CN109492830A - A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning - Google Patents
A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning Download PDFInfo
- Publication number
- CN109492830A CN109492830A CN201811541300.5A CN201811541300A CN109492830A CN 109492830 A CN109492830 A CN 109492830A CN 201811541300 A CN201811541300 A CN 201811541300A CN 109492830 A CN109492830 A CN 109492830A
- Authority
- CN
- China
- Prior art keywords
- time
- space
- feature
- model
- concentration
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000013135 deep learning Methods 0.000 title claims abstract description 9
- 239000013598 vector Substances 0.000 claims abstract description 22
- 230000006870 function Effects 0.000 claims abstract description 8
- 230000007246 mechanism Effects 0.000 claims abstract description 5
- 238000012544 monitoring process Methods 0.000 claims description 28
- 239000000809 air pollutant Substances 0.000 claims description 17
- 231100001243 air pollutant Toxicity 0.000 claims description 17
- 230000002123 temporal effect Effects 0.000 claims description 14
- 238000013507 mapping Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 8
- 238000003915 air pollution Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 6
- 238000009792 diffusion process Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 238000007689 inspection Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 239000000356 contaminant Substances 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 2
- 239000003344 environmental pollutant Substances 0.000 abstract description 17
- 231100000719 pollutant Toxicity 0.000 abstract description 17
- 238000013527 convolutional neural network Methods 0.000 abstract description 4
- 239000000284 extract Substances 0.000 abstract description 3
- 230000007774 longterm Effects 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 abstract 1
- 230000007787 long-term memory Effects 0.000 abstract 1
- 238000000714 time series forecasting Methods 0.000 abstract 1
- 235000004257 Cordia myxa Nutrition 0.000 description 4
- 244000157795 Cordia myxa Species 0.000 description 4
- 239000012855 volatile organic compound Substances 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 3
- 230000004927 fusion Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000013618 particulate matter Substances 0.000 description 2
- 206010024769 Local reaction Diseases 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000005416 organic matter Substances 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G06N3/045—Combinations of networks
-
- 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
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Educational Administration (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The mobile pollution source concentration of emission prediction technique based on deep learning that the invention discloses a kind of, the present invention propose the convolution shot and long term Memory Neural Networks prediction technique based on attention mechanism according to the region temporal-spatial distribution characteristic of mobile pollution source pollutant.Firstly, by the Granger causality between analysis website and developing hyper parameter Gaussian vectors weighting function to determine spatial autocorrelation variable as a part of input feature vector.Secondly, extracting the when space characteristics of the data of LSTM Web vector graphic using convolutional neural networks, while attention model is respectively used to weighted feature figure and channel with the validity of Enhanced feature.Finally, the time series forecasting device based on depth LSTM, for learning the long-term and short-term dependence of pollutant.The present invention extracts intrinsic useful feature from history atmosphere pollution data, and auxiliary data is included in proposed model to improve performance, thus concentration prediction method.
Description
Technical field
The present invention relates to a kind of prediction technique of data-driven more particularly to a kind of mobile dirts based on space-time deep learning
Contaminate source emission concentration prediction method.
Background technique
Atmosphere pollution, especially mobile pollution source superfine particulate matter and volatile organic matter (Volatile Organic
Compounds, VOCs) caused by wide range of haze have become one of China's environmental problem most outstanding.Superfine particulate matter and
VOCs not only has serious direct harm to human health, while as PM2.5Important as precursors object and photochemical fog master
Component part is wanted, the formation of compound atmosphere pollution is played a crucial role.Therefore, mobile pollution source ultra-fine grain is monitored
The discharge amount of object and VOCs is to reduce haze weather and photochemical pollution, improves having for region city atmospheric environment quality
One of effect means.It is necessary to understand the source of these pollutants and quantity, to reduce atmosphere pollution to the greatest extent to the unfavorable shadow of health
It rings.On Spatial Dimension, the feature of the pollutant variation of different regions is analyzed, dissects pollutant concentration in time and sky
Between on otherness and Crack cause, and then improve pollutant concentration prediction efficiency and accurate reliability, be government control
Atmosphere pollution and traffic control, living trip provide decision-making foundation.
In general, air pollution processes are usually directed to a plurality of types of pollutants of interaction, and by local reaction,
The Spatio-temporal Evolution characteristic of air pollutant concentration and the influence of Confounding Factor, such as the direction of wind and humidity.Therefore, to air dirt
The research of dye object concentration prediction still faces following two challenge: (1) LSTM network is generally used for temporal characteristics extraction, but its disadvantage
It is the having differences property when list entries has long-rang dependence and complexity task.Model is difficult to learn reasonable vector
It indicates, keeps the learning efficiency of model poor;(2) air pollution causal path is complicated between the different location of nature.
Correlation between monitoring station is very complicated, because it may be by geography, natural process, the shadow of meteor or other factors
It rings.
Summary of the invention
The present invention provides a kind of base for the not high problem of the accuracy of air pollutant concentration is predicted in the prior art
In the mobile pollution source concentration of emission prediction technique of space-time deep learning.
A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning, comprising the following steps:
Step 1: solving the problems, such as temporal correlation using granger-causality test, Granger causality is as measure time
The index to interact between sequence, in recent decades always by favor.For complicated space factor, using Granger because
Fruit relationship analyzes the time series of air concentration.The time series of the air pollutants of one monitoring station is defined as Yi, separately
The time series of one monitoring point is Xi.The format and null hypothesis of Granger causality are:
This is NdThe neighborhood collection of space clustering (considers that all space factors carry out monitoring point using K-Means algorithm
Cluster);εtIt is a white Gaussian random vector, n is the quantity of timestamp, vector ΦiIt is corresponding weight Yi;μiRepresentation space
Position XiAnd YiBetween space weight.
However, atmosphere pollution is also different to the dependence of varying environment condition, pollutant is in diffusion process vulnerable to wind
To influence.Therefore, it is necessary to study the space-time causalities of different pollutants under wind directions different between different monitoring points.Based on
Lower example such as Fig. 1, it can be seen that the causality analysis between room and time should reflect two aspects: the 1) dependence of space-time structure
Property, it reflects propagation of the various pollutants on room and time;2) predictability shows that different environmental conditions may be led
Cause different space-time causalities.
For XiAnd YiSpace weight mui, wind direction and distance are combined with the concentration diffusion process of air pollutants, mentioned
Go out a kind of hyper parameter Gauss weight vectors based on gaussian kernel function, can be described as:
Wherein α (j) is learning rate, wherein dX/dYAnd θXYIt respectively represents apart from variable and angle variables.Distance can pass through
Euclidean distance directly calculates.First formula is determined using F inspection, and whether than second formula is more important.If so,
XiIt is YiCausality, so XiIt can be used for predicting YiNote that having used a bandwidth parameter σ, it indicates transformed
Compromise between direction and distant effect;
Step 2: space time feature extraction, usage history air pollutant concentration data establish a time-space relationship and are
Input of the three-dimensional matrice as CNN.Assuming that the monitoring station S Si={ s1,s2,…,s25Sum be spatially arranged sequentially section
On.The station S can provide one group of time detection data Di={ dt-r,…,dt-2,dt-1, wherein the inspection of t record at every point of time
Measured data isD is time detection data, and S is monitoring station, and ο is attribute.According to the sky of monitoring point on road
Between and Annual distribution be combined with all data, obtained a size be M={ S, D, ο } three-dimensional data matrix.This paper's is pre-
Survey problem can be with is defined as: for one group of given monitoring station S in given different time intervals, by its newest maximum time
Step t-r ..., and t-2, t-1 } historical data is modeled for temporal correlation and air pollutant concentration prediction.In order to improve effect
Rate, addition batch standardizes after second and third convolutional layer of model.Used here as nonlinear activation function SELU function,
SELU has better convergence and the problem of gradient disappears can be effectively avoided, this is specific in Klambauer et al.
It discusses.
Step 3: establishing characteristic pattern attention model, carry out convolution algorithm using convolution kernel to obtain output room and time
Feature Mapping.In attention model, F={ f (1), f (2) ..., f (j) } is that Feature Mapping, i.e. j are hidden in the output of convolutional layer
The quantity of ∈ R convolution kernel.It is calculated using following equation by equation:
F '=W ⊙ F
Here, W={ ω1,ω2,…,ωjIt is one group of weight matrix, size is identical as Feature Mapping.Generate the note of W
Meaning power model is made of three convolutional layers with stride 1.First convolutional layer has k × s filter, convolution kernel having a size of 5 ×
5, second and third layer have k filter, convolution kernel is having a size of 3 × 3, and each convolutional layer has 100 filters.⊙ is
The product of element.
Step 4: establishing channel attention model, the three-dimensional space-time data matrix of input model includes many different categories
Property, different attributes has different influences to the estimated performance of model.It is expected that this 34 kinds of useful informations are integrated into machine learning
In model, the study and analysis that have supervision are carried out, realizes accurately prediction.Meanwhile the different channels output one from convolutional layer
It is fixed that there is different influence on model.Therefore, different channels is weighted, to improve feature to model prediction performance
Validity.
F={ f (1), f (2) ..., f (j) } is mapped for input feature vector, wherein j is the port number in Feature Mapping, in F
The maximum pondization operation of middle execution is to obtain each channel C={ c1,c2,…,cjMaximum value.Attention mechanism will generate attention
Weight vectors V and weighting indicate that the following equation of F ' calculates
ci=maxf (j)
V=softmax (C)
F '=V ⊙ F
Step 5: multiple monitoring point air pollution concentration predictions obtain weight temporal feature and sky from convolution sum merging treatment
Between after feature, F can be passed throughST=F 's⊕F′tIt by element adds two characteristic patterns and compresses them into one-dimensional characteristic X
=(x1,x2,…,xt).These features are broken down into continuous component and are fed to duplicate LSTM unit to carry out the time point
Analysis.Then using LSTM layers from X=(x1,x2,…,xt) extracting time information input, and another input is walked from final time
Suddenly the hidden unit h startedt-1.The forward direction training process of Attention-CNN-LSTM can be indicated with following equation:
ft=σ (Wf·g[ht-1,xt]+bf)
it=σ (Wi·g[ht-1,xt]+bi)
Ct=ft×Ct-1+it×tanh(WC·g[ht-1,xt]+bC)
ot=σ (Wo·g[ht-1,xt]+bo)
ht=ot·tanh(Ct)
Wherein it, otAnd ftRespectively indicate input gate, out gate and the activation for forgeing door;CtAnd htRespectively indicate each cell
With the activation carrier of block of memory;W and b respectively indicate weight matrix and bias vector.Then, only by the defeated of the final step of LSTM
It is fed in the layer being fully connected out, is used for space-time Air Pollution Forecast.
Training terminates, and the weighting parameter of model determines, model at this time is using granger-causality test spatial coherence
Wind direction influence factor is added in analysis, weighs hiding feature using attention model with the validity of Enhanced feature, it will be real-time
Weather information, contaminant information and timestamp information be input in model as mobile pollution source emissions concentration, model
The concentration results of i.e. exportable following 24 hours multi-scale predictions.
In prediction technique of the invention, the Granger causality between analysis website is used, and generate spatial data
As opposite website and time-space attribute, space characteristics and temporal characteristics are extracted using common convolution unit, and by using attention
Power mechanism is weighted feature, by stacked multilayer LSTM, can successively automatically extract the space phase with long-rang dependence
The feature of pollutant data is closed, and it is pre- come the multiple dimensioned time series for predicting air pollutant concentration that fusion feature can be used
It surveys.The method achieve higher Stability and veracities.
Detailed description of the invention
Fig. 1 is pollutant spatial Correlation Analysis figure;
Fig. 2 is based on space-time deep learning mobile pollution source concentration prediction model structure chart.
Specific embodiment
As depicted in figs. 1 and 2, the method is specifically implemented by the following steps:
In view of the historic state of temporal correlation and monitoring station between 25 monitoring stations, select two or more
CNN layers are extracted the internal characteristics of long-term span study from history air pollutants data, then using single heat coding.The party
Method encodes data per hour, and the feature of extraction is mutually tied to current meteorological data and relevant pollutant data
It closes, to improve the estimated performance of model.Space characteristics and temporal characteristics are extracted using Liang Ge branch, then use attention model
To weigh hiding feature with the validity of Enhanced feature.By stacked multilayer LSTM, can successively automatically extract have for a long time according to
Rely property space correlation pollutant data feature, and can be used fusion feature come predict air pollutants it is multiple dimensioned when
Between sequence prediction.
Step 1: solving the problems, such as temporal correlation using granger-causality test, Granger causality is as measure time
The index to interact between sequence, in recent decades always by favor.For complicated space factor, using Granger because
Fruit relationship analyzes the time series of air concentration.The time series of the air pollutants of one monitoring station is defined as Yi, separately
The time series of one monitoring point is Xi.The format and null hypothesis of Granger causality are:
This is NdThe neighborhood collection of space clustering (considers that all space factors carry out monitoring point using K-Means algorithm
Cluster);εtIt is a white Gaussian random vector, n is the quantity of timestamp, vector ΦiIt is corresponding weight Yi;μiRepresentation space
Position XiAnd YiBetween space weight.
However, atmosphere pollution is also different to the dependence of varying environment condition, pollutant is in diffusion process vulnerable to wind
To influence.Therefore, it is necessary to study the space-time causalities of different pollutants under wind directions different between different monitoring points.Based on
Lower example such as Fig. 1, it can be seen that the causality analysis between room and time should reflect two aspects: the 1) dependence of space-time structure
Property, it reflects propagation of the various pollutants on room and time;2) predictability shows that different environmental conditions may be led
Cause different space-time causalities.
For XiAnd YiSpace weight mui, wind direction and distance are combined with the concentration diffusion process of air pollutants, mentioned
Go out a kind of hyper parameter Gauss weight vectors based on gaussian kernel function, can be described as:
Wherein α (j) is learning rate, wherein dX/dYAnd θXYIt respectively represents apart from variable and angle variables.Distance can pass through
Euclidean distance directly calculates.First formula is determined using F inspection, and whether than second formula is more important.If so,
XiIt is YiCausality, so XiIt can be used for predicting Yi;Note that having used a bandwidth parameter σ, it indicates transformed
Compromise between direction and distant effect.
Step 2: space time feature extraction, usage history air pollutant concentration data establish a time-space relationship and are
Input of the three-dimensional matrice as CNN.Assuming that the monitoring station S Si={ s1,s2,…,s25Sum be spatially arranged sequentially section
On.The station S can provide one group of time detection data Di={ dt-r,…,dt-2,dt-1, wherein the inspection of t record at every point of time
Measured data isD is time detection data, and S is monitoring station, and ο is attribute.According to the sky of monitoring point on road
Between and Annual distribution be combined with all data, obtained a size be M={ S, D, ο } three-dimensional data matrix.This paper's is pre-
Survey problem can be with is defined as: for one group of given monitoring station S in given different time intervals, by its newest maximum time
Step t-r ..., and t-2, t-1 } historical data is modeled for temporal correlation and air pollutant concentration prediction.In order to improve effect
Rate, addition batch standardizes after second and third convolutional layer of model.Used here as nonlinear activation function SELU function,
SELU has better convergence and the problem of gradient disappears can be effectively avoided.
Step 3: establishing characteristic pattern attention model, carry out convolution algorithm using convolution kernel to obtain output room and time
Feature Mapping.In attention model, F={ f (1), f (2) ..., f (j) } is that Feature Mapping, i.e. j are hidden in the output of convolutional layer
The quantity of ∈ R convolution kernel.It is calculated using following equation by equation:
F '=W ⊙ F
Here, W={ ω1,ω2,…,ωjIt is one group of weight matrix, size is identical as Feature Mapping.Generate the note of W
Meaning power model is made of three convolutional layers with stride 1.First convolutional layer has k × s filter, convolution kernel having a size of 5 ×
5, second and third layer have k filter, convolution kernel is having a size of 3 × 3, and each convolutional layer has 100 filters.⊙ is
The product of element.
Step 4: establishing channel attention model, the three-dimensional space-time data matrix of input model includes many different categories
Property, different attributes has different influences to the estimated performance of model.It is expected that this 34 kinds of useful informations are integrated into machine learning
In model, the study and analysis that have supervision are carried out, realizes accurately prediction.Meanwhile the different channels output one from convolutional layer
It is fixed that there is different influence on model.Therefore, different channels is weighted, to improve feature to model prediction performance
Validity.
F={ f (1), f (2) ..., f (j) } is mapped for input feature vector, wherein j is the port number in Feature Mapping, in F
The maximum pondization operation of middle execution is to obtain each channel C={ c1,c2,…,cjMaximum value.Attention mechanism will generate attention
Weight vectors V and weighting indicate that the following equation of F ' calculates
ci=maxf (j)
V=softmax (C)
F '=V ⊙ F
Step 5: multiple monitoring point air pollution concentration predictions obtain weight temporal feature and sky from convolution sum merging treatment
Between after feature, F can be passed throughST=F 's⊕F′tIt by element adds two characteristic patterns and compresses them into one-dimensional characteristic X
=(x1,x2,…,xt).These features are broken down into continuous component and are fed to duplicate LSTM unit to carry out the time point
Analysis.Then using LSTM layers from X=(x1,x2,…,xt) extracting time information input, and another input is walked from final time
Suddenly the hidden unit h startedt-1.The forward direction training process of Attention-CNN-LSTM can be indicated with following equation:
ft=σ (Wf·g[ht-1,xt]+bf)
it=σ (Wi·g[ht-1,xt]+bi)
Ct=ft×Ct-1+it×tanh(WC·g[ht-1,xt]+bC)
ot=σ (Wo·g[ht-1,xt]+bo)
ht=ot·tanh(Ct)
Wherein it, otAnd ftRespectively indicate input gate, out gate and the activation for forgeing door;CtAnd htRespectively indicate each cell
With the activation carrier of block of memory;W and b respectively indicate weight matrix and bias vector.Then, only by the defeated of the final step of LSTM
It is fed in the layer being fully connected out, is used for space-time Air Pollution Forecast.
Training terminates, and the weighting parameter of model determines, model at this time is using granger-causality test spatial coherence
Wind direction influence factor is added in analysis, weighs hiding feature using attention model with the validity of Enhanced feature, it will be real-time
Weather information, contaminant information and timestamp information be input in model as mobile pollution source emissions concentration, model
The concentration results of i.e. exportable following 24 hours multi-scale predictions.
Step 7: the space-time deep learning model obtained to training is analyzed and is compared, and is analyzed and is compared, compared to
Other existing methods, can be preferably to the mobile pollution source emission of solution with the prediction model based on space-time deep learning
Concentration temporal correlation, and there is preferable prediction accuracy.
The above examples are only used to illustrate the technical scheme of the present invention and are not limiting, reference only to preferred embodiment to this hair
It is bright to be described in detail.Those skilled in the art should understand that can modify to technical solution of the present invention
Or equivalent replacement should all cover in claim model of the invention without departing from the spirit and scope of the technical solution of the present invention
In enclosing.
Claims (1)
1. a kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning, it is characterised in that: this method is specific
The following steps are included:
Step 1: solving the problems, such as temporal correlation using granger-causality test, for complicated space factor, use Granger
Causality analyzes the time series of air concentration;The time series of the air pollutants of one monitoring station is defined as Yi,
The time series of another monitoring point is Xi;The format and null hypothesis of Granger causality are:
Wherein NdThe neighborhood collection of space clustering, εtIt is a white Gaussian random vector, n is the quantity of timestamp, vector ΦiIt is pair
The weight Y answeredi, μiRepresentation space position XiAnd YiBetween space weight;
For XiAnd YiSpace weight mui, wind direction and distance are combined with the concentration diffusion process of air pollutants, proposed
A kind of hyper parameter Gauss weight vectors based on gaussian kernel function, description are as follows:
Wherein α (j) is learning rate, wherein dX/dYAnd θXYIt respectively represents apart from variable and angle variables, distance passes through euclidean
Distance directly calculates, and first formula is determined using F inspection, and whether than second formula is more important;If so, XiIt is YiBecause
Fruit relationship, so XiIt can be used for predicting Yi, bandwidth parameter σ, it indicates the compromise between transformed direction and distant effect;
Step 2: space time feature extraction, usage history air pollutant concentration data establish the three-dimensional of a time-space relationship
Matrix, the input as CNN;Assuming that the monitoring station S Si={ s1,s2,…,s25Sum be spatially arranged sequentially on section;S
One group of time detection data D can be provided by standingi={ dt-r,…,dt-2,dt-1, wherein the testing number of t record at every point of time
According to beingD is time detection data, and S is monitoring station, and ο is attribute;According to the space of monitoring point on road and
Annual distribution is combined with all data, has obtained the three-dimensional data matrix that a size is M={ S, D, ο };Thus forecasting problem
Is defined as: for one group of given monitoring station S in given different time intervals, by its newest maximum time step { t-
R ..., t-2, t-1 } historical data is modeled for temporal correlation and air pollutant concentration prediction;In order to improve efficiency, in mould
Batch is added after second and third convolutional layer of type to standardize;
Step 3: establishing characteristic pattern attention model, carry out convolution algorithm using convolution kernel to obtain output room and time feature
Mapping;In attention model, F={ f (1), f (2) ..., f (j) } is that Feature Mapping, i.e. j ∈ R volume are hidden in the output of convolutional layer
The quantity of product core;It is calculated using following equation by equation:
F '=W ⊙ F
Here, W={ ω1,ω2,…,ωjIt is one group of weight matrix, size is identical as Feature Mapping;Generate the attention of W
Model is made of three convolutional layers with stride 1;First convolutional layer has k × s filter, and convolution kernel is having a size of 5 × 5, and the
Two and third layer have k filter, convolution kernel is having a size of 3 × 3, and each convolutional layer has 100 filters;⊙ is element
Product;
Step 4: establishing channel attention model
F={ f (1), f (2) ..., f (j) } is mapped for input feature vector, wherein j is the port number in Feature Mapping, is held in F
Row maximum pondization operation is to obtain each channel C={ c1,c2,…,cjMaximum value;Attention mechanism, which will generate, pays attention to weight
Vector V and weighting indicate that the following equation of F ' calculates
ci=maxf (j)
V=softmax (C)
F '=V ⊙ F
Step 5: multiple monitoring point air pollution concentration predictions obtain weight temporal feature from convolution sum merging treatment and space are special
After sign, pass through FST=F 's⊕F′tIt by element adds two characteristic patterns and compresses them into one-dimensional characteristic X=(x1,
x2,…,xt);These features are broken down into continuous component and are fed to duplicate LSTM unit to carry out time analysis;Then
Using LSTM layers from X=(x1,x2,…,xt) extracting time information input, and another input is since final time step
Hidden unit ht-1;The forward direction training process of Attention-CNN-LSTM can be indicated with following equation:
ft=σ (Wf·g[ht-1,xt]+bf)
it=σ (Wi·g[ht-1,xt]+bi)
Ct=ft×Ct-1+it×tanh(WC·g[ht-1,xt]+bC)
ot=σ (Wo·g[ht-1,xt]+bo)
ht=ot·tanh(Ct)
Wherein it, otAnd ftRespectively indicate input gate, out gate and the activation for forgeing door, CtAnd htRespectively indicate each cell and note
Recall the activation carrier of block, W and b respectively indicate weight matrix and bias vector;Then, only the output of the final step of LSTM is presented
It is sent in the layer being fully connected, is used for space-time Air Pollution Forecast;
Training terminates, and the weighting parameter of model determines, model at this time is using granger-causality test spatial Correlation Analysis
Middle addition wind direction influence factor weighs hiding feature using attention model with the validity of Enhanced feature, by real-time gas
Image information, contaminant information and timestamp information are input in model as mobile pollution source emissions concentration, model
Export the concentration results of following 24 hours multi-scale predictions.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811541300.5A CN109492830B (en) | 2018-12-17 | 2018-12-17 | Mobile pollution source emission concentration prediction method based on time-space deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811541300.5A CN109492830B (en) | 2018-12-17 | 2018-12-17 | Mobile pollution source emission concentration prediction method based on time-space deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109492830A true CN109492830A (en) | 2019-03-19 |
CN109492830B CN109492830B (en) | 2021-08-31 |
Family
ID=65710608
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811541300.5A Expired - Fee Related CN109492830B (en) | 2018-12-17 | 2018-12-17 | Mobile pollution source emission concentration prediction method based on time-space deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109492830B (en) |
Cited By (40)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109993282A (en) * | 2019-03-28 | 2019-07-09 | 北京航空航天大学 | A kind of typhoon involves the prediction technique of range |
CN110135057A (en) * | 2019-05-14 | 2019-08-16 | 北京工业大学 | Solid waste burning process dioxin concentration flexible measurement method based on multilayer feature selection |
CN110163449A (en) * | 2019-05-31 | 2019-08-23 | 杭州电子科技大学 | A kind of motor vehicle blowdown monitoring node dispositions method based on active space-time diagram convolution |
CN110197307A (en) * | 2019-06-03 | 2019-09-03 | 上海海洋大学 | A kind of domain type sea-surface temperature prediction technique of combination attention mechanism |
CN110263977A (en) * | 2019-05-24 | 2019-09-20 | 河南大学 | The method and device of Optimization Prediction PM2.5 based on LSTM neural network model |
CN110544304A (en) * | 2019-07-18 | 2019-12-06 | 长春市万易科技有限公司 | space-time reasoning-based site pollution digitization and graphical display system and method |
CN110807577A (en) * | 2019-10-15 | 2020-02-18 | 中国石油天然气集团有限公司 | Pollution emission prediction method and device |
CN110929793A (en) * | 2019-11-27 | 2020-03-27 | 谢国宇 | Time-space domain model modeling method and system for ecological environment monitoring |
CN111047012A (en) * | 2019-12-06 | 2020-04-21 | 重庆大学 | Air quality prediction method based on deep bidirectional long-short term memory network |
CN111341386A (en) * | 2020-02-17 | 2020-06-26 | 大连理工大学 | Attention-introducing multi-scale CNN-BilSTM non-coding RNA interaction relation prediction method |
CN111340669A (en) * | 2020-02-21 | 2020-06-26 | 中国科学技术大学 | Crowd funding project initial stage financing performance prediction system |
CN111401605A (en) * | 2020-02-17 | 2020-07-10 | 北京石油化工学院 | Interpretable prediction method for atmospheric pollution |
CN111768038A (en) * | 2020-06-30 | 2020-10-13 | 平安国际智慧城市科技股份有限公司 | Pollutant monitoring method and device, terminal equipment and storage medium |
CN111814956A (en) * | 2020-06-23 | 2020-10-23 | 哈尔滨工程大学 | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction |
CN111832814A (en) * | 2020-07-01 | 2020-10-27 | 北京工商大学 | Air pollutant concentration prediction method based on graph attention machine mechanism |
CN111882138A (en) * | 2020-08-07 | 2020-11-03 | 中国农业大学 | Water quality prediction method, device, equipment and storage medium based on space-time fusion |
CN112288156A (en) * | 2020-10-26 | 2021-01-29 | 哈尔滨工程大学 | Air quality prediction method based on meta-learning and graph attention space-time neural network |
CN112580859A (en) * | 2020-06-01 | 2021-03-30 | 北京理工大学 | Haze prediction method based on global attention mechanism |
CN112651553A (en) * | 2020-12-24 | 2021-04-13 | 航天科工网络信息发展有限公司 | Urban atmospheric pollution early warning supervision and decision analysis method |
CN112819218A (en) * | 2021-01-27 | 2021-05-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
CN112884243A (en) * | 2021-03-16 | 2021-06-01 | 杭州职业技术学院 | Air quality analysis and prediction method based on deep learning and Bayesian model |
CN112906941A (en) * | 2021-01-21 | 2021-06-04 | 哈尔滨工程大学 | Prediction method and system for dynamic correlation air quality time series |
CN112950924A (en) * | 2019-12-10 | 2021-06-11 | 东北大学秦皇岛分校 | Complex traffic network traffic speed prediction method based on deep learning |
CN112966442A (en) * | 2021-03-08 | 2021-06-15 | 浙江传媒学院 | Haze analysis and identification method based on causal centrality |
CN112986492A (en) * | 2019-12-17 | 2021-06-18 | 华为技术有限公司 | Method and device for establishing gas concentration prediction model |
CN113077081A (en) * | 2021-03-26 | 2021-07-06 | 航天科工智能运筹与信息安全研究院(武汉)有限公司 | Traffic pollution emission prediction method |
CN113128776A (en) * | 2021-04-26 | 2021-07-16 | 中国科学技术大学先进技术研究院 | Data self-migration multi-vehicle type diesel vehicle emission prediction method and system |
CN113255995A (en) * | 2021-05-24 | 2021-08-13 | 北京建筑大学 | Air pollution prediction method |
CN113326981A (en) * | 2021-05-26 | 2021-08-31 | 北京交通大学 | Atmospheric environment pollutant prediction model based on dynamic space-time attention mechanism |
CN113488058A (en) * | 2021-06-23 | 2021-10-08 | 武汉理工大学 | Voiceprint recognition method based on short voice |
CN113570156A (en) * | 2021-08-18 | 2021-10-29 | 中国农业大学 | Thermal environment prediction model and thermal environment prediction method based on agricultural facilities |
CN113657582A (en) * | 2021-08-20 | 2021-11-16 | 西安科技大学 | Spatiotemporal graph neural network gas concentration prediction method based on spatiotemporal data |
CN113742893A (en) * | 2021-08-06 | 2021-12-03 | 中煤科工集团沈阳研究院有限公司 | Gas concentration prediction method based on multi-element fusion space-time feature convolution network |
CN113919235A (en) * | 2021-10-29 | 2022-01-11 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Method and medium for detecting abnormal emission of mobile source pollution based on LSTM evolution clustering |
CN114511239A (en) * | 2022-02-18 | 2022-05-17 | 中国农业科学院农业信息研究所 | Method and device for dividing soil heavy metal pollution risk area, electronic equipment and medium |
CN114529825A (en) * | 2022-04-24 | 2022-05-24 | 城云科技(中国)有限公司 | Target detection model, method and application for fire fighting channel occupation target detection |
CN116008481A (en) * | 2023-01-05 | 2023-04-25 | 山东理工大学 | Air pollutant monitoring method and device based on large-range ground monitoring station |
CN116227756A (en) * | 2023-05-10 | 2023-06-06 | 北京航空航天大学 | Evaluation method of urban space-time data prediction causal model |
CN116578761A (en) * | 2023-05-18 | 2023-08-11 | 图林科技(深圳)有限公司 | Deep learning-based big data intelligent analysis method |
CN117744704A (en) * | 2024-02-21 | 2024-03-22 | 云南宇松科技有限公司 | Flue gas pollution source acquisition monitoring system, method and readable storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102682209A (en) * | 2012-05-03 | 2012-09-19 | 桂林理工大学 | Variable selection method for modeling organic pollutant quantitative structure and activity relationship |
CN104616060A (en) * | 2014-12-23 | 2015-05-13 | 南京工程学院 | Method for predicating contamination severity of insulator based on BP neural network and fuzzy logic |
CN107341576A (en) * | 2017-07-14 | 2017-11-10 | 河北百斛环保科技有限公司 | A kind of visual air pollution of big data is traced to the source and trend estimate method |
CN107346459A (en) * | 2017-05-22 | 2017-11-14 | 天津科技大学 | One kind is based on the improved multi-mode pollutant Integrated Forecast Method of genetic algorithm |
CN108009674A (en) * | 2017-11-27 | 2018-05-08 | 上海师范大学 | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks |
CN108052793A (en) * | 2017-12-26 | 2018-05-18 | 杭州电子科技大学 | A kind of mobile pollution source concentration of emission Forecasting Methodology based on FUZZY WEIGHTED ELM |
KR20180084417A (en) * | 2017-01-17 | 2018-07-25 | 국민대학교산학협력단 | Energy control process and system for seawater desalination plant |
CN108491970A (en) * | 2018-03-19 | 2018-09-04 | 东北大学 | A kind of Predict Model of Air Pollutant Density based on RBF neural |
CN108593557A (en) * | 2018-03-13 | 2018-09-28 | 杭州电子科技大学 | Based on TE-ANN-AWF mobile pollution source telemetry errors compensation methodes |
-
2018
- 2018-12-17 CN CN201811541300.5A patent/CN109492830B/en not_active Expired - Fee Related
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102682209A (en) * | 2012-05-03 | 2012-09-19 | 桂林理工大学 | Variable selection method for modeling organic pollutant quantitative structure and activity relationship |
CN104616060A (en) * | 2014-12-23 | 2015-05-13 | 南京工程学院 | Method for predicating contamination severity of insulator based on BP neural network and fuzzy logic |
KR20180084417A (en) * | 2017-01-17 | 2018-07-25 | 국민대학교산학협력단 | Energy control process and system for seawater desalination plant |
CN107346459A (en) * | 2017-05-22 | 2017-11-14 | 天津科技大学 | One kind is based on the improved multi-mode pollutant Integrated Forecast Method of genetic algorithm |
CN107341576A (en) * | 2017-07-14 | 2017-11-10 | 河北百斛环保科技有限公司 | A kind of visual air pollution of big data is traced to the source and trend estimate method |
CN108009674A (en) * | 2017-11-27 | 2018-05-08 | 上海师范大学 | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks |
CN108052793A (en) * | 2017-12-26 | 2018-05-18 | 杭州电子科技大学 | A kind of mobile pollution source concentration of emission Forecasting Methodology based on FUZZY WEIGHTED ELM |
CN108593557A (en) * | 2018-03-13 | 2018-09-28 | 杭州电子科技大学 | Based on TE-ANN-AWF mobile pollution source telemetry errors compensation methodes |
CN108491970A (en) * | 2018-03-19 | 2018-09-04 | 东北大学 | A kind of Predict Model of Air Pollutant Density based on RBF neural |
Non-Patent Citations (2)
Title |
---|
OSAMA A.GHONEIM.ETC: "Forecasting of Ozone Concentration in Smart City", 《2017 INTERNATIONAL CONFERENCE ON ADVANCES IN COMPUTING, COMMUNICATIONS AND INFORMATICS (ICACCI)》 * |
尹文君 等: "基于深度学习的大数据空气污染预报", 《中国环境管理》 * |
Cited By (58)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109993282B (en) * | 2019-03-28 | 2020-06-12 | 北京航空航天大学 | Typhoon wave and range prediction method |
CN109993282A (en) * | 2019-03-28 | 2019-07-09 | 北京航空航天大学 | A kind of typhoon involves the prediction technique of range |
CN110135057A (en) * | 2019-05-14 | 2019-08-16 | 北京工业大学 | Solid waste burning process dioxin concentration flexible measurement method based on multilayer feature selection |
CN110263977A (en) * | 2019-05-24 | 2019-09-20 | 河南大学 | The method and device of Optimization Prediction PM2.5 based on LSTM neural network model |
CN110163449A (en) * | 2019-05-31 | 2019-08-23 | 杭州电子科技大学 | A kind of motor vehicle blowdown monitoring node dispositions method based on active space-time diagram convolution |
CN110197307A (en) * | 2019-06-03 | 2019-09-03 | 上海海洋大学 | A kind of domain type sea-surface temperature prediction technique of combination attention mechanism |
CN110197307B (en) * | 2019-06-03 | 2023-07-25 | 上海海洋大学 | Regional sea surface temperature prediction method combined with attention mechanism |
CN110544304A (en) * | 2019-07-18 | 2019-12-06 | 长春市万易科技有限公司 | space-time reasoning-based site pollution digitization and graphical display system and method |
CN110544304B (en) * | 2019-07-18 | 2023-03-14 | 长春市万易科技有限公司 | Space-time reasoning-based site pollution digitization and graphical display system and method |
CN110807577A (en) * | 2019-10-15 | 2020-02-18 | 中国石油天然气集团有限公司 | Pollution emission prediction method and device |
CN110929793A (en) * | 2019-11-27 | 2020-03-27 | 谢国宇 | Time-space domain model modeling method and system for ecological environment monitoring |
CN111047012A (en) * | 2019-12-06 | 2020-04-21 | 重庆大学 | Air quality prediction method based on deep bidirectional long-short term memory network |
CN112950924A (en) * | 2019-12-10 | 2021-06-11 | 东北大学秦皇岛分校 | Complex traffic network traffic speed prediction method based on deep learning |
CN112986492A (en) * | 2019-12-17 | 2021-06-18 | 华为技术有限公司 | Method and device for establishing gas concentration prediction model |
CN111401605A (en) * | 2020-02-17 | 2020-07-10 | 北京石油化工学院 | Interpretable prediction method for atmospheric pollution |
CN111401605B (en) * | 2020-02-17 | 2023-05-02 | 北京石油化工学院 | Interpreted prediction method for atmospheric pollution |
CN111341386B (en) * | 2020-02-17 | 2022-09-20 | 大连理工大学 | Attention-introducing multi-scale CNN-BilSTM non-coding RNA interaction relation prediction method |
CN111341386A (en) * | 2020-02-17 | 2020-06-26 | 大连理工大学 | Attention-introducing multi-scale CNN-BilSTM non-coding RNA interaction relation prediction method |
CN111340669A (en) * | 2020-02-21 | 2020-06-26 | 中国科学技术大学 | Crowd funding project initial stage financing performance prediction system |
CN112580859A (en) * | 2020-06-01 | 2021-03-30 | 北京理工大学 | Haze prediction method based on global attention mechanism |
CN111814956A (en) * | 2020-06-23 | 2020-10-23 | 哈尔滨工程大学 | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction |
CN111768038A (en) * | 2020-06-30 | 2020-10-13 | 平安国际智慧城市科技股份有限公司 | Pollutant monitoring method and device, terminal equipment and storage medium |
CN111832814A (en) * | 2020-07-01 | 2020-10-27 | 北京工商大学 | Air pollutant concentration prediction method based on graph attention machine mechanism |
CN111832814B (en) * | 2020-07-01 | 2023-06-23 | 北京工商大学 | Air pollutant concentration prediction method based on graph attention mechanism |
CN111882138A (en) * | 2020-08-07 | 2020-11-03 | 中国农业大学 | Water quality prediction method, device, equipment and storage medium based on space-time fusion |
CN111882138B (en) * | 2020-08-07 | 2024-02-23 | 中国农业大学 | Water quality prediction method, device, equipment and storage medium based on space-time fusion |
CN112288156B (en) * | 2020-10-26 | 2023-03-31 | 哈尔滨工程大学 | Air quality prediction method based on meta-learning and graph attention space-time neural network |
CN112288156A (en) * | 2020-10-26 | 2021-01-29 | 哈尔滨工程大学 | Air quality prediction method based on meta-learning and graph attention space-time neural network |
CN112651553A (en) * | 2020-12-24 | 2021-04-13 | 航天科工网络信息发展有限公司 | Urban atmospheric pollution early warning supervision and decision analysis method |
CN112906941A (en) * | 2021-01-21 | 2021-06-04 | 哈尔滨工程大学 | Prediction method and system for dynamic correlation air quality time series |
CN112819218A (en) * | 2021-01-27 | 2021-05-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
CN112819218B (en) * | 2021-01-27 | 2022-05-06 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
CN112966442B (en) * | 2021-03-08 | 2023-05-23 | 浙江传媒学院 | Haze analysis and identification method based on causality centrality |
CN112966442A (en) * | 2021-03-08 | 2021-06-15 | 浙江传媒学院 | Haze analysis and identification method based on causal centrality |
CN112884243A (en) * | 2021-03-16 | 2021-06-01 | 杭州职业技术学院 | Air quality analysis and prediction method based on deep learning and Bayesian model |
CN113077081A (en) * | 2021-03-26 | 2021-07-06 | 航天科工智能运筹与信息安全研究院(武汉)有限公司 | Traffic pollution emission prediction method |
CN113128776B (en) * | 2021-04-26 | 2023-07-07 | 中国科学技术大学先进技术研究院 | Multi-vehicle type diesel vehicle emission prediction method and system with data self-migration |
CN113128776A (en) * | 2021-04-26 | 2021-07-16 | 中国科学技术大学先进技术研究院 | Data self-migration multi-vehicle type diesel vehicle emission prediction method and system |
CN113255995A (en) * | 2021-05-24 | 2021-08-13 | 北京建筑大学 | Air pollution prediction method |
CN113326981B (en) * | 2021-05-26 | 2024-05-31 | 北京交通大学 | Atmospheric environmental pollutant prediction model based on dynamic space-time attention mechanism |
CN113326981A (en) * | 2021-05-26 | 2021-08-31 | 北京交通大学 | Atmospheric environment pollutant prediction model based on dynamic space-time attention mechanism |
CN113488058A (en) * | 2021-06-23 | 2021-10-08 | 武汉理工大学 | Voiceprint recognition method based on short voice |
CN113742893A (en) * | 2021-08-06 | 2021-12-03 | 中煤科工集团沈阳研究院有限公司 | Gas concentration prediction method based on multi-element fusion space-time feature convolution network |
CN113742893B (en) * | 2021-08-06 | 2024-02-06 | 中煤科工集团沈阳研究院有限公司 | Gas concentration prediction method based on multi-element fusion space-time characteristic convolution network |
CN113570156A (en) * | 2021-08-18 | 2021-10-29 | 中国农业大学 | Thermal environment prediction model and thermal environment prediction method based on agricultural facilities |
CN113657582A (en) * | 2021-08-20 | 2021-11-16 | 西安科技大学 | Spatiotemporal graph neural network gas concentration prediction method based on spatiotemporal data |
CN113657582B (en) * | 2021-08-20 | 2024-04-09 | 西安科技大学 | Gas concentration prediction method based on space-time data and space-time graph neural network |
CN113919235A (en) * | 2021-10-29 | 2022-01-11 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Method and medium for detecting abnormal emission of mobile source pollution based on LSTM evolution clustering |
CN114511239A (en) * | 2022-02-18 | 2022-05-17 | 中国农业科学院农业信息研究所 | Method and device for dividing soil heavy metal pollution risk area, electronic equipment and medium |
CN114529825A (en) * | 2022-04-24 | 2022-05-24 | 城云科技(中国)有限公司 | Target detection model, method and application for fire fighting channel occupation target detection |
CN114529825B (en) * | 2022-04-24 | 2022-07-22 | 城云科技(中国)有限公司 | Target detection model, method and application for fire fighting access occupied target detection |
CN116008481A (en) * | 2023-01-05 | 2023-04-25 | 山东理工大学 | Air pollutant monitoring method and device based on large-range ground monitoring station |
CN116227756B (en) * | 2023-05-10 | 2023-07-07 | 北京航空航天大学 | Evaluation method of urban space-time data prediction causal model |
CN116227756A (en) * | 2023-05-10 | 2023-06-06 | 北京航空航天大学 | Evaluation method of urban space-time data prediction causal model |
CN116578761A (en) * | 2023-05-18 | 2023-08-11 | 图林科技(深圳)有限公司 | Deep learning-based big data intelligent analysis method |
CN116578761B (en) * | 2023-05-18 | 2024-04-02 | 西安韵通网络科技有限公司 | Deep learning-based big data intelligent analysis method |
CN117744704A (en) * | 2024-02-21 | 2024-03-22 | 云南宇松科技有限公司 | Flue gas pollution source acquisition monitoring system, method and readable storage medium |
CN117744704B (en) * | 2024-02-21 | 2024-04-30 | 云南宇松科技有限公司 | Flue gas pollution source acquisition monitoring system, method and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109492830B (en) | 2021-08-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109492830A (en) | A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning | |
Pak et al. | A hybrid model based on convolutional neural networks and long short-term memory for ozone concentration prediction | |
Le et al. | Spatiotemporal deep learning model for citywide air pollution interpolation and prediction | |
CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
Ibrahim et al. | predictSLUMS: A new model for identifying and predicting informal settlements and slums in cities from street intersections using machine learning | |
Ye et al. | Attention-based spatio-temporal graph convolutional network considering external factors for multi-step traffic flow prediction | |
Muthukumar et al. | Predicting PM2. 5 atmospheric air pollution using deep learning with meteorological data and ground-based observations and remote-sensing satellite big data | |
Ozbek et al. | Deep learning approach for one-hour ahead forecasting of energy production in a solar-PV plant | |
Esquivel et al. | Spatio-temporal prediction of Baltimore crime events using CLSTM neural networks | |
CN109117883A (en) | SAR image sea ice classification method and system based on long memory network in short-term | |
CN112149887A (en) | PM2.5 concentration prediction method based on data space-time characteristics | |
CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
CN114936691A (en) | Temperature forecasting method integrating relevance weighting and space-time attention | |
CN113516304A (en) | Space-time joint prediction method and device for regional pollutants based on space-time graph network | |
CN115629160A (en) | Air pollutant concentration prediction method and system based on space-time diagram | |
Zhou et al. | Discovering spatio-temporal dependencies based on time-lag in intelligent transportation data | |
Gul et al. | Forecasting hazard level of air pollutants using LSTM’s | |
Gómez-Navarro et al. | Event selection for dynamical downscaling: a neural network approach for physically-constrained precipitation events | |
Liu et al. | Air quality index forecasting via genetic algorithm-based improved extreme learning machine | |
Zhang et al. | A spatial correlation prediction model of urban PM2. 5 concentration based on deconvolution and LSTM | |
Kvanum | Developing a deep learning forecasting system for short-term and high-resolution prediction of sea ice concentration | |
Guan et al. | Predicting short-term PM2. 5 concentrations at fine temporal resolutions using a multi-branch temporal graph convolutional neural network | |
Bonet et al. | Conditional variational graph autoencoder for air quality forecasting | |
Lu et al. | Application of offshore visibility forecast based on temporal convolutional network and transfer learning | |
CN116913098B (en) | Short-time traffic flow prediction method integrating air quality and vehicle flow data |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210831 |