A diesel engine instantaneous fuel consumption calculation method and system based on a deep neural network

Abstract

This application discloses a method and system for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network, relating to the field of vehicle system development. The method includes: S1, data acquisition: a vehicle-side T-BOX collects vehicle status information and sends it to the cloud; S2, data cleaning: the received raw vehicle status information is cleaned in the cloud; S3, data slicing: based on the continuity of GPS time, the most recently acquired N cleaned data points are extracted and assembled into a data slice; S4, feature parameter calculation: for the N data points in the data slice, M feature parameters are calculated according to a set rule; S5, feature sample construction: the last N-1 data points are extracted from the data slice as samples; S6, model inference: the feature samples are input into a pre-trained deep neural network model, and the deep neural network model outputs the instantaneous fuel consumption value of the engine.

Description

Technical Field

[0001] This application relates to the field of vehicle system development, and in particular to a method for calculating instantaneous fuel consumption of diesel engines based on deep neural networks, a system for calculating instantaneous fuel consumption of diesel engines based on deep neural networks, electronic devices, storage media, and a China VI diesel engine vehicle platform. Background Technology

[0002] In the development of control functions related to vehicle fuel economy, such as AMT shifting strategies, predictive cruise control, hybrid power system control strategies, and vehicle energy management, a steady-state fuel consumption calculation model (steady-state fuel consumption map) is typically used to calculate overall vehicle fuel economy. The data used in constructing this model is obtained through engine bench tests, where the engine operates under steady-state conditions (stable operating conditions such as speed and torque). The steady-state fuel consumption calculation model only reflects the engine's fuel consumption level under steady-state conditions. However, actual engine operating conditions are typically transient, requiring priority to be given to power, response speed, emission control, and component protection. Not only do engine speed and torque change frequently, but operating conditions such as intake pressure, intake temperature, and intake flow also change continuously over time. The steady-state fuel consumption model cannot accurately reflect fuel consumption under transient conditions; its calculated results are significantly lower than actual fuel consumption. This leads to actual fuel consumption for fuel-saving functions such as predictive cruise control being higher than the model's calculation results, reducing overall vehicle fuel economy and failing to fully realize the energy-saving effects of the control strategy.

[0003] To improve the accuracy of engine fuel consumption calculation and develop control functions that can truly improve the fuel economy of the whole vehicle, automakers and research institutions have adopted a variety of methods to build engine transient fuel consumption calculation models.

[0004] CN202510978465 discloses a method for constructing a dynamic fuel consumption model for heavy-duty commercial vehicles. Based on steady-state test data from an engine bench, this method calculates weighting coefficients for normal, regenerative, and SCR thermal management states, and then calculates the dynamic fuel consumption model by linearly weighted summing of steady-state fuel consumption. However, this method relies on steady-state data weighting and fails to adequately learn transient nonlinear characteristics.

[0005] CN202511861859 discloses a data-driven method and system for correcting transient fuel consumption in engines. This method develops correction sub-models for factors such as engine combustion boundary conditions, torque and torque change rate, and electric motor assistance in hybrid vehicles. A fuel consumption correction model is obtained through linear weighted fusion, and finally, the fuel consumption correction model is used to process the steady-state fuel consumption to obtain the transient fuel consumption. Although this method involves correction of multiple factors, the model architecture is linear fusion, limiting its complexity.

[0006] CN202511623546 discloses a method and system for real-time fuel consumption estimation of non-road machinery vehicles. Targeting non-road vehicles such as excavators, bulldozers, and loaders, it constructs a hybrid architecture of multi-sensor fusion and adaptive operating conditions for real-time fuel consumption estimation. It combines multi-sensor data fusion, operating condition identification, adaptive Kalman filtering, and rule-based event processing to obtain a real-time fuel consumption estimate for the engine. However, this method relies on a physical model to construct the engine's transient fuel consumption, resulting in high computational complexity. Summary of the Invention

[0007] The purpose of this invention is to provide a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, a system for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, an electronic device, a storage medium, and a China VI diesel engine vehicle platform, thereby solving at least one of a number of technical problems.

[0008] Core technical issue: In the development of existing vehicle fuel economy control functions, the engine steady-state fuel consumption calculation model is usually used. However, it is only based on engine bench steady-state test data and cannot accurately reflect the fuel consumption of the engine under actual transient operating conditions. The calculation results are significantly lower, resulting in poor actual effect of related energy-saving functions.

[0009] This invention provides the following solution:

[0010] According to a first aspect of the present invention, a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network is provided, comprising:

[0011] S1. Data Collection: The vehicle-side T-BOX collects vehicle status information and sends it to the cloud.

[0012] Vehicle status information, including GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-required torque.

[0013] The vehicle-mounted T-BOX also collects one or more of the following information: GPS longitude, GPS latitude, GPS altitude, GPS positioning status, instrument speed, engine torque percentage, engine coolant temperature, and DPF regeneration status. After being uploaded to the cloud, this information, along with the vehicle status information, serves as the raw data for data cleaning.

[0014] S2. Data Cleaning: Cleaning the received raw vehicle status information in the cloud;

[0015] Data cleaning strategies include sorting by GPS time, removing duplicate values, and correcting outliers;

[0016] S3, Data Slicing: Based on the continuity of GPS time, extract the 11 most recently acquired cleaned data points and assemble them into a data slice;

[0017] When the total amount of data after cleaning is less than 11 records, it is padded with zeros forward to complete the data to 11 records, forming a data slice.

[0018] S4. Feature parameter calculation: 14 feature parameters are calculated for 11 data points in the data slice according to the set rules;

[0019] S5. Feature Sample Construction: Extract the last 10 data points from the data slice as samples;

[0020] The sample data is sorted by GPS time from smallest to largest, and combined with 14 feature parameters to construct a feature sample in the form of a numerical matrix with 10 rows and 14 columns;

[0021] S6. Model Inference: Input the feature samples into the pre-trained deep neural network model, and the deep neural network model outputs the instantaneous fuel consumption value of the engine.

[0022] The deep neural network model is a hybrid temporal model, including fully connected layers, batch normalization layers, one-dimensional convolutional layers, random deactivation layers, and an output layer, with ReLU as the activation function.

[0023] Furthermore, including:

[0024] In step S1, the attributes of each vehicle status information include:

[0025] GPS time is in timestamp format, in seconds, and is an integer data type.

[0026] The front wheel speed is measured in km / h and is a floating-point number.

[0027] The data types for accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque are all integers, with units of %, rpm, kPa, ℃, ℃, ℃, %.

[0028] Furthermore, including:

[0029] In step S2, the specific operations of the data cleaning strategy include:

[0030] Step S2.1: Sort the original vehicle status information in ascending order of GPS time;

[0031] Step S2.2: Delete the original vehicle status information with duplicate GPS times;

[0032] Step S2.3: Delete the original vehicle status information containing null values;

[0033] Step S2.4: Delete abnormal data according to preset filtering rules;

[0034] The preset filtering rules include:

[0035] Data showing a front wheel speed greater than 140 km / h is considered abnormal.

[0036] Accelerator pedal opening less than 0% or greater than 100% is considered abnormal data;

[0037] Data showing engine speeds below 0 rpm or above 8000 rpm is considered abnormal.

[0038] Engine intake manifold pressure less than 20 kPa or greater than 300 kPa is considered abnormal data;

[0039] Engine intake manifold temperatures below -50°C or above 150°C are considered abnormal data.

[0040] Data showing an engine DOC temperature below 100°C or above 800°C is considered abnormal.

[0041] Data with an ambient temperature below -50℃ or above 50℃ is considered abnormal.

[0042] Data showing a driver's required torque less than 0% or greater than 100% is considered abnormal.

[0043] Furthermore, including:

[0044] In step S3, the specific method for filling "zero" forward is as follows:

[0045] The zero values ​​are placed at the beginning of the data slice, and the existing cleaned data is sequentially appended after the zero values ​​according to GPS time order, forming a complete slice of 11 data points.

[0046] Furthermore, including:

[0047] In step S4, the 14 feature parameters include 8 basic feature parameters and 6 variable feature parameters;

[0048] The eight basic characteristic parameters include: front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque, all of which directly use the corresponding data in the data slices.

[0049] The six variable characteristic parameters include: speed change, accelerator pedal opening change, engine speed change, engine intake manifold pressure change, engine DOC inlet temperature change, and driver-demanded torque change.

[0050] The calculation rule for each change characteristic parameter is as follows: Change at any time t = Basic characteristic parameter value at time t - Basic characteristic parameter value at the previous time t-1;

[0051] When t=0 is the first time point, the change is 0.

[0052] Furthermore, the calculation formulas for each characteristic parameter of the change are as follows:

[0053] Change in velocity: In the formula yes Front wheel speed at all times yes The speed of the front wheels at time t=0 (the first time). ;

[0054] Change in accelerator pedal opening: In the formula yes Constantly increase the accelerator pedal opening. yes The accelerator pedal opening at any given moment, when t=0 (the first moment). ;

[0055] Engine speed change: In the formula yes Constant engine speed. yes The engine speed at time t=0 (the first time step). ;

[0056] Engine intake manifold pressure change: In the formula yes Constant engine intake manifold pressure yes The engine intake manifold pressure at time t=0 (the first time). ;

[0057] Engine DOC temperature change: In the formula yes Constant engine DOC front temperature yes Engine DOC inlet temperature at time t=0 (first time). ;

[0058] Driver's required torque variation: In the formula yes The driver always needs torque. yes The driver's torque requirement at any given moment, when t=0 (the first moment). .

[0059] Furthermore, including:

[0060] In step S6, the structural components and core parameters of the deep neural network model are as follows:

[0061] Step S6.1, Fully connected layer 1: Input feature count = 14, Output feature count = 33, Enable bias = True;

[0062] Step S6.2, Batch Normalization Layer 1: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0063] Step S6.3, Fully connected layer 2: Input feature count = 33, Output feature count = 121, Enable bias = True;

[0064] Step S6.4, Batch Normalization Layer 2: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0065] Step S6.5, one-dimensional convolutional layer: number of input channels = 121, number of output channels = 45, kernel size = (3,), stride = (1,), padding = (1,);

[0066] Step S6.6, Batch Normalization Layer 3: Feature Dimension = 45, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0067] Step S6.7, Fully connected layer 3: Input feature count = 450, Output feature count = 29, Enable bias = True;

[0068] Step S6.8, Fully Connected Layer 4: Input feature count = 29, Output feature count = 23, Enable bias = True;

[0069] Step S6.9, Output layer: Input feature count = 23, Output feature count = 1, Enable bias = True;

[0070] Step S6.10, random deactivation layer: drop probability = 0.21649165607921678.

[0071] According to a second aspect of the present invention, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0072] The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps such as a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network.

[0073] According to a third aspect of the present invention, a computer-readable storage medium is provided, comprising: storing a computer program executable by an electronic device, wherein when the computer program is run on the electronic device, the electronic device performs steps such as a method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network.

[0074] According to a fourth aspect of the present invention, a China VI diesel engine vehicle platform is provided, comprising:

[0075] Electronic equipment for implementing steps such as a method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network;

[0076] The processor runs programs, and when the programs are running, they execute steps such as a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, based on data output from electronic devices.

[0077] Storage medium for storing programs that, when running, perform steps such as a method for calculating instantaneous fuel consumption of a diesel engine based on deep neural networks, on data output from electronic devices.

[0078] The above solution achieves the following beneficial technical effects:

[0079] This application systematically integrates the influence mechanism of multiple factors such as vehicle driving conditions, torque demand, regeneration state and energy control on engine transient fuel consumption, and realizes accurate modeling and calculation of transient fuel consumption through a deep neural network architecture.

[0080] This application supports the customized development of vehicle-level engine transient fuel consumption calculation models, and achieves differentiated simulation calculation capabilities with "a thousand vehicles, a thousand faces" through personalized modeling, adapting to the operating characteristics of different vehicles.

[0081] This application collects vehicle status information and calculates characteristic indicators through a T-BOX, inputs the characteristic indicators into a trained deep neural network model, and outputs an estimated value of instantaneous engine fuel consumption.

[0082] This application specifies the construction specifications for 9 types of data to be collected, 14 types of feature parameters, 11 data slicing rules, and 10 rows and 14 columns of feature samples.

[0083] This application specifies the specific structural components and corresponding core parameters of the deep neural network used in this method. Attached Figure Description

[0084] Figure 1 This is a flowchart of a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, provided by one or more embodiments of the present invention.

[0085] Figure 2 This is a structural diagram of a diesel engine instantaneous fuel consumption calculation system based on a deep neural network, provided by one or more embodiments of the present invention.

[0086] Figure 3 This is a schematic diagram of the process of calculating instantaneous fuel consumption of an engine according to a specific embodiment of the present invention.

[0087] Figure 4 This is a block diagram of an electronic device for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, provided by one or more embodiments of the present invention. Detailed Implementation

[0088] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0089] Figure 1 This is a flowchart of a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, provided by one or more embodiments of the present invention.

[0090] like Figure 1 The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network, as shown, includes:

[0091] S1. Data Collection: The vehicle-side T-BOX collects vehicle status information and sends it to the cloud.

[0092] Vehicle status information, including GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-required torque.

[0093] The vehicle-mounted T-BOX also collects one or more of the following information: GPS longitude, GPS latitude, GPS altitude, GPS positioning status, instrument speed, engine torque percentage, engine coolant temperature, and DPF regeneration status. After being uploaded to the cloud, this information, along with the vehicle status information, serves as the raw data for data cleaning.

[0094] S2. Data Cleaning: Cleaning the received raw vehicle status information in the cloud;

[0095] Data cleaning strategies include sorting by GPS time, removing duplicate values, and correcting outliers;

[0096] S3, Data Slicing: Based on the continuity of GPS time, extract the 11 most recently acquired cleaned data points and assemble them into a data slice;

[0097] When the total amount of data after cleaning is less than 11 records, it is padded with zeros forward to complete the data to 11 records, forming a data slice.

[0098] S4. Feature parameter calculation: 14 feature parameters are calculated for 11 data points in the data slice according to the set rules;

[0099] S5. Feature Sample Construction: Extract the last 10 data points from the data slice as samples;

[0100] The sample data is sorted by GPS time from smallest to largest, and combined with 14 feature parameters to construct a feature sample in the form of a numerical matrix with 10 rows and 14 columns;

[0101] S6. Model Inference: Input the feature samples into the pre-trained deep neural network model, and the deep neural network model outputs the instantaneous fuel consumption value of the engine.

[0102] The deep neural network model is a hybrid temporal model, including fully connected layers, batch normalization layers, one-dimensional convolutional layers, random deactivation layers, and an output layer, with ReLU as the activation function.

[0103] In this embodiment, it includes:

[0104] In step S1, the attributes of each vehicle status information include:

[0105] GPS time is in timestamp format, in seconds, and is an integer data type.

[0106] The front wheel speed is measured in km / h and is a floating-point number.

[0107] The data types for accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque are all integers, with units of %, rpm, kPa, ℃, ℃, ℃, %.

[0108] In this embodiment, it includes:

[0109] In step S2, the specific operations of the data cleaning strategy include:

[0110] Step S2.1: Sort the original vehicle status information in ascending order of GPS time;

[0111] Step S2.2: Delete the original vehicle status information with duplicate GPS times;

[0112] Step S2.3: Delete the original vehicle status information containing null values;

[0113] Step S2.4: Delete abnormal data according to preset filtering rules;

[0114] The preset filtering rules include:

[0115] Data showing a front wheel speed greater than 140 km / h is considered abnormal.

[0116] Accelerator pedal opening less than 0% or greater than 100% is considered abnormal data;

[0117] Data showing engine speeds below 0 rpm or above 8000 rpm is considered abnormal.

[0118] Engine intake manifold pressure less than 20 kPa or greater than 300 kPa is considered abnormal data;

[0119] Engine intake manifold temperatures below -50°C or above 150°C are considered abnormal data.

[0120] Data showing an engine DOC temperature below 100°C or above 800°C is considered abnormal.

[0121] Data with an ambient temperature below -50℃ or above 50℃ is considered abnormal.

[0122] Data showing a driver's required torque less than 0% or greater than 100% is considered abnormal.

[0123] In this embodiment, it includes:

[0124] In step S3, the specific method for filling "zero" forward is as follows:

[0125] The zero values ​​are placed at the beginning of the data slice, and the existing cleaned data is sequentially appended after the zero values ​​according to GPS time order, forming a complete slice of 11 data points.

[0126] In this embodiment, it includes:

[0127] In step S4, the 14 feature parameters include 8 basic feature parameters and 6 variable feature parameters;

[0128] The eight basic characteristic parameters include: front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque, all of which directly use the corresponding data in the data slices.

[0129] The six variable characteristic parameters include: speed change, accelerator pedal opening change, engine speed change, engine intake manifold pressure change, engine DOC inlet temperature change, and driver-demanded torque change.

[0130] The calculation rule for each change characteristic parameter is as follows: Change at any time t = Basic characteristic parameter value at time t - Basic characteristic parameter value at the previous time t-1;

[0131] When t=0 is the first time point, the change is 0.

[0132] In this embodiment, the calculation formulas for each characteristic parameter of the change are as follows:

[0133] Change in velocity: In the formula yes Front wheel speed at all times yes The speed of the front wheels at time t=0 (the first time). ;

[0134] Change in accelerator pedal opening: In the formula yes Constantly increase the accelerator pedal opening. yes The accelerator pedal opening at any given moment, when t=0 (the first moment). ;

[0135] Engine speed change: In the formula yes Constant engine speed. yes The engine speed at time t=0 (the first time step). ;

[0136] Engine intake manifold pressure change: In the formula yes Constant engine intake manifold pressure yes The engine intake manifold pressure at time t=0 (the first time). ;

[0137] Engine DOC temperature change: In the formula yes Constant engine DOC front temperature yes Engine DOC inlet temperature at time t=0 (first time). ;

[0138] Driver's required torque variation: In the formula yes The driver always needs torque. yes The driver's torque requirement at any given moment, when t=0 (the first moment). .

[0139] In this embodiment, it includes:

[0140] In step S6, the structural components and core parameters of the deep neural network model are as follows:

[0141] Step S6.1, Fully connected layer 1: Input feature count = 14, Output feature count = 33, Enable bias = True;

[0142] Step S6.2, Batch Normalization Layer 1: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0143] Step S6.3, Fully connected layer 2: Input feature count = 33, Output feature count = 121, Enable bias = True;

[0144] Step S6.4, Batch Normalization Layer 2: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0145] Step S6.5, one-dimensional convolutional layer: number of input channels = 121, number of output channels = 45, kernel size = (3,), stride = (1,), padding = (1,);

[0146] Step S6.6, Batch Normalization Layer 3: Feature Dimension = 45, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True;

[0147] Step S6.7, Fully connected layer 3: Input feature count = 450, Output feature count = 29, Enable bias = True;

[0148] Step S6.8, Fully Connected Layer 4: Input feature count = 29, Output feature count = 23, Enable bias = True;

[0149] Step S6.9, Output layer: Input feature count = 23, Output feature count = 1, Enable bias = True;

[0150] Step S6.10, random deactivation layer: drop probability = 0.21649165607921678.

[0151] Figure 2 This is a flowchart of a diesel engine instantaneous fuel consumption calculation system based on a deep neural network, provided by one or more embodiments of the present invention.

[0152] like Figure 2 The diesel engine instantaneous fuel consumption calculation system based on deep neural networks shown includes:

[0153] The data acquisition module is used by the vehicle-side T-BOX to collect vehicle status information and send the vehicle status information to the cloud;

[0154] Vehicle status information, including GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-required torque.

[0155] The vehicle-mounted T-BOX also collects one or more of the following information: GPS longitude, GPS latitude, GPS altitude, GPS positioning status, instrument speed, engine torque percentage, engine coolant temperature, and DPF regeneration status. After being uploaded to the cloud, this information, along with the vehicle status information, serves as the raw data for data cleaning.

[0156] The data cleaning module is used to clean the received raw vehicle status information in the cloud.

[0157] Data cleaning strategies include sorting by GPS time, removing duplicate values, and correcting outliers;

[0158] The data slicing module is used to extract the 11 most recently acquired cleaned data points and assemble them into a data slice based on the continuity of GPS time.

[0159] When the total amount of data after cleaning is less than 11 records, it is padded with zeros forward to complete the data to 11 records, forming a data slice.

[0160] The feature parameter calculation module is used to calculate 14 feature parameters from 11 data points within a data slice according to set rules.

[0161] The feature sample construction module is used to extract the last 10 data points from the data slice as samples.

[0162] The sample data is sorted by GPS time from smallest to largest, and combined with 14 feature parameters to construct a feature sample in the form of a numerical matrix with 10 rows and 14 columns;

[0163] The model inference module is used to input feature samples into a pre-trained deep neural network model, and the deep neural network model outputs the instantaneous fuel consumption value of the engine.

[0164] The deep neural network model is a hybrid temporal model, including fully connected layers, batch normalization layers, one-dimensional convolutional layers, random deactivation layers, and an output layer, with ReLU as the activation function.

[0165] It is worth noting that although this system / device only discloses the above-mentioned modules / units, it does not mean that this system / device is limited to the above-mentioned basic functional modules. On the contrary, what this invention intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can add one or more functional modules in combination with the prior art to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. It cannot be assumed that the scope of protection of the claims of this invention is limited to the above-disclosed basic functional modules just because this embodiment only discloses a few basic functional modules.

[0166] In one specific embodiment, a method for calculating the instantaneous fuel consumption of a China VI diesel engine based on a deep neural network is disclosed, such as... Figure 3 As shown. This embodiment specifically includes the following steps:

[0167] S1. Data Acquisition and Transmission: The vehicle-mounted T-BOX collects information such as GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC temperature, ambient temperature, and driver's required torque (see Table 1 for details), and transmits the above information to the vehicle network cloud platform.

[0168] Table 1, Field Description

[0169] serial number Field Name illustrate 1 GPS time Timestamp format, unit: seconds, integer 2 Front wheel speed Unit: km / h, floating-point number 3 Accelerator pedal opening Unit %, integer 4 Engine speed Unit: rpm, integer 5 Engine intake manifold pressure Unit: kPa, integer 6 Engine intake manifold temperature Unit: °C, integer 7 Engine DOC inlet temperature Unit: °C, integer 8 Ambient temperature Unit: °C, integer 9 Driver's required torque Unit %, integer

[0170] S2. Data Cleaning: Cleaning the received raw data on the cloud platform. Data cleaning methods include, but are not limited to, sorting by time, deleting duplicate values, and correcting outliers. Specifically, this includes: Data is arranged according to GPS time from smallest to largest. Delete duplicate GPS time data; Delete empty data; Delete data that meets the filtering rules shown in Table 2.

[0171] Table 2, Summary Table of Filtering Rules for Specific Fields

[0172] serial number Field Name Filtering rules 1 GPS time —— 2 Front wheel speed Greater than 140 3 Accelerator pedal opening Less than 0 or greater than 100 4 Engine speed Less than 0 or greater than 8000 5 Engine intake manifold pressure Less than 20 or greater than 300 6 Engine intake manifold temperature Less than -50 or greater than 150 7 Engine DOC inlet temperature Less than 100 or greater than 800 8 Ambient temperature Less than -50 or greater than 50 9 Driver's required torque Less than 0 or greater than 100

[0173] S3. Data Slicing: Based on the continuity of GPS time, extract the 11 most recently acquired data points and assemble these 11 data points into a data slice. When the total amount of data is less than 11, fill the data forward with zeros. For example, if there are only 3 data points, fill the first 7 data points with zeros. The existing 3 data points are then connected in chronological order after the filled data to form a complete set of 11 data points.

[0174] S4. Feature Parameter Calculation: For the data within the slice, calculate the 14 feature parameters required to estimate engine oil performance according to the following rules {REF _Ref194068238\h \*MERGEFORMAT}: front wheel speed, accelerator pedal position, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC temperature, ambient temperature, driver-demanded torque, speed change, accelerator pedal position change, engine speed change, engine intake manifold pressure change, engine DOC temperature change, and driver-demanded torque change.

[0175] Front wheel speed: Directly use the front wheel speed data within the slice;

[0176] Accelerator pedal opening: Directly use the accelerator pedal opening data within the slice;

[0177] Engine speed: Engine speed data within the slice is used directly;

[0178] Engine intake manifold pressure: Engine intake manifold pressure data from within the slice is used directly;

[0179] Engine intake manifold temperature: Engine intake manifold temperature data from the slice is used directly;

[0180] Engine DOC inlet temperature: Directly use engine DOC inlet temperature data from the slice;

[0181] Ambient temperature: Use the ambient temperature data within the slice directly;

[0182] Driver torque demand: Directly use the driver torque demand data within the slice;

[0183] Change in velocity: In the formula yes Front wheel speed at all times yes The speed of the front wheels at time t=0 (the first time). ;

[0184] Change in accelerator pedal opening: In the formula yes Constantly increase the accelerator pedal opening. yes The accelerator pedal opening at any given moment, when t=0 (the first moment). ;

[0185] Engine speed change: In the formula yes Constant engine speed. yes The engine speed at time t=0 (the first time step). ;

[0186] Engine intake manifold pressure change: In the formula yes Constant engine intake manifold pressure yes The engine intake manifold pressure at time t=0 (the first time). ;

[0187] Engine DOC temperature change: In the formula yes Constant engine DOC front temperature yes Engine DOC inlet temperature at time t=0 (first time). ;

[0188] Driver's required torque variation: In the formula yes The driver always needs torque. yes The driver's torque requirement at any given moment, when t=0 (the first moment). ;

[0189] S5. Feature Sample Construction: In the current slice, extract the last 10 data points as samples. The data in the samples are sorted in ascending order of GPS time and contain 14 features, which constitute a numerical matrix of 10 rows and 14 columns.

[0190] S6. Model Inference: Input the above feature samples into the pre-trained deep neural network model, and the model outputs the instantaneous fuel consumption value of the engine. This deep neural network model is a time series model, and its structure is shown in Table 3.

[0191] Table 3. Structural parameters of deep neural network models

[0192] serial number Structural component name Core parameters 1 Fully connected layer 1 Input feature count = 14, output feature count = 33, enable bias = True 2 Batch Normalization Layer 1 Feature dimension = 10, numerical precision = 1e-05, momentum coefficient = 0.1, enable affine transformation = True, tracking runtime statistics = True 3 Fully connected layer 2 Input feature count = 33, output feature count = 121, enable bias = True 4 Batch Normalization Layer 2 Feature dimension = 10, numerical precision = 1e-05, momentum coefficient = 0.1, enable affine transformation = True, tracking runtime statistics = True 5 One-dimensional convolutional layer Input channels = 121, output channels = 45, kernel size = (3,), stride = (1,), padding = (1,) 6 Batch Normalization Layer 3 Feature dimension = 45, numerical precision = 1e-05, momentum coefficient = 0.1, enable affine transformation = True, tracking runtime statistics = True 7 Fully connected layer 3 Input feature count = 450, output feature count = 29, enable bias = True 8 Fully connected layer 4 Input feature count = 29, output feature count = 23, enable bias = True 9 Output layer Input feature count = 23, output feature count = 1, enable bias = True 10 Random deactivation layer Discard probability = 0.21649165607921678 11 Activation function ReLU

[0193] In another specific embodiment, the key technical points and protection points are listed below:

[0194] A method for calculating the instantaneous fuel consumption of a China VI diesel engine based on a deep neural network is proposed. This method collects vehicle status information and calculates characteristic indicators through a T-BOX, and inputs the characteristic indicators into a trained deep neural network model. After calculation by the deep neural network model, the estimated value of the engine's instantaneous fuel consumption is output.

[0195] In this embodiment, a method for calculating the instantaneous fuel consumption of a China VI diesel engine based on a deep neural network is proposed. The method includes:

[0196] S1. Data Acquisition: The vehicle-mounted T-BOX collects information such as GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC temperature, ambient temperature, and driver's required torque, and uploads the above information to the cloud.

[0197] S2. Data Cleaning: Cleaning the received raw data in the cloud. Data cleaning methods include, but are not limited to, sorting by time, deleting duplicate values, and correcting outliers.

[0198] S3. Data Slicing: Based on the continuity of GPS time, extract the 11 most recently acquired data points and assemble these 11 data points into a data slice; when the total amount of data is less than 11, fill the data forward with zeros.

[0199] S4. Feature Parameter Calculation: Calculate 14 feature parameters for the data within the slice according to the set rules.

[0200] S5. Feature Sample Construction: In the current slice, extract the last 10 data points as samples. The data in the samples are sorted in ascending order of GPS time and contain 14 features, which constitute a numerical matrix of 10 rows and 14 columns.

[0201] S6. Model Inference: Input the above feature samples into the pre-trained deep neural network model, and the model outputs the instantaneous fuel consumption value of the engine. In step S1, the collected and transmitted data are GPS time, GPS longitude, GPS latitude, GPS altitude, GPS positioning status, instrument vehicle speed, engine speed, engine torque percentage, engine coolant temperature, and DPF regeneration status;

[0202] The characteristic parameters calculated in step S4 are front wheel speed, accelerator pedal position, engine speed, intake manifold pressure, intake manifold temperature, DOC inlet temperature, ambient temperature, required torque, speed change, accelerator pedal position change, engine speed change, intake manifold pressure change, DOC inlet temperature change, and required torque change.

[0203] In step S6, the deep neural network model structure and main parameters used are shown in Table 3.

[0204] Figure 4 This is a block diagram of an electronic device for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network, provided by one or more embodiments of the present invention.

[0205] like Figure 4 As shown, this application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0206] The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps of a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network.

[0207] This application also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a method for calculating the instantaneous fuel consumption of a diesel engine based on a deep neural network.

[0208] This application also provides a China VI diesel engine vehicle platform, including:

[0209] Electronic equipment for implementing a method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network;

[0210] The processor runs a program, and when the program runs, it executes the steps of a diesel engine instantaneous fuel consumption calculation method based on deep neural networks from the data output by the electronic device.

[0211] Storage medium for storing programs that, when running, execute steps of a method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network on data output from an electronic device.

[0212] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0213] The electronic device comprises a hardware layer, an operating system layer running on top of the hardware layer, and an application layer running on the operating system. The hardware layer includes hardware such as a central processing unit (CPU), a memory management unit (MMU), and memory. The operating system can be any one or more computer operating systems that control the electronic device through processes, such as Linux, Unix, Android, iOS, or Windows. Furthermore, in this embodiment of the invention, the electronic device can be a smartphone, tablet computer, or other handheld device, or a desktop computer, portable computer, or other electronic device; there is no particular limitation in this embodiment.

[0214] In this embodiment of the invention, the executing entity for electronic device control can be an electronic device itself, or a functional module within an electronic device capable of calling and executing a program. The electronic device can obtain the firmware corresponding to the storage medium. This firmware is provided by the supplier, and different storage media may have the same or different firmware; no limitation is made here. After obtaining the firmware corresponding to the storage medium, the electronic device can write this firmware into the storage medium; specifically, it burns the firmware corresponding to the storage medium into the storage medium. The process of burning the firmware into the storage medium can be implemented using existing technology, and will not be elaborated upon in this embodiment of the invention.

[0215] Electronic devices can also obtain reset commands corresponding to the storage media. The reset commands corresponding to the storage media are provided by the supplier. The reset commands corresponding to different storage media can be the same or different, and no restrictions are imposed here.

[0216] At this time, the storage medium of the electronic device is a storage medium on which the corresponding firmware has been written. The electronic device can respond to the reset command corresponding to the storage medium on which the corresponding firmware has been written, thereby resetting the storage medium on which the corresponding firmware has been written according to the reset command. The process of resetting the storage medium according to the reset command can be implemented by existing technology and will not be described in detail in this embodiment of the invention.

[0217] For ease of description, the above devices are described separately by function as various units and modules. Of course, in implementing this application, the functions of each unit and module can be implemented in one or more software and / or hardware.

[0218] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined.

[0219] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0220] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network, characterized in that, include: S1. Data Acquisition: The vehicle-mounted T-BOX collects vehicle status information and sends the vehicle status information to the cloud; S2. Data Cleaning: Cleaning the received raw vehicle status information in the cloud; Data cleaning strategies include sorting by GPS time, removing duplicate values, and correcting outliers; S3, Data Slicing: Based on the continuity of GPS time, extract the latest N cleaned data and assemble them into a data slice; When the total amount of data after cleaning is less than N records, it is padded with "zero" forward to complete the data to N records, forming a data slice; S4. Feature parameter calculation: For N data points within the data slice, M feature parameters are calculated according to a set rule; S5. Feature Sample Construction: Extract the last N-1 data points from the data slice as samples; The sample data are sorted by GPS time from smallest to largest, and combined with the M feature parameters, a feature sample in the form of a numerical matrix with N-1 rows and M columns is constructed. S6. Model Inference: Input the feature samples into a pre-trained deep neural network model, and the deep neural network model outputs the instantaneous fuel consumption value of the engine. The deep neural network model is a hybrid temporal model, comprising a fully connected layer, a batch normalization layer, a one-dimensional convolutional layer, a random deactivation layer, and an output layer, with ReLU as the activation function.

2. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 1, characterized in that, include: In step S1, the vehicle status information includes GPS time, front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-required torque. The attributes of each vehicle status information include: GPS time is in timestamp format, in seconds, and is an integer data type. The front wheel speed is measured in km / h and is a floating-point number. The data types for accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque are all integers, with units of %, rpm, kPa, ℃, ℃, ℃, %.

3. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 1, characterized in that, include: In step S2, the specific operations of the data cleaning strategy include: Step S2.1: Sort the original vehicle status information in ascending order of GPS time; Step S2.2: Delete the original vehicle status information with duplicate GPS times; Step S2.3: Delete the original vehicle status information containing null values; Step S2.4: Delete abnormal data according to preset filtering rules; The preset filtering rules include: Data showing a front wheel speed greater than 140 km / h is considered abnormal. Accelerator pedal opening less than 0% or greater than 100% is considered abnormal data; Data showing engine speeds below 0 rpm or above 8000 rpm is considered abnormal. Engine intake manifold pressure less than 20 kPa or greater than 300 kPa is considered abnormal data; Engine intake manifold temperatures below -50°C or above 150°C are considered abnormal data. Data showing an engine DOC temperature below 100°C or above 800°C is considered abnormal. Data with an ambient temperature below -50℃ or above 50℃ is considered abnormal. Data showing a driver's required torque less than 0% or greater than 100% is considered abnormal.

4. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 1, characterized in that, include: In step S3, the specific method for filling "zero" forward is as follows: The zero values ​​are placed at the beginning of the data slice, and the existing cleaned data is sequentially appended after the zero values ​​according to GPS time order to form a complete slice of N data.

5. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 1, characterized in that, include: In step S4, the M feature parameters include 8 basic feature parameters and 6 variable feature parameters; The eight basic characteristic parameters include: front wheel speed, accelerator pedal opening, engine speed, engine intake manifold pressure, engine intake manifold temperature, engine DOC inlet temperature, ambient temperature, and driver-demanded torque, all of which directly use the corresponding data in the data slices. The six variable characteristic parameters include: speed change, accelerator pedal opening change, engine speed change, engine intake manifold pressure change, engine DOC inlet temperature change, and driver-demanded torque change. The calculation rule for each change characteristic parameter is as follows: Change at any time t = Basic characteristic parameter value at time t - Basic characteristic parameter value at the previous time t-1; When t=0 is the first time point, the change is 0.

6. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 5, characterized in that, The calculation formulas for each characteristic parameter of the change are as follows: Change in velocity: In the formula yes Front wheel speed at all times yes The speed of the front wheels at time t=0 (the first time). ; Change in accelerator pedal opening: In the formula yes Constantly increase the accelerator pedal opening. yes The accelerator pedal opening at any given moment, when t=0 (the first moment). ; Engine speed change: In the formula yes Constant engine speed. yes The engine speed at time t=0 (the first time step). ; Engine intake manifold pressure change: In the formula yes Constant engine intake manifold pressure yes The engine intake manifold pressure at time t=0 (the first time). ; Engine DOC temperature change: In the formula yes Constant engine DOC front temperature yes Engine DOC inlet temperature at time t=0 (first time). ; Driver's required torque variation: In the formula yes The driver always needs torque. yes The driver's torque requirement at any given moment, when t=0 (the first moment). .

7. The method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network according to claim 1, characterized in that, include: In step S6, the structural components and core parameters of the deep neural network model are as follows: Step S6.1, Fully connected layer 1: Input feature count = 14, Output feature count = 33, Enable bias = True; Step S6.2, Batch Normalization Layer 1: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True; Step S6.3, Fully connected layer 2: Input feature count = 33, Output feature count = 121, Enable bias = True; Step S6.4, Batch Normalization Layer 2: Feature Dimension = 10, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True; Step S6.5, one-dimensional convolutional layer: number of input channels = 121, number of output channels = 45, kernel size = (3,), stride = (1,), padding = (1,); Step S6.6, Batch Normalization Layer 3: Feature Dimension = 45, Numerical Precision = 1e-05, Momentum Coefficient = 0.1, Enable Affine Transformation = True, Track Run Statistics = True; Step S6.7, Fully connected layer 3: Input feature count = 450, Output feature count = 29, Enable bias = True; Step S6.8, Fully Connected Layer 4: Input feature count = 29, Output feature count = 23, Enable bias = True; Step S6.9, Output layer: Input feature count = 23, Output feature count = 1, Enable bias = True; Step S6.10, random deactivation layer: drop probability = 0.21649165607921678.

8. An electronic device, characterized in that, include: The processor, communication interface, memory, and communication bus are connected, with the processor, communication interface, and memory communicating with each other via the communication bus. The memory stores a computer program that, when executed by a processor, causes the processor to perform the steps of the diesel engine instantaneous fuel consumption calculation method based on a deep neural network as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, include: The device stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of the method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network as described in any one of claims 1 to 7.

10. A China VI diesel engine vehicle platform, characterized in that, include: An electronic device for implementing the steps of the diesel engine instantaneous fuel consumption calculation method based on a deep neural network as described in any one of claims 1 to 7; The processor runs a program, and when the program runs, it executes the steps of the diesel engine instantaneous fuel consumption calculation method based on a deep neural network as described in any one of claims 1 to 7 from the data output by the electronic device. A storage medium for storing a program that, when running, performs the steps of the method for calculating instantaneous fuel consumption of a diesel engine based on a deep neural network as described in any one of claims 1 to 7 on data output from an electronic device.

Images