CN112598170B - Vehicle exhaust emission prediction method and system based on multi-component fusion time network - Google Patents
Vehicle exhaust emission prediction method and system based on multi-component fusion time network Download PDFInfo
- Publication number
- CN112598170B CN112598170B CN202011507315.7A CN202011507315A CN112598170B CN 112598170 B CN112598170 B CN 112598170B CN 202011507315 A CN202011507315 A CN 202011507315A CN 112598170 B CN112598170 B CN 112598170B
- Authority
- CN
- China
- Prior art keywords
- exhaust emission
- time
- data
- input data
- external factor
- 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.)
- Active
Links
- 230000004927 fusion Effects 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims description 51
- 238000013527 convolutional neural network Methods 0.000 claims description 31
- 230000007246 mechanism Effects 0.000 claims description 13
- 210000002569 neuron Anatomy 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000000946 synaptic effect Effects 0.000 claims description 2
- 239000007789 gas Substances 0.000 description 20
- 230000006870 function Effects 0.000 description 11
- 241000728173 Sarima Species 0.000 description 7
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 6
- 230000000306 recurrent effect Effects 0.000 description 4
- 230000014509 gene expression Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000001932 seasonal effect Effects 0.000 description 3
- 230000002457 bidirectional effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000036962 time dependent Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 230000003442 weekly effect Effects 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- 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
- G06N3/045—Combinations of 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Human Resources & Organizations (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a vehicle exhaust emission prediction method and a vehicle exhaust emission prediction system based on a multi-component fusion time network, wherein the method comprises the following steps: step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the vehicle exhaust concentration data and the meteorological factor data, and respectively constructing time series input data of exhaust emission and external factor input data; step S2: inputting the time sequence input data of the tail gas emission and the external factor input data into a multi-component fusion time network for training to obtain an attention time characteristic result of the tail gas emission and an external factor characteristic result; and step S3: and fusing the attention time characteristic result and the external factor characteristic result of the exhaust emission to obtain an exhaust emission concentration prediction result. According to the invention, the exhaust emission concentration of the vehicle is predicted by combining the exhaust emission time series data and external factor data input data, so that the accuracy and precision are improved.
Description
Technical Field
The invention belongs to the field of vehicle tail gas emission concentration prediction and mode recognition, and particularly relates to a vehicle tail gas emission prediction method and system based on a multi-component fusion time network.
Background
Along with the rapid increase of the exhaust emission of motor vehicles, the natural and environmental problems caused by the emission of urban exhaust gas are increasingly serious, and a significant social problem is brought. Greenhouse gases (such as CO) generated by vehicle exhaust emission can cause certain harm to human health.
The prediction of the exhaust emission concentration of the urban vehicle can be regarded as a time series prediction problem. There are several conventional linear models that can solve this problem. The historical average model may use the average of the historical time series to predict future values of the time series. However, the historical average model does not reflect the time dependence. ARIMA [1 ]]([1]Contreras,J.,Espinola,R.,Nogales,F.J.,&The Conejo, A.J. (2003). ARIMA models to predict next-day electric properties systems,18 (3), 1014-1020.) assume that in predicting a future time series, the future value of the time series is a linear combination of the previous value and the residual. The ARIMA model is built on a stationary time series basis. Therefore, to obtain stationarity, non-stationary time series need to be differentially analyzed before analysis [2]([2]Box,G.E.,Jenkins,G.M.,Reinsel,G.C.,&Ljung,G.M.(2015).Time series analysis:forecasting and control.John Wiley&Sons.), but does not take into account external factors, resulting in a low prediction accuracy. A method for obtaining sequence stationarity involving seasonal characteristics using additional seasonal differences is called SARIMA [3 ]]([3]Martinez,Edson Zangiacomi,
Aparecida Soares da Silva,and Amaury Lelis Dal Fabbro."A SARIMA forecasting model to predict the number of cases ofdengue in Campinas,State of
Paulo,Brazil."Revista da Sociedade Brasileira de Medicinal Tropical 44.4 (2011): 436-440.). However, SARIMA suffers from the disadvantage of being time consuming and not suitable for real-time online prediction of exhaust gas concentration [4]([4]Smith, brian L., billy M.Williams, and R.Keith Oswald. "comprehensive of parameter and nonparametric models for streaming flow for evaluation." Transportation Research Part C: operating Technologies 10.4 (2002): 303-321.). In addition to the linear models described above, artificial Neural Network (ANN) and multi-layer perceptron (MLP) models are often used as non-linear models for time series prediction [5 ]]-[6]([5]Florio,L.I.V.I.O.,&Mussone,L.O.R.E.N.Z.O.(1996).Neural-network models for classification and forecasting of freeway traffic flow stability.Control Engineering Practice,4(2),153-164.，[6]Dougherty,M.S.,&Cobbett, m.r. (1997). Short-term inter-url traffic using neural networks. International journel of learning, 13 (1), 21-31.). Artificial neural networks have good nonlinear fitting capabilities, but do not capture time dependencies well [7]([7]Zhang,G.P.,&Qi,M.(2005).Neural network forecasting for seasonal and trend time series.Europeanjournal ofoperational research,160(2),501-514.)。
Therefore, the existing exhaust emission concentration prediction defect and the defect of low accuracy of the artificial intelligence technology for predicting the exhaust emission concentration need a new exhaust emission concentration prediction method to further improve the accuracy and precision of the exhaust emission concentration prediction.
Disclosure of Invention
In order to solve the technical problems, the invention provides a vehicle exhaust emission prediction method and system based on a multi-component fusion time network. Because the exhaust emission concentration data has certain time dependency, the influence of the exhaust emission data of a far time period and a near time period within a target prediction time period on the prediction concentration is introduced, and meanwhile, external meteorological factors are considered, so that the method can obtain better prediction accuracy. The method provided by the invention can obtain the vehicle exhaust emission concentration information in a certain period of time in the future, has important significance for supporting urban traffic pollution treatment and environmental protection, and can send out regional pollution early warning by predicting the exhaust emission concentration conditions of some remote monitoring stations in a city in a period of time in the future, thereby helping to improve the design of urban traffic infrastructure and the control of urban vehicle exhaust emission.
The technical solution of the invention is as follows: a vehicle exhaust emission prediction method based on a multi-component fusion time network comprises the following steps:
step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the vehicle exhaust concentration data and the meteorological factor data, and respectively constructing time series input data of exhaust emission and external factor input data;
step S2: inputting the time sequence input data of the tail gas emission and the external factor input data into a multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, the time series input data of the exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of the exhaust emission, and the external factor input data pass through the artificial neural network to obtain an external factor characteristic result;
and step S3: and fusing the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result.
Compared with the prior art, the invention has the following advantages:
1. the vehicle exhaust emission prediction method of the multi-component fusion time network is based on exhaust emission time sequence data and external factor data in a certain time period. Simultaneously selecting a plurality of time interval components (recent period, daily period and weekly period) on the time axis of the tail gas time sequence data to construct time sequence input data of tail gas emission; meanwhile, the influence of external factor input data of meteorological factors on the exhaust emission concentration is considered. The vehicle exhaust emission concentration is predicted by combining two input data, and the accuracy is improved.
2. The vehicle exhaust emission prediction method of the multi-component fusion time network can predict the vehicle exhaust emission concentration in a certain period of time in the future. The time dependency of the time-series input data of the exhaust emission constructed by a plurality of time components is firstly captured by using a plurality of one-dimensional convolutional neural networks, and the accuracy of the prediction of the exhaust emission concentration of the model is improved by using the attention mechanism to give the captured time characteristic weight.
Drawings
FIG. 1 is a flow chart of a method for predicting vehicle exhaust emissions based on a multi-component fusion time network according to an embodiment of the present invention;
FIG. 2 shows a step S1 of a vehicle exhaust emission prediction method based on a multi-component fusion time network in the embodiment of the present invention: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the data, and respectively constructing a flow chart of time series input data of exhaust emission and external factor input data;
FIG. 3 shows a step S11 of a vehicle exhaust emission prediction method based on a multi-component fusion time network in an embodiment of the present invention: collecting vehicle exhaust concentration data of different periods, and constructing a flow chart of time series input data of exhaust emission;
fig. 4 shows a method for predicting vehicle exhaust emission based on a multi-component fusion time network in the embodiment of the present invention, S2: inputting time sequence input data of tail gas emission and external factor input data into a multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, time series input data of exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of the exhaust emission, and external factor input data pass through the artificial neural network to obtain a flow chart of the external factor characteristic result;
fig. 5 shows a step S21 in the vehicle exhaust emission prediction method based on the multi-component fusion time network in the embodiment of the present invention: inputting the time sequence of the exhaust emission into data, inputting the data into a convolutional neural network for training, and adding an attention characteristic mechanism to obtain a flow chart of an attention time characteristic result of the exhaust emission;
FIG. 6 is a block diagram of a multi-component fusion time network in an embodiment of the present invention, where CNN is a convolutional neural network and ANN is an artificial neural network;
fig. 7 is a block diagram of a vehicle exhaust emission prediction system based on a multi-component fusion time network according to an embodiment of the present invention.
Detailed Description
The invention provides a vehicle exhaust emission prediction method based on a multi-component fusion time network, which predicts the vehicle exhaust emission concentration by combining exhaust emission time sequence data and external factor data input data, improves the accuracy, captures the time dependence of the exhaust emission time sequence input data constructed by a plurality of one-dimensional convolution neural networks on a plurality of time components, and improves the prediction accuracy of the exhaust emission concentration by using an attention mechanism to endow the capture arrival time with characteristic weight.
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings.
Example one
As shown in fig. 1, a vehicle exhaust emission prediction method based on a multi-component fusion time network according to an embodiment of the present invention includes the following steps:
step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the vehicle exhaust concentration data and the meteorological factor data, and respectively constructing time series input data of exhaust emission and external factor input data;
step S2: inputting time sequence input data of tail gas emission and external factor input data into a multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, the time series input data of the exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of the exhaust emission, and the external factor input data pass through the artificial neural network to obtain an external factor characteristic result;
and step S3: and fusing the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result.
The vehicle exhaust emission prediction method of the multi-component fusion time network provided by the invention is based on exhaust emission time sequence data and external factor data in a certain time period. The tail gas time sequence input data is used, and meanwhile the influence of external factor input data of meteorological factors on the tail gas emission concentration is considered. The vehicle exhaust emission concentration is predicted by combining the two input data, so that the accuracy is improved.
As shown in fig. 2, in one embodiment, the step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the data, and respectively constructing time series input data and external factor input data of exhaust emission, wherein the data comprises the following steps:
step S11: collecting vehicle exhaust concentration data of different periods, and constructing time series input data of exhaust emission;
step S12: and collecting various meteorological factor data and constructing external factor input data.
As shown in fig. 3, in one embodiment, the step S11: collecting vehicle exhaust concentration data of different periods, and constructing time series input data of exhaust emission, wherein the time series input data comprises the following steps:
step S111: recent time interval assembly for collecting and constructing vehicle exhaust concentration data
Step S112: daily cycle time interval assembly for collecting and constructing vehicle exhaust concentration data
Step S113: week cycle time interval assembly for collecting and constructing vehicle exhaust concentration data
The emission concentration of the vehicle exhaust concentration data collected by the remote exhaust gas monitoring system at the moment t is X t (ii) a Vehicle exhaust gas concentration data according to T p Accumulating within hours, in the embodiment of the invention, T is selected p =0.5 hour, so the number of vehicle exhaust emission concentration data per day is
A plurality of; l. the c 、l p And l q Respectively, the lengths of the sequence segments of the recent period, the daily period and the week period are respectively intercepted along a time axis and are T p Integer multiples of; for example, assuming that the target prediction interval is one frame in a time interval of friday (e.g., 8 c =3 denotes 3 frames of the three time interval before the time interval; time interval length of day cycle l p =3 represents 3 frames at the same time interval on thursday, wednesday, tuesday; periodic time interval length l q =3 represents 3 frames for the same time interval every friday in the past three weeks; p and q are 1 day and 1 week, respectively.
Step S114: constructing a time series of exhaust emissions with input data { X ] according to the 3 sets of time interval components c ,X p ,X q }。
In one embodiment, the step S12: collecting various meteorological factor data, and constructing external factor input data, wherein the external factor input data comprises the following steps:
external factor input data E collected by exhaust remote sensing monitoring system t The method comprises 5 numerical attribute fields of wind speed, wind direction, outdoor temperature, relative humidity and atmospheric pressure, and uses a normalization method to process and adjust the data in numerical scale, wherein the numerical range is adjusted to be [ -1,1 [ -1]。
As shown in fig. 4, in one embodiment, the step S2: inputting time sequence input data of tail gas emission and external factor input data into a multi-component fusion time network for training; wherein, multicomponent fuses time network and includes convolution neural network and artificial neural network, and exhaust emissions's time series input data passes through convolution neural network, obtains exhaust emissions's attention time characteristic result, and external factor input data passes through artificial neural network, obtains external factor characteristic result, includes:
step S21: inputting the time sequence of the exhaust emission into data, inputting the data into a convolutional neural network for training, and adding an attention characteristic mechanism to obtain an attention time characteristic result of the exhaust emission;
step S22: and inputting the external factor into data, and inputting the data into an artificial neural network for training to obtain an external factor characteristic result.
The multi-component fusion time network disclosed by the invention consists of two parts, namely a convolutional neural network and an artificial neural network, which are respectively used for different input data. The time series input data of the exhaust emission are input into the convolutional neural network, and the attention time characteristic result of the exhaust emission can be obtained. The external factors are input into data and input into an artificial neural network for training, and external factor characteristic results can be obtained. The vehicle exhaust emission concentration is predicted by combining two input data, and the accuracy is improved.
As shown in fig. 5, in one embodiment, the step S21: inputting the time sequence of the exhaust emission into data, inputting the data into a convolutional neural network for training, and adding an attention characteristic mechanism to obtain an attention time characteristic result of the exhaust emission, wherein the attention time characteristic result comprises the following steps:
step S211: inputting the time sequence of exhaust emission into data, calculating by at least one-dimensional convolution neural network according to the following formula (1), and outputting a time characteristic result X f ；
Wherein, the convolution operator is represented by,
is the feature vector, M, corresponding to the jth convolution kernel of the l-th layer j Is the receptive field of the neurons of the jth convolution kernel,
and
respectively representWeight matrix and bias coefficient of jth convolution kernel of ith layer, f (l) (. Cndot.) represents the activation function of the l layer.
As shown in fig. 6, in this step, in order to capture the time-dependent features in the time-series data of exhaust emissions from various angles, the embodiment of the present invention uses 3 one-dimensional convolutional neural networks, each convolutional neural network is composed of an input layer, 3 convolutional layers, and a fully-connected output layer, and the layers are normalized in batch and Dropout to reduce the overfitting during the training process, where the input layer is the time-series input data { X } of exhaust emissions mentioned above c ,X p ,X q And (4) taking the output of the previous layer as the input of the next layer and the output of the next layer as the input of the other layer by the convolution layer, and so on until the time characteristic result X from the output of the output layer to the exhaust emission f 。
Through a plurality of experiments, the invention adopts 3 one-dimensional 3-layer convolutional neural networks, and can obtain better results. The invention does not specifically limit the convolutional neural network, and can determine which structure of convolutional neural network to use according to actual requirements.
Step S212: time feature result X f Performing cascade operation according to the following formula (2) and performing head-to-tail splicing;
as shown in fig. 6, in the embodiment of the present invention, after the results of 3 convolution operations are obtained, they are concatenated end to end using a concatenation operation. Wherein,
representing a set of real numbers, f i Representing the ith time signature result
The number of the characteristics of (a) is,
represents the convolution using the nthThe number of features obtained by the neural network is f n The characteristic result of (c), concat (. Cndot.) represents a cascading operation.
Step S213: calculating a weight coefficient alpha of the attention mechanism to the time characteristic result through the following formula (3) j And calculating the attention time characteristic result X Att ＝α j ·X j ；
Wherein, X j Represents X Cas Is the jth value of (1), r is X j The number of features of (2).
In this step, since there is an importance difference between the time-dependent features of the plurality of angles, the attention time feature result X is finally obtained by giving a weight coefficient to the time feature result using an attention mechanism Att 。
The vehicle exhaust emission prediction method of the multi-component fusion time network can predict the vehicle exhaust emission concentration in a certain period of time in the future. The time dependency of the time-series input data of the exhaust emission constructed by a plurality of time components is firstly captured by using a plurality of one-dimensional convolutional neural networks, and the accuracy of the prediction of the exhaust emission concentration of the model is improved by using the attention mechanism to give the captured time characteristic weight.
In one embodiment, the step S22: inputting external factor input data into the artificial neural network for training to obtain the external factor characteristic result, wherein the external factor characteristic result comprises the following steps:
inputting external factors into data, and calculating to obtain an external feature result X through an artificial neural network shown in the following formula (4) Ext ；
Wherein f is an activation function, w i Is the ith synaptic weight, x, of the neuron i Represents the outsideThe i-th component of the input data, b, is the bias parameter for the neuron.
As shown in fig. 5, the artificial neural network structure in the embodiment of the present invention is composed of an input layer, 3 hidden layers, and a fully connected output layer, wherein the layers use batch normalization and Dropout operations to reduce overfitting occurring in the training process, each layer is composed of a plurality of neurons, and the input data of the input layer is the external factor input data E t The hidden layer takes the output of the previous layer as the input of the next layer, takes the output of the next layer as the input of the other layer, and so on until the output of the fully-connected output layer obtains the external factor characteristic result X Ext 。
The activation functions in the convolutional neural network and the artificial neural network in the above steps are both Relu functions as the activation functions, and the mathematical expression of the activation functions is as the following formula (5):
wherein x i Is the ith input of the neuron, w i Is the weight value of the ith input of the neuron, L is the total input number of the neuron, theta is a bias coefficient, and f is the output function of the neuron.
In one embodiment, the step S3: fusing the attention time characteristic result and the external factor characteristic result of the exhaust emission to obtain an exhaust emission concentration prediction result, wherein the prediction result comprises the following steps:
attention time feature result X after using attention mechanism Att And external factor characteristic result X Ext Obtaining the exhaust emission concentration prediction result X in a certain period by using summation average, as the following formula (6) Avg 。
The value range of the treated exhaust emission data is [ -1,1]Exhaust emission concentration prediction result X Avg Selecting tThe anh function is taken as an activation function, namely the result of the prediction of the exhaust emission concentration in a certain period during training is
Obtaining the exhaust emission concentration prediction result
Then, using the Huber function as a loss function of the multi-component fusion time network, the mathematical expression of which is as the following formula (7):
wherein, x and
the target predicted exhaust emission concentration observed value and the predicted value within a certain time period are delta, the prediction deviation is delta, and the default is 1.
In the embodiment of the invention, RMSProp is selected as an optimization algorithm and a back propagation algorithm for training.
After the multi-component fusion time network model is trained, the exhaust concentration data of the vehicle and the meteorological factor data are combined, so that the exhaust concentration in a certain time period can be accurately predicted. Selecting a root mean square error RMSE and an average absolute error MAE as statistical analysis performance indexes of the vehicle exhaust emission prediction method of the multi-component fusion time network. Here, the smaller the values of RMSE and MAE, the better the estimation performance of the model. The mathematical expressions of RMSE and MAPE are respectively shown in the following formulas (8) to (9):
wherein Z is the total number of samples of the exhaust emission concentration,
and
respectively predicting the concentration of the target tail gas of the sample and observing the concentration of the target tail gas of the sample.
The exhaust gas concentrations over a period of time were predicted using the multi-component fusion time network model (MCFT-Net), the averaging model (HA), other single machine learning models (including ARIMA, SARIMA) and neural network models (including Artificial Neural Network (ANN), recurrent Neural Network (RNN), long-short term memory artificial neural network (LSTM), gated Recurrent Unit (GRU) and bidirectional long-short term memory artificial neural network (BiLSTM)) of the present invention, respectively, with associated statistical performance pairs such as those shown in table 1.
TABLE 1 comparison of statistical properties of exhaust gas concentrations for MCFT-Net, HA, ARIMA, SARIMA, ANN, LSTM, GRU, and BilSTM model predictive stochastic test vehicle samples
| Model (model) | RMSE | MAE |
| HA | 0.2757 | 0.2023 |
| ARIMA | 0.1950 | 0.1403 |
| SARIMA | 0.1890 | 0.1360 |
| ANN | 0.1651±0.0031 | 0.1245±0.0032 |
| RNN | 0.1606±0.0014 | 0.1228±0.0015 |
| LSTM | 0.1599±0.0026 | 0.1216±0.0024 |
| GRU | 0.1600±0.0031 | 0.1218±0.0028 |
| BiLSTM | 0.1536±0.0120 | 0.1178±0.0300 |
| MCFT-Net | 0.1525±0.0019 | 0.1155±0.0016 |
As can be seen from table 1, compared with the MCFT-Net model and the average model (HA), and other single machine learning models (including ARIMA, SARIMA) and neural network models (including Artificial Neural Network (ANN), recurrent Neural Network (RNN), long-short term memory artificial neural network (LSTM), gated Recurrent Unit (GRU), and bidirectional long-short term memory artificial neural network (BiLSTM)) disclosed in the present invention, the root mean square error RMSE and the average absolute error MAE obtained by the exhaust gas concentration prediction method based on the MCFT-Net model are smaller, which indicates that the exhaust gas concentration prediction performance of the MCFT-Net model of the present invention is better than that of other prediction models.
Example two
As shown in fig. 7, an embodiment of the present invention provides a vehicle exhaust emission prediction system based on a multi-component fusion time network, including the following modules:
the construction training data module 31 is used for collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the data and respectively constructing time sequence input data of exhaust emission and external factor input data;
the multi-component fusion time network training module 32 is used for inputting the time sequence input data of the exhaust emission and the external factor input data into the multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, the time series input data of the exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of the exhaust emission, and the external factor input data pass through the artificial neural network to obtain an external factor characteristic result;
and the fusion module 33 is configured to fuse the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result.
The above examples are provided for the purpose of describing the present invention only and are not intended to limit the scope of the present invention. The scope of the invention is defined by the appended claims. Various equivalent substitutions and modifications can be made without departing from the spirit and principles of the invention, and are intended to be within the scope of the invention.
Claims (6)
1. A vehicle exhaust emission prediction method based on a multi-component fusion time network is characterized by comprising the following steps:
step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the vehicle exhaust concentration data and the meteorological factor data, and respectively constructing time series input data of exhaust emission and external factor input data;
step S2: inputting the time sequence input data of the tail gas emission and the external factor input data into a multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, the time series input data of exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of exhaust emission, and the external factor input data pass through the artificial neural network to obtain an external factor characteristic result, and the method specifically comprises the following steps:
step S21: inputting the time sequence input data of the exhaust emission into the convolutional neural network for training, and adding an attention characteristic mechanism to obtain an attention time characteristic result of the exhaust emission, wherein the method specifically comprises the following steps:
step S211: inputting the time sequence of the exhaust emission into data, calculating through at least one-dimensional convolution neural network of at least one layer according to the following formula (1), and outputting a time characteristic result X f ；
Wherein, the convolution operator is represented by,
is the feature vector, M, corresponding to the jth convolution kernel of the ith layer j Is the receptive field of the neurons of the jth convolution kernel,
and
weight matrix and bias coefficient respectively representing the jth convolution kernel of the ith layer, f (l) (. Cndot.) represents the activation function of the l layer;
step S212: the time characteristic result X is obtained f Performing cascade operation according to the following formula (2) and performing head-to-tail splicing;
wherein,
representing a set of real numbers, f i Representing the ith time characteristic result
The number of the characteristics of (a) is,
f represents the number of features obtained by using the n-th convolutional neural network n The characteristic result of (c), concat () represents a cascading operation;
step S213: calculating a weight coefficient alpha of the attention mechanism to the time characteristic result by the following formula (3) j And calculating to obtain attention time characteristic result X Att ＝α j ·X j ；
Wherein X j Represents X Cas Is the jth value of (1), r is X j The number of features of (c);
step S22: inputting the external factor input data into the artificial neural network for training to obtain the external factor characteristic result;
and step S3: and fusing the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result.
2. The method for predicting vehicle exhaust emission based on the multi-component fusion time network according to claim 1, wherein the step S1: collecting vehicle exhaust concentration data and meteorological factor data, preprocessing the data, and respectively constructing time series input data and external factor input data of exhaust emission, wherein the data comprises the following steps:
step S11: collecting vehicle exhaust concentration data of different periods, and constructing time series input data of exhaust emission;
step S12: and collecting various meteorological factor data and constructing external factor input data.
3. The method for predicting vehicle exhaust emission based on the multi-component fusion time network according to claim 2, wherein the step S11: collecting vehicle exhaust concentration data of different periods, and constructing time series input data of exhaust emission, wherein the time series input data comprises the following steps:
step S111: recent time interval component X for collecting and constructing vehicle exhaust concentration data c ；
Step S112: daily cycle time interval component X for collecting and constructing vehicle exhaust concentration data p ；
Step S113: cycle time interval component X for collecting and constructing vehicle exhaust concentration data q ；
Step S114: constructing a time series of exhaust emissions with input data { X ] according to the 3 sets of time interval components c ,X p ,X q }。
4. The method for predicting vehicle exhaust emission based on multi-component fusion time network according to claim 1, wherein the step S22: inputting the external factor input data into the artificial neural network for training to obtain the external factor characteristic result, wherein the external factor characteristic result comprises the following steps:
inputting the external factors into data, and calculating to obtain an external feature result X through an artificial neural network shown in the following formula (4) Ext ；
Wherein f is an activation function, w i Is the ith synaptic weight, x, of the neuron i Representing the ith component of the input signal, b is the bias parameter for the neuron.
5. The method for predicting vehicle exhaust emission based on the multicomponent fused time network according to claim 4, wherein the step S3: fusing the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result, wherein the prediction result comprises the following steps:
calculating to obtain the exhaust emission concentration prediction result X through the following formula (5) Avg ：
6. A vehicle exhaust emission prediction system based on a multi-component fusion time network is characterized by comprising the following modules:
the system comprises a training data construction module, a weather factor data acquisition module, a data processing module and a data processing module, wherein the training data construction module is used for acquiring vehicle exhaust concentration data and weather factor data, preprocessing the data and respectively constructing time sequence input data and external factor input data of exhaust emission;
the multi-component fusion time network training module is used for inputting the time sequence input data of the exhaust emission and the external factor input data into a multi-component fusion time network for training; the multi-component fusion time network comprises a convolutional neural network and an artificial neural network, the time series input data of exhaust emission pass through the convolutional neural network to obtain an attention time characteristic result of exhaust emission, and the external factor input data pass through the artificial neural network to obtain an external factor characteristic result, and the method specifically comprises the following steps:
step S21: inputting the time sequence input data of the exhaust emission into the convolutional neural network for training, and adding an attention characteristic mechanism to obtain an attention time characteristic result of the exhaust emission, wherein the method specifically comprises the following steps:
step S211: inputting the time sequence of the exhaust emission into data, calculating through at least one-dimensional convolution neural network of at least one layer according to the following formula (1), and outputting a time characteristic result X f ；
Wherein, denotes a convolution operator,
is the feature vector, M, corresponding to the jth convolution kernel of the ith layer j Is the receptive field of the neurons of the jth convolution kernel,
and
weight matrix and bias coefficient respectively representing the jth convolution kernel of the ith layer, f (l) (. Cndot.) represents the activation function of the l layer;
step S212: the time characteristic result X is obtained f Performing cascade operation according to the following formula (2) and performing head-to-tail splicing;
wherein,
representing a set of real numbers, f i Representing the ith time signature result
The number of the characteristics of (a) is,
the number of features obtained by using the n-th convolutional neural network is represented by f n Concat (-) represents a cascade operation;
step S213: calculating a weight coefficient alpha of the attention mechanism to the time characteristic result by the following formula (3) j And calculating the attention time characteristic result X Att ＝α j ·X j ；
Wherein, X j Represents X Cas Is the jth value of (1), r is X j The number of features of (a);
step S22: inputting the external factor input data into the artificial neural network for training to obtain the external factor characteristic result;
and the fusion module is used for fusing the attention time characteristic result of the exhaust emission and the external factor characteristic result to obtain an exhaust emission concentration prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011507315.7A CN112598170B (en) | 2020-12-18 | 2020-12-18 | Vehicle exhaust emission prediction method and system based on multi-component fusion time network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011507315.7A CN112598170B (en) | 2020-12-18 | 2020-12-18 | Vehicle exhaust emission prediction method and system based on multi-component fusion time network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112598170A CN112598170A (en) | 2021-04-02 |
| CN112598170B true CN112598170B (en) | 2022-10-28 |
Family
ID=75199346
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011507315.7A Active CN112598170B (en) | 2020-12-18 | 2020-12-18 | Vehicle exhaust emission prediction method and system based on multi-component fusion time network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112598170B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113222236A (en) * | 2021-04-30 | 2021-08-06 | 中国科学技术大学先进技术研究院 | Data distribution self-adaptive cross-regional exhaust emission prediction method and system |
| CN113469331B (en) * | 2021-06-28 | 2024-02-13 | 中国科学技术大学 | Vehicle tail gas prediction method and system based on global and local space-time diagram convolution |
| CN113887373B (en) * | 2021-09-27 | 2022-12-16 | 中关村科学城城市大脑股份有限公司 | Attitude identification method and system based on urban intelligent sports parallel fusion network |
| CN114692993B (en) * | 2022-04-18 | 2024-06-04 | 西南科技大学 | Water conservancy facility deformation prediction method integrating ARIMA and BiLSTM |
| CN118032062B (en) * | 2024-04-11 | 2024-08-02 | 克拉玛依市城投油砂矿勘探有限责任公司 | SAGD downhole temperature and pressure monitoring system and method based on artificial intelligence |
| CN118194140B (en) * | 2024-05-17 | 2024-09-10 | 吉林交通职业技术学院 | Intelligent diagnosis platform and method for vehicle exhaust emission faults |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107577910A (en) * | 2017-08-24 | 2018-01-12 | 中国科学技术大学 | A kind of Vehicular exhaust retrieving concentration method based on profound neutral net |
| CN108205889A (en) * | 2017-12-29 | 2018-06-26 | 长春理工大学 | Freeway traffic flow Forecasting Methodology based on convolutional neural networks |
| CN108288109A (en) * | 2018-01-11 | 2018-07-17 | 安徽优思天成智能科技有限公司 | Motor-vehicle tail-gas concentration prediction method based on LSTM depth space-time residual error networks |
| CN109214570A (en) * | 2018-09-03 | 2019-01-15 | 安徽优思天成智能科技有限公司 | A kind of urban area tail gas pollution prediction technique based on depth space-time correction model |
| CN109802430A (en) * | 2018-12-29 | 2019-05-24 | 上海电力学院 | A kind of wind-powered electricity generation power grid control method based on LSTM-Attention network |
| CN110168573A (en) * | 2016-11-18 | 2019-08-23 | 易享信息技术有限公司 | Spatial attention model for image labeling |
| CN110503139A (en) * | 2019-08-07 | 2019-11-26 | 中国科学技术大学 | A kind of urban area tail gas pollution prediction technique |
| CN110807365A (en) * | 2019-09-29 | 2020-02-18 | 浙江大学 | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network |
| 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 |
| CN111695731A (en) * | 2020-06-09 | 2020-09-22 | 中国电力科学研究院有限公司 | Load prediction method, system and equipment based on multi-source data and hybrid neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
-
2020
- 2020-12-18 CN CN202011507315.7A patent/CN112598170B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110168573A (en) * | 2016-11-18 | 2019-08-23 | 易享信息技术有限公司 | Spatial attention model for image labeling |
| CN107577910A (en) * | 2017-08-24 | 2018-01-12 | 中国科学技术大学 | A kind of Vehicular exhaust retrieving concentration method based on profound neutral net |
| CN108205889A (en) * | 2017-12-29 | 2018-06-26 | 长春理工大学 | Freeway traffic flow Forecasting Methodology based on convolutional neural networks |
| CN108288109A (en) * | 2018-01-11 | 2018-07-17 | 安徽优思天成智能科技有限公司 | Motor-vehicle tail-gas concentration prediction method based on LSTM depth space-time residual error networks |
| CN109214570A (en) * | 2018-09-03 | 2019-01-15 | 安徽优思天成智能科技有限公司 | A kind of urban area tail gas pollution prediction technique based on depth space-time correction model |
| CN109802430A (en) * | 2018-12-29 | 2019-05-24 | 上海电力学院 | A kind of wind-powered electricity generation power grid control method based on LSTM-Attention network |
| CN110503139A (en) * | 2019-08-07 | 2019-11-26 | 中国科学技术大学 | A kind of urban area tail gas pollution prediction technique |
| CN110807365A (en) * | 2019-09-29 | 2020-02-18 | 浙江大学 | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network |
| 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 |
| CN111695731A (en) * | 2020-06-09 | 2020-09-22 | 中国电力科学研究院有限公司 | Load prediction method, system and equipment based on multi-source data and hybrid neural network |
Non-Patent Citations (6)
| Title |
|---|
| On the Use of a Convolutional Neural Network to Predict Perceptual Quality of Images without Reference for Different Viewing Distances;Aladine Chetouani;《IEEE Xplore》;20190826;全文 * |
| 基于BP神经网络的汽油机尾气排放预测;左付山;《江苏大学学报(自然科学版)》;20200531;第41卷(第3期);全文 * |
| 基于注意力机制的显著性目标检测方法;王凯诚;《 智能系统学报》;20200930;第15卷(第5期);全文 * |
| 基于深度学习的城市大气PM2.5浓度预测研究;赵鹏飞等;《现代计算机》;20200425(第12期);全文 * |
| 基于视觉特征的二维图像文本描述生成关键算法研究;管志斌;《中国博士学位论文全文数据库 (信息科技辑)》;20200915;第27-67页 * |
| 试题知识点预测:一种教研知识强化的卷积神经网络模型;胡国平等;《中文信息学报》;20180531;第32卷(第5期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112598170A (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112598170B (en) | Vehicle exhaust emission prediction method and system based on multi-component fusion time network | |
| CN109492830B (en) | Mobile pollution source emission concentration prediction method based on time-space deep learning | |
| Zhang et al. | Dual-aspect self-attention based on transformer for remaining useful life prediction | |
| CN107045788B (en) | Traffic road condition prediction method and device | |
| CN111832222B (en) | Pollutant concentration prediction model training method, pollutant concentration prediction method and pollutant concentration prediction device | |
| Zheng et al. | Investigation of model ensemble for fine-grained air quality prediction | |
| CN114707772B (en) | Power load prediction method and system based on multi-feature decomposition and fusion | |
| Narwariya et al. | Graph neural networks for leveraging industrial equipment structure: An application to remaining useful life estimation | |
| CN116092294A (en) | Expressway traffic flow prediction method based on Transformer and graph meaning network | |
| CN115796434A (en) | Management and control method and device for power distribution network, electronic equipment and storage medium | |
| CN114120637A (en) | Intelligent high-speed traffic flow prediction method based on continuous monitor | |
| Al_Janabi et al. | Pragmatic method based on intelligent big data analytics to prediction air pollution | |
| CN114662389B (en) | Air pollutant-oriented self-correlation error Informer model long time sequence prediction method and system | |
| CN114062812A (en) | Fault diagnosis method and system for metering cabinet | |
| CN116013426A (en) | Site ozone concentration prediction method with high space-time resolution | |
| CN114648095A (en) | Air quality concentration inversion method based on deep learning | |
| Latyshev | Sensor Data Preprocessing, Feature Engineering and Equipment Remaining Lifetime Forecasting for Predictive Maintenance. | |
| CN114239990A (en) | Time series data prediction method based on time series decomposition and LSTM | |
| Huang et al. | A decomposition‐based multi‐time dimension long short‐term memory model for short‐term electric load forecasting | |
| CN117474529A (en) | Intelligent operation and maintenance system for power grid | |
| CN117391257A (en) | Road congestion condition prediction method and device | |
| CN117408833A (en) | Short-term load prediction method based on load identification | |
| CN115730456A (en) | Motor vehicle multielement tail gas prediction method and system based on double attention fusion network | |
| Tong et al. | A prediction model for complex equipment remaining useful life using gated recurrent unit complex networks | |
| CN116224950A (en) | Intelligent fault diagnosis method and system for self-organizing reconstruction of unmanned production line |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |