CN113657651A - Diesel vehicle emission prediction method, medium and equipment based on deep migration learning - Google Patents
Diesel vehicle emission prediction method, medium and equipment based on deep migration learning Download PDFInfo
- Publication number
- CN113657651A CN113657651A CN202110852663.6A CN202110852663A CN113657651A CN 113657651 A CN113657651 A CN 113657651A CN 202110852663 A CN202110852663 A CN 202110852663A CN 113657651 A CN113657651 A CN 113657651A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- source domain
- data
- target domain
- domain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000013508 migration Methods 0.000 title claims abstract description 26
- 230000005012 migration Effects 0.000 title claims abstract description 26
- 230000006870 function Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000009826 distribution Methods 0.000 claims abstract description 23
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000010276 construction Methods 0.000 claims abstract description 3
- 238000004590 computer program Methods 0.000 claims description 13
- 239000007789 gas Substances 0.000 claims description 11
- 239000000446 fuel Substances 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- XSQUKJJJFZCRTK-UHFFFAOYSA-N Urea Chemical compound NC(N)=O XSQUKJJJFZCRTK-UHFFFAOYSA-N 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 4
- 239000004202 carbamide Substances 0.000 claims description 4
- 239000007788 liquid Substances 0.000 claims description 4
- 210000002569 neuron Anatomy 0.000 claims description 4
- BUCXEFZXWKUCCY-UHFFFAOYSA-N 4-methyl-3-(2-phenylethyl)-1,2,4-oxadiazol-5-one Chemical compound O1C(=O)N(C)C(CCC=2C=CC=CC=2)=N1 BUCXEFZXWKUCCY-UHFFFAOYSA-N 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000002229 photoelectron microspectroscopy Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 2
- 230000002159 abnormal effect Effects 0.000 claims description 2
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 2
- 238000012217 deletion Methods 0.000 claims description 2
- 230000037430 deletion Effects 0.000 claims description 2
- 239000002828 fuel tank Substances 0.000 claims description 2
- 238000002347 injection Methods 0.000 claims description 2
- 239000007924 injection Substances 0.000 claims description 2
- 239000010705 motor oil Substances 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 239000001301 oxygen Substances 0.000 claims description 2
- 229910052760 oxygen Inorganic materials 0.000 claims description 2
- 230000001902 propagating effect Effects 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 238000011144 upstream manufacturing Methods 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 2
- 238000007599 discharging Methods 0.000 claims 1
- 238000012805 post-processing Methods 0.000 claims 1
- 230000000717 retained effect Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 10
- 239000000243 solution Substances 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Abstract
The invention relates to a diesel vehicle emission prediction method, medium and equipment based on deep migration learning, which comprises the steps of constructing a diesel vehicle emission prediction model by utilizing computer equipment, and then predicting NOx emission of a target domain vehicle by using the model; the construction steps of the prediction model are as follows: s10: acquiring source domain vehicle related data and target domain vehicle driving related data and preprocessing the acquired data; s20: the pre-training feature projection module is used for projecting the source domain vehicle features and the target domain vehicle features into a public subspace; s30: the pre-training tail gas prediction module is used for building a full-connection neural network prediction model with double hidden layers by using the reconstructed source domain data characteristics and labels; s40: and combining the pre-trained projection module and the prediction module to construct a migration model, considering the data distribution difference between the source domain vehicle and the target domain vehicle, adding KL divergence between the source domain and the target domain into the loss function part, enabling the target domain data distribution to be close to the source domain data distribution, and finely adjusting the whole model.
Description
Technical Field
The invention relates to the technical field of diesel vehicle exhaust prediction in the field of environmental monitoring, in particular to a diesel vehicle emission prediction method, medium and equipment based on deep migration learning.
Background
With the development of human society, the usage amount of motor vehicles is continuously increased, and the exhaust emission of the motor vehicles brings great harm to the environment. Therefore, it is necessary to strictly control the exhaust emission to realize effective treatment of the air environment. The exhaust emission of diesel vehicles in the exhaust pollution accounts for a considerable proportion, so that the monitoring of the exhaust emission of the diesel vehicles is particularly important.
Most of traditional diesel vehicle exhaust prediction schemes are based on deep learning or physical modeling, and exhaust prediction models are constructed by using relevant data of diesel vehicles during running. These methods often require a large amount of high-quality data to construct an accurate model, however, in actual monitoring, it is difficult to acquire a large amount of high-quality data. Therefore, most of data given to common predicted vehicle models is sparse. Aiming at sparse data, a reliable prediction model is established by utilizing an exhaust gas prediction model of an existing vehicle model and existing sparse data, so that a scheme for reducing time and economic cost in data monitoring is very necessary.
Disclosure of Invention
The diesel vehicle emission prediction method, medium and equipment based on deep migration learning can solve the problem of poor diesel vehicle exhaust prediction accuracy under the condition of insufficient training data.
In order to achieve the purpose, the invention adopts the following technical scheme:
a diesel vehicle emission prediction method based on deep migration learning comprises the following steps:
the method comprises the steps of constructing a diesel vehicle emission prediction model by utilizing computer equipment, and then predicting NOx emission of a target domain vehicle by using the model;
wherein, the construction of the diesel vehicle emission prediction model by utilizing the computer equipment comprises the following steps,
s10: acquiring source domain vehicle related data and target domain vehicle driving related data and preprocessing the acquired data;
s20: the pre-training feature projection module is used for projecting the source domain vehicle features and the target domain vehicle features into a public subspace;
s30: the pre-training tail gas prediction module is used for building a full-connection neural network prediction model with double hidden layers by using the reconstructed source domain data characteristics and labels;
s40: and combining the pre-trained projection module and the prediction module to construct a migration model, considering the data distribution difference between the source domain vehicle and the target domain vehicle, adding KL divergence between the source domain and the target domain to the loss function part to enable the target domain data distribution to be close to the source domain data distribution, and finely adjusting the whole model to obtain a stable diesel vehicle emission prediction model.
In another aspect, the present invention also discloses a computer readable storage medium storing a computer program, which when executed by a processor causes the processor to perform the steps of the method as described above.
In a third aspect, the present invention also discloses a computer device comprising a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method as described above.
According to the technical scheme, the diesel vehicle emission prediction method based on deep migration learning is a field-adaptive method, the concentration of NOx is estimated under the condition that the related data of the NOx of the diesel vehicle is sparse, the accuracy is high, the method migrates the high-quality NOx data of other diesel vehicles, does not need to collect the sparse data, can greatly reduce the cost, is good in real-time performance and high in practicability, and is beneficial to popularization.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic diagram of a network model of the method of the present invention;
FIG. 3 is a line graph illustrating accuracy verification of reconstructed features according to an embodiment of the present invention;
FIG. 4 is a predicted line graph according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention.
As shown in fig. 1, the method for predicting the emission of a diesel vehicle based on deep migration learning according to the embodiment includes:
s10: acquiring source domain vehicle related data with sufficient training data and target domain vehicle related data with sparse training data during driving, and preprocessing the acquired data;
s20: the pre-training feature projection module is used for projecting the source domain vehicle features and the target domain vehicle features into a public subspace;
s30: the pre-training tail gas prediction module is used for building a full-connection neural network prediction model with double hidden layers by using the reconstructed source domain data characteristics and labels;
s40: and combining the pre-trained projection module and the prediction module to construct a migration model, considering the data distribution difference between the source domain vehicle and the target domain vehicle, adding KL divergence between the source domain and the target domain into the loss function part, enabling the target domain data distribution to be close to the source domain data distribution, and finely adjusting the whole model.
The following is a detailed description:
the S10 obtains source domain vehicle-related data with sufficient training data and target domain vehicle-related data with sparse training data during driving, and preprocesses the collected data, specifically including the following steps:
s11: various data of the source domain diesel vehicle during running are obtained through vehicle-mounted exhaust gas detection equipment (PEMS), and a small amount of data of the target domain diesel vehicle during running are collected.
The data collected from PEMS includes license plate, terminal number, engine speed, actual output torque percentage, engine water temperature, engine fuel temperature, engine oil temperature, aftertreatment downstream NOx value, aftertreatment downstream oxygen percentage, barometric pressure, ambient temperature, aftertreatment exhaust gas mass flow, urea tank liquid level percentage, urea tank temperature, vehicle speed, accelerator pedal opening, single trip mileage, total mileage, engine instantaneous fuel injection, engine instantaneous fuel consumption, engine average fuel consumption, engine accumulated fuel consumption, battery voltage, fuel tank liquid level, engine accumulated run time, longitude, latitude, SCR upstream temperature, SCR downstream temperature;
s12: and carrying out abnormal value processing, missing value processing, non-correlation data deletion and normalization operation on the data.
The S20 pre-training feature projection module projects the source domain vehicle features and the target domain vehicle features into a common subspace, and specifically includes:
s21: training a sparse self-encoder, taking the feature data of a source domain vehicle and a target domain vehicle as input and output actual values, setting the number of neurons of a hidden layer to be three times of the input, and setting a loss function as follows:
whereinIn order to be the initial source-domain feature,to reconstruct the source domain features, DKLIn order to solve the KL divergence function, rho is the preset neuron activation probability,w is the weight, and beta and lambda are empirical parameters;
s22: minimizing the loss function in S21 by gradient descent, reversely propagating and updating the weight, stopping iteration after network convergence, and retaining the weight from the input layer to the hidden layer, which is marked as W1Output the hidden layer h1Retention, h1The calculation formula is as follows:
h1=fSAE_1(W1,Xs)=W1·Xs,
s23: outputting S22 hidden layer h1Training as new sparse self-encoder input according to S22, after network convergence, keeping weight input layer to hidden layer weight, and recording as W2Output the hidden layer h2Retention, h2The calculation formula is as follows:
h2=fSAE_2(W2,h1)=W2·h1
s24: and then taking the output of the S23 sparse self-coding hidden layer as the input of a new sparse self-coder, training according to S23, after network convergence, keeping the weight from the input layer to the hidden layer, and recording as W3(ii) a Outputting the hidden layer h3Retention, h3The calculation formula is as follows:
h3=fSAE_3(W3,h2)=W3·h2
s25: stacking sparse autoencoders, with W1,W2,W3Adding an output layer (weight random initialization) as an initial weight, constructing a multi-layer sparse self-encoder, taking a loss function as a first item in S21, and after reverse iteration integral optimization, keeping an updated weight W1',W2',W3'
S26: verifying the reliability of the reconstruction characteristics, respectively establishing a source domain prediction model by using the source domain original characteristics and the reconstruction characteristics, selecting 100 samples as a test set, and enabling the rest samples to be 8: 2 into training and validation sets, the prediction results over 100 samples are shown in fig. 3, and the NOx prediction information using the original and reconstructed features as inputs on the source domain data, respectively, and the true values of the selected 100 source domain sample points are given in fig. 3. As can be seen from FIG. 3, the accuracy of the reconstruction characteristics is effectively verified by the high coincidence of the respective predicted values and the true values of the two input training networks;
the S30 pre-training exhaust gas prediction module builds a double-hidden-layer fully-connected neural network prediction model by using the reconstructed source domain data characteristics and labels, and comprises the following specific steps:
s31: based on S24, using reserved third-layer hidden layer output data as input of a prediction model, using a downstream NOx value as output, constructing a double-hidden-layer fully-connected neural network, selecting Relu as an activation function, selecting Adam as an optimizer, setting the iteration number to 15000 times, and stopping iteration when the loss value is not reduced for more than 10 continuous times;
s32: minimizing the Mean Square Error (MSE) of the real NOx value and the predicted value, and updating hidden layer weights W4, W5 and W6 through back propagation until convergence;
the S40 combines the pre-trained projection module and the prediction module to construct a migration model, considers the data distribution difference between the source domain vehicle and the target domain vehicle, adds the KL divergence between the source domain and the target domain in the loss function part to enable the target domain data distribution to be close to the source domain, and finely adjusts the whole model by the following specific steps:
s41: based on the weight W of steps S20, S301',W2',W3',W4,W5,W6Multiplexing the weights, building a 6-layer migration neural network, inputting various characteristics of a source domain vehicle and a target domain vehicle, and predicting an emission value with a label of NOx;
s42: by adding KL divergence in the loss function and minimizing the loss function, the distribution difference of the source domain vehicle and the target domain vehicle in the subspace is reduced, the similarity of the source domain vehicle and the target domain vehicle is improved, and therefore knowledge transfer from the source domain vehicle to the target domain vehicle is achieved, and the loss function is as follows:
whereinNO for i-th sample of source domain vehiclexThe value of the one or more of the one,is the ith vehicle NO of the source area vehiclexBeta is an empirical parameter, DKLTo solve the KL divergence function, psFor the feature distribution of the source domain features in the subspace, ρtIs the characteristic distribution of the target domain vehicle among the subspaces.
S43: the whole migration prediction network is finely adjusted through gradient descent back propagation, and after the network is converged, the model is used for predicting NO of the target domain vehiclexEmission, NOxThe emission formula is as follows:
preNOx=freg_3(W6,freg_2(W5,freg_1(W4,fSAE_3(W3',fSAE_2(W2',fSAE_1(W1',XS))))))
wherein f isreg_x/SAE_x(W,X)=W·X。
Fig. 4 gives an exemplary graph of the results obtained on the basis of the present invention. As shown in fig. 4, the predicted values of the 100 sample points in the deep migration learning and the deep learning, and the true values of the 100 sample points are given in the final result diagram. From the information in the figure, the method based on the patent has better prediction performance and is beneficial to popularization. From the information in fig. 4, the method provided by the present invention can realize effective prediction by relying on high-quality NOx data of other diesel vehicles under the condition that NOx data of the diesel vehicle is sparse, and the method does not need to use other expensive equipment, so that the cost can be greatly reduced.
In another aspect, the present invention also discloses a computer readable storage medium storing a computer program, which when executed by a processor causes the processor to perform the steps of the method as described above.
In a third aspect, the present invention also discloses a computer device comprising a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method as described above.
It is understood that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, the example and the beneficial effects of the related contents can refer to the corresponding parts in the method.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (9)
1. A diesel vehicle emission prediction method based on deep migration learning is characterized by comprising the following steps: includes constructing a diesel vehicle emission prediction model using a computer device and then using the model to predict NO for a target domain vehiclexDischarging;
wherein, the construction of the diesel vehicle emission prediction model by utilizing the computer equipment comprises the following steps,
s10: acquiring source domain vehicle related data and target domain vehicle driving related data and preprocessing the acquired data;
s20: the pre-training feature projection module is used for projecting the source domain vehicle features and the target domain vehicle features into a public subspace;
s30: the pre-training tail gas prediction module is used for building a full-connection neural network prediction model with double hidden layers by using the reconstructed source domain data characteristics and labels;
s40: and combining the pre-trained projection module and the prediction module to construct a migration model, considering the data distribution difference between the source domain vehicle and the target domain vehicle, adding KL divergence between the source domain and the target domain to the loss function part to enable the target domain data distribution to be close to the source domain data distribution, and finely adjusting the whole model to obtain a stable diesel vehicle emission prediction model.
2. The diesel vehicle emission prediction method based on deep migration learning of claim 1, wherein: the S10: the method comprises the steps of obtaining source domain vehicle related data and target domain vehicle related data during running and preprocessing the collected data, and specifically comprises the following steps:
s11: acquiring various data of a source domain diesel vehicle during running through vehicle-mounted exhaust gas detection equipment (PEMS), and collecting a set amount of data of a target domain diesel vehicle during running;
s12: and carrying out abnormal value processing, missing value processing, non-correlation data deletion and normalization operation on the data.
3. The diesel vehicle emission prediction method based on deep migration learning of claim 2, wherein: the data collected from the vehicle-mounted tail gas detection device in the step S11 includes a license plate, a terminal number, an engine speed, an actual output torque percentage, an engine water temperature, an engine fuel temperature, an engine oil temperature, an aftertreatment downstream NOx value, an aftertreatment downstream oxygen percentage, an atmospheric pressure, an ambient temperature, an aftertreatment exhaust gas mass flow, a urea tank liquid level percentage, a urea tank temperature, a vehicle speed, an accelerator pedal opening, a single-trip mileage, a total mileage, an engine instantaneous fuel injection quantity, an engine instantaneous fuel consumption rate, an engine average fuel consumption rate, an engine accumulated fuel consumption, a battery voltage, a fuel tank liquid level, an engine accumulated operation time, a longitude, a latitude, an SCR upstream temperature, and an SCR downstream temperature.
4. The diesel vehicle emission prediction method based on deep migration learning of claim 3, wherein: the S20 pre-training feature projection module projects the source domain vehicle features and the target domain vehicle features into a common subspace, and specifically includes:
s21: training a sparse self-encoder, taking the feature data of a source domain vehicle and a target domain vehicle as input and output actual values, setting the number of neurons of a hidden layer to be three times of the input, and setting a loss function as follows:
whereinIn order to be the initial source-domain feature,to reconstruct the source domain features, DKLIn order to solve the KL divergence function, rho is the preset neuron activation probability,w is the weight, and beta and lambda are empirical parameters;
s22: minimizing the loss function in S21 by gradient descent, reversely propagating and updating the weight, stopping iteration after network convergence, and retaining the weight from the input layer to the hidden layer, which is marked as W1Output the hidden layer h1Retention, h1The calculation formula is as follows:
h1=fSAE_1(W1,Xs)=W1·Xs,
s23: outputting S22 hidden layer h1Training as new sparse self-encoder input according to S22, after network convergence, keeping weight input layer to hidden layer weight, and recording as W2Output the hidden layer h2Retention, h2The calculation formula is as follows:
h2=fSAE_2(W2,h1)=W2·h1
s24: and then taking the output of the S23 sparse self-coding hidden layer as the input of a new sparse self-coder, training according to S23, after network convergence, keeping the weight from the input layer to the hidden layer, and recording as W3(ii) a Outputting the hidden layer h3Retention, h3The calculation formula is as follows:
h3=fSAE_3(W3,h2)=W3·h2;
s25: stacking sparse autoencoders, with W1,W2,W3Adding an output layer as an initial weight, building a multi-layer sparse self-encoder, wherein a loss function is a first item in S21, and after reverse iteration integral optimization, keeping an updated weight W1',W2',W3';;
S26: verifying the authenticity of the reconstructed features.
5. The diesel vehicle emission prediction method based on deep migration learning of claim 4, wherein: the verifying the reliability of the reconstruction features of S26 includes respectively establishing a source domain prediction model using the source domain original features and the reconstruction features, selecting 100 samples as a test set, and performing, on the remaining samples, a test of 8: 2 into a training set and a validation set.
6. The diesel vehicle emission prediction method based on deep migration learning of claim 4, wherein: the S30: the pre-training exhaust gas prediction module builds a full-connection neural network prediction model with double hidden layers by using the reconstructed source domain data characteristics and labels, and specifically comprises the following steps:
s31: based on the step S24, using the retained output data of the third hidden layer as the input of a prediction model, post-processing a downstream NOx value as the output, constructing a double-hidden-layer fully-connected neural network, selecting Relu as an activation function, selecting Adam as an optimizer, setting the iteration number to 15000 times, and stopping iteration when the loss value is not reduced for more than 10 continuous times;
s32: and minimizing the Mean Square Error (MSE) of the real NOx value and the predicted value, and updating hidden layer weights W4, W5 and W6 in a back propagation mode until convergence.
7. The diesel vehicle emission prediction method based on deep migration learning of claim 6, wherein: s40: combining a pre-trained projection module and a prediction module to construct a migration model, considering the data distribution difference between a source domain vehicle and a target domain vehicle, adding KL divergence between the source domain and the target domain in a loss function part, enabling the target domain data distribution to be close to the source domain data distribution, and finely adjusting the whole model to obtain a stable diesel vehicle emission prediction model, which specifically comprises the following steps:
s41: based on the weight W of steps S20, S301',W2',W3',W4,W5,W6Multiplexing the weights, building a 6-layer migration neural network, inputting various characteristics of a source domain vehicle and a target domain vehicle, and predicting an emission value with a label of NOx;
s42: by adding KL divergence in the loss function and minimizing the loss function, the distribution difference of the source domain vehicle and the target domain vehicle in the subspace is reduced, the similarity of the source domain vehicle and the target domain vehicle is improved, and therefore knowledge transfer from the source domain vehicle to the target domain vehicle is achieved, and the loss function is as follows:
whereinNO for i-th sample of source domain vehiclexThe value of the one or more of the one,is the ith vehicle NO of the source area vehiclexBeta is an empirical parameter, DKLTo solve the KL divergence function, psFor the feature distribution of the source domain features in the subspace, ρtCharacteristic distribution of the target domain vehicle in the subspace;
s43: the whole migration prediction network is finely adjusted through gradient descent back propagation, and after the network is converged, the model is used for predicting NO of the target domain vehiclexEmission, NOxThe emission formula is as follows:
wherein f isreg_x/SAE_x(W,X)=W·X。
8. A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to carry out the steps of the method according to any one of claims 1 to 7.
9. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110852663.6A CN113657651A (en) | 2021-07-27 | 2021-07-27 | Diesel vehicle emission prediction method, medium and equipment based on deep migration learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110852663.6A CN113657651A (en) | 2021-07-27 | 2021-07-27 | Diesel vehicle emission prediction method, medium and equipment based on deep migration learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113657651A true CN113657651A (en) | 2021-11-16 |
Family
ID=78478810
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110852663.6A Pending CN113657651A (en) | 2021-07-27 | 2021-07-27 | Diesel vehicle emission prediction method, medium and equipment based on deep migration learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113657651A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114372941A (en) * | 2021-12-16 | 2022-04-19 | 佳源科技股份有限公司 | Low-illumination image enhancement method, device, equipment and medium |
CN116070796A (en) * | 2023-03-29 | 2023-05-05 | 中国科学技术大学 | Diesel vehicle emission level evaluation method and system |
CN116128049A (en) * | 2023-04-04 | 2023-05-16 | 厦门大学 | XGBoost model-based migration condition selection method for water quality prediction model |
CN114372941B (en) * | 2021-12-16 | 2024-04-26 | 佳源科技股份有限公司 | Low-light image enhancement method, device, equipment and medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105843734A (en) * | 2016-03-18 | 2016-08-10 | 厦门美图移动科技有限公司 | Automatic test method and device and computing device |
US20180157963A1 (en) * | 2016-12-02 | 2018-06-07 | Fleetmatics Ireland Limited | Vehicle classification using a recurrent neural network (rnn) |
CN108898218A (en) * | 2018-05-24 | 2018-11-27 | 阿里巴巴集团控股有限公司 | A kind of training method of neural network model, device and computer equipment |
CN109060001A (en) * | 2018-05-29 | 2018-12-21 | 浙江工业大学 | A kind of multiple operating modes process soft-measuring modeling method based on feature transfer learning |
CN110321926A (en) * | 2019-05-24 | 2019-10-11 | 北京理工大学 | A kind of moving method and system based on depth residual GM network |
CN110325929A (en) * | 2016-12-07 | 2019-10-11 | 阿瑞路资讯安全科技股份有限公司 | System and method for detecting the waveform analysis of cable network variation |
CN110378366A (en) * | 2019-06-04 | 2019-10-25 | 广东工业大学 | A kind of cross-domain image classification method based on coupling knowledge migration |
CN110826772A (en) * | 2019-10-15 | 2020-02-21 | 深圳市赛梅斯凯科技有限公司 | Vehicle exhaust emission prediction method, device, equipment and storage medium |
CN111612254A (en) * | 2020-05-22 | 2020-09-01 | 中国科学院合肥物质科学研究院 | Road motor vehicle exhaust emission prediction method based on improved attention bidirectional long-short term memory network |
US20200372345A1 (en) * | 2019-05-21 | 2020-11-26 | Alibaba Group Holding Limited | System and method for training neural networks |
US20210012198A1 (en) * | 2018-05-31 | 2021-01-14 | Huawei Technologies Co., Ltd. | Method for training deep neural network and apparatus |
-
2021
- 2021-07-27 CN CN202110852663.6A patent/CN113657651A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105843734A (en) * | 2016-03-18 | 2016-08-10 | 厦门美图移动科技有限公司 | Automatic test method and device and computing device |
US20180157963A1 (en) * | 2016-12-02 | 2018-06-07 | Fleetmatics Ireland Limited | Vehicle classification using a recurrent neural network (rnn) |
CN110325929A (en) * | 2016-12-07 | 2019-10-11 | 阿瑞路资讯安全科技股份有限公司 | System and method for detecting the waveform analysis of cable network variation |
CN108898218A (en) * | 2018-05-24 | 2018-11-27 | 阿里巴巴集团控股有限公司 | A kind of training method of neural network model, device and computer equipment |
CN109060001A (en) * | 2018-05-29 | 2018-12-21 | 浙江工业大学 | A kind of multiple operating modes process soft-measuring modeling method based on feature transfer learning |
US20210012198A1 (en) * | 2018-05-31 | 2021-01-14 | Huawei Technologies Co., Ltd. | Method for training deep neural network and apparatus |
US20200372345A1 (en) * | 2019-05-21 | 2020-11-26 | Alibaba Group Holding Limited | System and method for training neural networks |
CN110321926A (en) * | 2019-05-24 | 2019-10-11 | 北京理工大学 | A kind of moving method and system based on depth residual GM network |
CN110378366A (en) * | 2019-06-04 | 2019-10-25 | 广东工业大学 | A kind of cross-domain image classification method based on coupling knowledge migration |
CN110826772A (en) * | 2019-10-15 | 2020-02-21 | 深圳市赛梅斯凯科技有限公司 | Vehicle exhaust emission prediction method, device, equipment and storage medium |
CN111612254A (en) * | 2020-05-22 | 2020-09-01 | 中国科学院合肥物质科学研究院 | Road motor vehicle exhaust emission prediction method based on improved attention bidirectional long-short term memory network |
Non-Patent Citations (1)
Title |
---|
邱宁佳;王晓霞;王鹏;周思丞;王艳春;: "结合迁移学习模型的卷积神经网络算法研究", 计算机工程与应用, no. 05 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114372941A (en) * | 2021-12-16 | 2022-04-19 | 佳源科技股份有限公司 | Low-illumination image enhancement method, device, equipment and medium |
CN114372941B (en) * | 2021-12-16 | 2024-04-26 | 佳源科技股份有限公司 | Low-light image enhancement method, device, equipment and medium |
CN116070796A (en) * | 2023-03-29 | 2023-05-05 | 中国科学技术大学 | Diesel vehicle emission level evaluation method and system |
CN116070796B (en) * | 2023-03-29 | 2023-06-23 | 中国科学技术大学 | Diesel vehicle emission level evaluation method and system |
CN116128049A (en) * | 2023-04-04 | 2023-05-16 | 厦门大学 | XGBoost model-based migration condition selection method for water quality prediction model |
CN116128049B (en) * | 2023-04-04 | 2023-06-30 | 厦门大学 | XGBoost model-based migration condition selection method for water quality prediction model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022077693A1 (en) | Load prediction model training method and apparatus, storage medium, and device | |
US10635976B2 (en) | Machine learning system for estimating a temperature of an exhaust purification catalyst | |
CN113657651A (en) | Diesel vehicle emission prediction method, medium and equipment based on deep migration learning | |
Kong et al. | A novel torque distribution strategy based on deep recurrent neural network for parallel hybrid electric vehicle | |
Lin et al. | A driving-style-oriented adaptive control strategy based PSO-fuzzy expert algorithm for a plug-in hybrid electric vehicle | |
CN113505926B (en) | Fuel cell fault prediction method based on impedance prediction model self-updating | |
CN111931413A (en) | Method for predicting performance of high-altitude diesel engine of optimized extreme learning machine based on chaotic particle swarm with extreme disturbance | |
Asher et al. | Economic and efficient hybrid vehicle fuel economy and emissions modeling using an artificial neural network | |
Montazeri-Gh et al. | Driving condition recognition for genetic-fuzzy HEV control | |
Motallebiaraghi et al. | High-fidelity modeling of light-duty vehicle emission and fuel economy using deep neural networks | |
CN116298936A (en) | Intelligent lithium ion battery health state prediction method in incomplete voltage range | |
CN113715629B (en) | Residual driving range prediction method based on improved symbolic regression and XGBoost algorithm | |
CN116523177A (en) | Vehicle energy consumption prediction method and device integrating mechanism and deep learning model | |
Huang et al. | Correlation analysis and prediction of PEM fuel cell voltage during start-stop operation based on real-world driving data | |
Liao et al. | A comparative investigation of advanced machine learning methods for predicting transient emission characteristic of diesel engine | |
CN113379145B (en) | Vehicle emission remote online monitoring platform, system and vehicle-mounted terminal | |
Malikopoulos et al. | Optimal engine calibration for individual driving styles | |
CN116741304A (en) | Diesel engine original exhaust nitrogen oxide prediction method based on Internet of vehicles big data | |
Miranda et al. | Particle swarm optimization of Elman neural network applied to battery state of charge and state of health estimation | |
Xu et al. | A deep transfer NOx emission inversion model of diesel vehicles with multisource external influence | |
Witaszek | Modeling of fuel consumption using artificial neural networks | |
CN114282680A (en) | Vehicle exhaust emission prediction method and system based on machine learning algorithm | |
CN116628395B (en) | Urban terrain feature-based road vehicle carbon emission measuring and calculating method and system | |
JP7056794B1 (en) | Model learning system and model learning device | |
Liao et al. | Prediction of the transient emission characteristics from diesel engine using temporal convolutional networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |