor's transient torque, output power and its rate of change, battery SOC, current, voltage, and motor power allocation commands issued by the energy management system.

[0056] In a preferred embodiment of the present invention, the method for constructing the engine combustion boundary condition correction sub-model includes the following steps:

[0057] Using engine combustion boundary condition parameters as input, a sub-model for correcting engine combustion boundary conditions is established using a multilayer perceptron (MLP) and / or a backpropagation (BP) neural network, and the first prediction correction factor is output.

[0058] In practical applications, engine combustion boundary condition parameters such as throttle opening, intake pressure, intake air temperature, and coolant temperature are used as inputs. These parameters are mostly static operating condition parameters, and while there is a strong nonlinear relationship between the input parameters and fuel consumption, the temporal relationship is not strong. Therefore, a multilayer perceptron (MLP) and / or backpropagation (BP) neural network can be used to establish a modified sub-model of the engine combustion boundary conditions on the steady-state fuel consumption (MAP), resulting in high computational efficiency. The specific modeling process is as follows:

[0059] (1) During the engine bench test and the vehicle road operation, boundary condition parameters related to engine combustion are collected, including throttle opening, intake pressure, intake temperature, coolant temperature, etc., and the engine speed, torque and actual fuel consumption obtained by fuel flow meter are recorded.

[0060] (2) The collected data is time-synchronized, preprocessed and cleaned, and the transient change characteristics of fuel consumption rate under different engine combustion boundary conditions are extracted and compared with the fuel consumption characteristic map (steady-state fuel consumption MAP) under steady-state conditions to obtain the deviation between the actual fuel consumption under transient conditions and the steady-state predicted value.

[0061] (3) The sensitivity analysis method was used to quantify the influence weight of different engine combustion boundary conditions on transient fuel consumption and screen out the main weight boundary condition parameters of the sub-model.

[0062] (4) Using the selected main weight boundary condition parameters as input, a multilayer perceptron MLP is used to establish a modified sub-model of engine combustion boundary condition with a strong nonlinear relationship between engine combustion boundary condition parameters and fuel consumption, so as to realize the mapping from steady-state fuel consumption MAP to transient fuel consumption.

[0063] (5) The engine combustion boundary condition correction sub-model obtained by training is embedded into the energy management system, and the first prediction correction factor is output in real time during vehicle operation, so as to dynamically correct the fuel consumption prediction under different operating conditions and improve the accuracy of fuel consumption prediction.

[0064] In a preferred embodiment of the present invention, the torque and torque change rate correction sub-model is based on the transient operating conditions of the engine and mainly considers the impact of torque and its change rate on fuel consumption. Its construction method includes the following steps:

[0065] Using the engine's actual output torque, torque change rate, engine start-stop status, rapid acceleration and deceleration parameters, and short-term vehicle speed prediction as inputs, a long short-term memory network (LSTM) is used to establish a torque and torque change rate correction sub-model, and outputs a second prediction correction factor. The short-term vehicle speed prediction result is obtained by using a time-series prediction model to predict the future short-term vehicle speed change trend based on historical vehicle speed, throttle opening, and acceleration data.

[0066] In practical applications, the actual output torque and torque change rate of the engine, the engine's start-stop state, rapid acceleration and deceleration parameters, and short-term vehicle speed prediction are used as inputs. This approach exhibits both steady-state characteristics (torque magnitude determines basic fuel consumption) and dynamic characteristics (the rate of change determines transient additional consumption). The fluctuation of transient torque and vehicle speed changes need to consider their time dependence. The engine's start-stop state and rapid acceleration / deceleration are essentially time-series dynamics problems. Since LSTM can capture the dynamic effects of acceleration and deceleration, this embodiment of the invention uses LSTM to establish a correction sub-model for transient load fluctuations on fuel consumption, reflecting the impact of engine dynamic response on fuel consumption and accurately reflecting the correlation between "historical torque changes and current fuel consumption." Simultaneously, a short-term vehicle speed prediction algorithm is introduced, which predicts the vehicle speed change trend within a short period (0.5-2 seconds) to capture driving intentions in advance, achieving feedforward correction of transient fuel consumption. The specific modeling process is as follows:

[0067] (1) Collect dynamic parameters such as engine speed, torque, torque change rate, vehicle longitudinal acceleration, and rapid acceleration / deceleration signs. In the data preprocessing, the input at continuous time moments is constructed into time-series features through a sliding time window to reflect the dynamic response process of the engine.

[0068] (2) Based on the vehicle’s historical speed, throttle opening, acceleration and other data, a time-series prediction model (preferably a long short-term memory network LSTM or a gated recurrent unit GRU) is used to predict the vehicle speed change trend in the future short time (e.g., 0.5-2s). The prediction output includes the vehicle speed and rate of change (Δv / Δt) at future time, which is the short-term vehicle speed prediction result.

[0069] (3) Compare the transient fuel consumption corresponding to the above data with the steady-state fuel consumption under the corresponding steady-state conditions to obtain the prediction deviation of the steady-state fuel consumption MAP under transient conditions. Quantify the deviation in the form of a ratio and use the deviation as the transient target correction factor.

[0070] (4) The steady-state fuel consumption MAP is used as the baseline input, and together with the above-mentioned time-series features and the predicted short-term vehicle speed, it is input into the time-series neural network model (such as a long short-term memory network LSTM) to establish a nonlinear mapping relationship between the input and the transient target correction factor. In the process of calculating the predicted transient correction factor, higher weights are given to the data samples of key transient conditions such as rapid acceleration, rapid deceleration, and engine start-stop, so as to improve the prediction accuracy of the model under extreme conditions.

[0071] (5) The model outputs a second prediction correction factor, which is superimposed on the baseline MAP to obtain the torque and torque change rate correction sub-model.

[0072] In a preferred embodiment of the present invention, the motor-assisted correction sub-model is used to characterize the impact of the motor on the transient fuel consumption of the engine in a hybrid power system, especially its corrective effect on vehicle energy distribution and fuel economy under conditions such as acceleration, deceleration, and engine start-stop; the method for constructing the motor-assisted correction sub-model includes the following steps:

[0073] Using relevant motor operating parameters as input, a motor-aided correction sub-model is established using a Long Short-Term Memory (LSTM) network, and a third prediction correction factor is output.

[0074] In practical applications, the dynamic interaction between the motor and battery, using the motor's transient torque, output power, and rate of change as inputs, is a typical time-series problem, requiring consideration of the impact of historical states on current corrections. Therefore, this embodiment of the invention also employs LSTM to establish a motor-assisted correction sub-model to characterize the compensation effect of motor-shared power and energy recovery on fuel consumption. The specific modeling process is as follows:

[0075] (1) Collect relevant operating parameters of the motor, including the actual output / absorbed torque of the motor, the motor power and its rate of change, the battery SOC (state of charge), current, voltage, and motor power allocation commands issued by the energy management system. In order to reflect the real-time auxiliary effect of the motor on the dynamic process of the engine, the time-series input characteristics of the change process of motor torque and power are constructed by using a sliding time window method. At the same time, rapid acceleration / deceleration flags and engine start-stop status flags are introduced to capture the rapid intervention or withdrawal characteristics of the motor under typical transient conditions. The collected data are then cleaned and extracted.

[0076] (2) Based on steady-state fuel consumption (MAP), a nonlinear mapping model for motor-assisted correction factors is constructed. This model preferably uses a long short-term memory (LSTM) network to learn the influence of motor output and battery status on engine transient fuel consumption. For samples where the battery SOC is in a low charge or high current state, higher weights are given during training to ensure the prediction accuracy of the model under battery-constrained operating conditions.

[0077] (3) The output of the motor-assisted correction sub-model is the third prediction correction factor for steady-state fuel consumption, which can dynamically reflect the reduction or increase effect of motor intervention and battery status on engine transient fuel consumption.

[0078] (4) To ensure the real-time performance and reliability of the model in the vehicle control unit (ECU / EMS), the motor-assisted correction sub-model is compressed and optimized before deployment. When the input features exceed the training range or the battery parameters are abnormal, the system automatically reverts to the experience-based motor-assisted correction formula to ensure the stable execution of the vehicle energy management strategy.

[0079] In a preferred embodiment of the present invention, the step of linearly weighting and fusing the three transient fuel consumption correction sub-models to obtain the total transient fuel consumption correction model specifically includes:

[0080] The outputs of the three transient fuel consumption correction sub-models are used as inputs, and the actual fuel consumption is used as the output. The parameters are fitted based on the multiple linear regression method to determine the initial weight factors.

[0081] Based on the genetic algorithm, the initial weight factors are optimized to obtain the optimized weight factors;

[0082] In actual vehicle operation, based on the adaptive weighting mechanism, the optimized weighting factors are dynamically adjusted according to the real-time operating status of the vehicle to obtain the final weighting factors.

[0083] Based on the final weighting factor, the three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model.

[0084] Specifically, the mechanism of the above linear weighted fusion is shown in the following equation:

[0085] f total (x)=w1f1(x)+w2f2(x)+w3f3(x);

[0086] In the formula, f total f(x) represents the output of the total transient fuel consumption correction model, i.e., the final corrected transient fuel consumption; f1(x), f2(x), and f3(x) represent the outputs of the three transient fuel consumption correction sub-models, respectively; x represents the input variable; w1, w2, and w3 represent the final weighting factors corresponding to the three transient fuel consumption correction sub-models, which are optimized based on vehicle operating data and optimal fuel consumption performance. The specific process is as follows:

[0087] (1) Initial determination stage: In the initial stage of establishing the total transient fuel consumption correction model, based on large-scale bench and road test data, the output results (fuel consumption prediction results) of the three transient fuel consumption correction sub-models are used as inputs and the actual fuel consumption is used as outputs. The parameters are fitted by the multiple linear regression method to obtain a set of initial weight factors.

[0088] (2) To further improve the model accuracy, intelligent optimization algorithms (such as genetic algorithms) are used to perform secondary optimization on the initial weight factors. Through global optimization, a better weight factor allocation under different working conditions is obtained, resulting in optimized weight factors;

[0089] (3) During the actual operation of the vehicle, an adaptive weighting mechanism of artificial neural network (such as MLP with adaptive dynamic adjustment) is adopted so that the optimized weight factors can be dynamically adjusted according to the real-time operating status of the vehicle (such as engine speed, load, battery SOC, driver's driving intention, etc.) to obtain the final weight factors.

[0090] It is worth noting that the aforementioned data-driven engine transient fuel consumption correction method can be applied to the energy management system (EMS) of hybrid electric vehicles. Specifically, the total transient fuel consumption correction model is embedded into the vehicle's EMS. During vehicle operation, the EMS calls the total transient fuel consumption correction model in real time to predict and compare transient fuel consumption for different engine-motor power distribution schemes, and selects the optimal scheme to execute. Especially under typical transient conditions such as engine start-stop, rapid acceleration, deceleration, and regenerative braking, this model can accurately capture the engine's dynamic response delay and the motor's auxiliary characteristics, thereby optimizing the vehicle's energy management decisions.

[0091] Furthermore, to ensure the accuracy and adaptability of the total transient fuel consumption correction model, this embodiment of the invention also designs an online learning mechanism: as the transient operating data of the vehicle under various complex operating conditions accumulates, the predicted value of the total transient fuel consumption correction model is compared with the actual value to calculate the prediction error. The error data is used as input to update and iterate the weight factor parameters of different transient fuel consumption correction sub-models using the backpropagation algorithm, thereby ensuring the long-term applicability and sustainability of the total transient fuel consumption correction model.

[0092] In summary, the data-driven engine transient fuel consumption correction method provided in this embodiment of the invention has the following beneficial technical effects compared to the prior art:

[0093] 1. A multi-dimensional, data-driven method for constructing a total transient fuel consumption correction model is proposed. By introducing the total transient fuel consumption correction model, fuel consumption can be predicted more accurately under complex operating conditions such as engine start-stop, rapid acceleration, and rapid deceleration, and the power distribution between the engine and the electric motor can be dynamically adjusted, thereby significantly improving fuel economy.

[0094] 2. Various artificial neural networks (such as MLP, LSTM, etc.) are used to establish transient fuel consumption correction sub-models of different dimensions, taking into account both nonlinear relationships and temporal characteristics, and can adapt to the performance requirements and prediction accuracy under various complex working conditions.

[0095] 3. A short-time vehicle speed prediction algorithm is introduced into the torque and torque change rate correction sub-model. This embodiment of the invention utilizes historical vehicle operating data (vehicle speed, throttle opening, acceleration, etc.) and employs a time-series prediction model (LSTM / GRU) to predict the vehicle speed change trend within a short timeframe (0.5-2 seconds). The prediction results, along with parameters such as engine torque and torque change rate, are input into the transient correction network. This method, by anticipating changes in driver power demand, achieves feedforward correction of fuel consumption, effectively reducing prediction bias caused by engine response delay and improving the temporal foresight and dynamic accuracy of the transient fuel consumption model.

[0096] 4. It has an online learning function. As vehicle driving data is accumulated and learned, the total transient fuel consumption correction model can be continuously optimized online according to real-time operating parameters to adapt to long-term use scenarios such as vehicle aging and changes in driving habits.

[0097] like Figure 2 As shown, in another embodiment of the present invention, a data-driven engine transient fuel consumption correction system is also provided to implement the above-described engine transient fuel consumption correction method, comprising:

[0098] The multi-source data acquisition module 10 is used to acquire multi-source data in real time during vehicle operation; the multi-source data includes engine operation data, motor operation data, battery operation data and vehicle operation data;

[0099] Data preprocessing module 20 is used to preprocess the multi-source data and output valid operating condition data;

[0100] The correction sub-model construction module 30 is used to construct three transient fuel consumption correction sub-models based on multiple artificial neural networks and the effective operating condition data; the transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction sub-model, and an electric motor auxiliary correction model;

[0101] The correction sub-model fusion module 40 is used to perform linear weighted fusion of three transient fuel consumption correction sub-models to obtain a total transient fuel consumption correction model; and to correct the engine transient fuel consumption according to the total transient fuel consumption correction model; wherein, the weight factors of the linear weighted fusion are dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

[0102] It should be noted that each of the above modules can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up each module, enabling the processor to execute each step of the above method.

[0103] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0104] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0105] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A data-driven method for correcting transient engine fuel consumption, characterized in that, Includes the following steps: Real-time acquisition of multi-source data during vehicle operation; the multi-source data includes engine operation data, motor operation data, battery operation data, and vehicle operation data; The multi-source data is preprocessed to output valid operating condition data; The valid operating condition data includes: Engine combustion boundary condition parameters; the engine combustion boundary condition parameters include throttle opening, intake pressure, intake temperature, and coolant temperature; Actual engine output torque and torque change rate; Engine start-stop status, rapid acceleration and rapid deceleration parameters; Short-term vehicle speed prediction results; Motor operating data; the motor operating data includes the motor's transient torque, output power and its rate of change, battery SOC, current, voltage, and motor power allocation instructions issued by the energy management system; Based on various artificial neural networks, three transient fuel consumption correction sub-models are constructed according to the effective operating condition data. The transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction model, and an electric motor auxiliary correction model. The method for constructing the modified sub-model of engine combustion boundary conditions includes the following steps: Using engine combustion boundary condition parameters as input, a sub-model for engine combustion boundary condition correction is established using a multilayer perceptron (MLP) and / or a backpropagation neural network, and the first prediction correction factor is output. The method for constructing the torque and torque change rate correction sub-model includes the following steps: Using the engine's actual output torque, torque change rate, engine start-stop status, rapid acceleration and deceleration parameters, and short-term vehicle speed prediction as inputs, a long short-term memory network (LSTM) is used to establish a torque and torque change rate correction sub-model, and outputs a second prediction correction factor; wherein, the short-term vehicle speed prediction result is obtained by predicting the future short-term vehicle speed change trend based on historical vehicle speed, throttle opening, and acceleration data through a time-series prediction model. The method for constructing the motor-assisted correction sub-model includes the following steps: Using motor operating data as input, a motor-assisted correction sub-model is established using a Long Short-Time Memory (LSTM) network, and a third prediction correction factor is output. The three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model; the engine transient fuel consumption is corrected according to the total transient fuel consumption correction model; wherein, the weight factors of the linear weighted fusion are dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

2. The data-driven engine transient fuel consumption correction method according to claim 1, characterized in that, The preprocessing method includes data synchronization, data cleaning, and data extraction; the data synchronization is used to interpolate and align all data according to a unified time base; the data cleaning is used to remove abnormal data, missing data, and static constant data segments; the data extraction is used to extract effective operating points based on fuel efficiency.

3. The data-driven engine transient fuel consumption correction method according to claim 1, characterized in that, The specific steps for linearly weighting and fusing the three transient fuel consumption correction sub-models to obtain the total transient fuel consumption correction model include: The outputs of the three transient fuel consumption correction sub-models are used as inputs, and the actual fuel consumption is used as the output. The parameters are fitted based on the multiple linear regression method to determine the initial weight factors. Based on the genetic algorithm, the initial weight factors are optimized to obtain the optimized weight factors; In actual vehicle operation, based on the adaptive weighting mechanism, the optimized weighting factors are dynamically adjusted according to the real-time operating status of the vehicle to obtain the final weighting factors. Based on the final weighting factor, the three transient fuel consumption correction sub-models are linearly weighted and fused to obtain the total transient fuel consumption correction model.

4. A data-driven engine transient fuel consumption correction system, used to implement the engine transient fuel consumption correction method according to any one of claims 1-3, characterized in that, include: The multi-source data acquisition module is used to collect multi-source data in real time during vehicle operation; The multi-source data includes engine operating data, motor operating data, battery operating data, and vehicle operating data; The data preprocessing module is used to preprocess the multi-source data and output valid operating condition data; The correction sub-model construction module is used to construct three transient fuel consumption correction sub-models based on various artificial neural networks and the effective operating condition data. The transient fuel consumption correction sub-models include an engine combustion boundary condition correction sub-model, a torque and torque change rate correction sub-model, and an electric motor-assisted correction model. The correction sub-model fusion module is used to linearly weight and fuse three transient fuel consumption correction sub-models to obtain a total transient fuel consumption correction model; the engine transient fuel consumption is corrected according to the total transient fuel consumption correction model; wherein, the weight factors of the linear weighted fusion are dynamically adjusted based on vehicle operating data and optimal fuel consumption performance through an optimization algorithm.

Images