CN112085163A - Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU - Google Patents
Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU Download PDFInfo
- Publication number
- CN112085163A CN112085163A CN202010870423.4A CN202010870423A CN112085163A CN 112085163 A CN112085163 A CN 112085163A CN 202010870423 A CN202010870423 A CN 202010870423A CN 112085163 A CN112085163 A CN 112085163A
- Authority
- CN
- China
- Prior art keywords
- data
- agc
- gru
- neural network
- air quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 11
- 125000004122 cyclic group Chemical group 0.000 title claims description 4
- 239000003344 environmental pollutant Substances 0.000 claims abstract description 16
- 231100000719 pollutant Toxicity 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000012360 testing method Methods 0.000 claims abstract description 8
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 4
- 238000012544 monitoring process Methods 0.000 claims description 24
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000003062 neural network model Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 239000000654 additive Substances 0.000 claims description 7
- 230000000996 additive effect Effects 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000002123 temporal effect Effects 0.000 claims description 3
- 238000011425 standardization method Methods 0.000 claims description 2
- 239000003570 air Substances 0.000 description 32
- 230000006870 function Effects 0.000 description 6
- 239000010410 layer Substances 0.000 description 6
- 238000003915 air pollution Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 239000012080 ambient air Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 210000000352 storage cell Anatomy 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
- G01N33/0068—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display using a computer specifically programmed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Combustion & Propulsion (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an air quality prediction method based on an attention enhancement graph convolutional neural network AGC and a gated circulation unit GRU, which specifically comprises the following steps: firstly, selecting pollutants to be predicted, acquiring a data set for air quality prediction, preprocessing, and dividing the preprocessed data set into a training set and a test set according to a proportion; secondly, an AGC-GRU model is constructed, a training set is input into the AGC-GRU model and is trained by adopting a back propagation algorithm, and optimal model parameters are obtained; and finally, inputting the test set into the trained AGC-GRU model for prediction to obtain the predicted value of the selected pollutants. The invention can simultaneously consider and extract the time sequence characteristics and the space characteristics of the related data sets, thereby achieving the purpose of accurate prediction.
Description
Technical Field
The invention belongs to the technical field of air quality prediction, and particularly relates to an air quality prediction method based on an attention enhancement graph convolutional neural network AGC and a gated circulation unit GRU.
Background
In recent years, with the development of industry, the problem of air pollution is becoming more serious, and the problem becomes a hot topic of attention. Various pollutants such as PM2.5 and PM10 have small particle radius, large area and strong diffusivity, are very easy to attach toxic and harmful substances, have great influence on human health and atmospheric environment, and are urgent to treat the air pollution problem. And accurate air quality prediction information can provide great help for air pollution control. In order to master the air pollution condition, the government establishes an air monitoring station to monitor the air quality of the area in real time, but the air monitoring station cannot predict the future air quality, so that the research on the air quality prediction can make up for the loss of the function of the monitoring station.
The study of air quality prediction has been a focus and has been addressed by researchers in different fields. The prediction of Air Quality Index (AQI) is somewhat challenging, mainly because of the excessive and unstable effects, not only geographical, but also human factors and chemical interactions between various pollutants. At present, air pollution prediction is mainly divided into two aspects, one is that a mechanism model based on an atmospheric chemical mode becomes a deterministic model, and the other is a statistical model based on machine learning. In recent years, many neural networks and related technologies of deep learning are applied to the field, and the model networks have good prediction effect and show certain prediction capability.
However, the existing air quality prediction model has some defects that (1) the factors influencing the air quality are not sufficiently considered. Due to the diffusivity of air, the air quality of a monitoring station to be predicted is influenced by the air quality of an adjacent area, and a proper method for acquiring spatial factors influencing air quality prediction is lacked. (2) The traditional recurrent neural network model has the problems of gradient disappearance, error accumulation and the like due to the increase of the length of a training sequence. (3) And the prediction is carried out only by means of short-term data modeling, and the characteristic that the air quality data has a periodic trend is ignored. At present, in a big data era, a large amount of ambient air data are possessed, model training can be carried out by utilizing massive data, and the parameter effect of the model is better under the big data training.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides an air quality prediction method based on an attention enhancement graph convolutional neural network (AGC) and a gated circulation unit (GRU). The invention simultaneously considers the air quality change of the same monitoring station at different time and also considers the spatial relationship of different monitoring stations.
The invention is realized by the following technical scheme, and provides an air quality prediction method based on an attention enhancement graph convolutional neural network AGC and a gated circulation unit GRU, which specifically comprises the following steps:
step 1: selecting pollutants to be predicted, collecting data for air quality prediction, and establishing a data set;
step 2: preprocessing the data acquired in the step 1;
step 3, dividing the preprocessed data obtained in the step 2 into a training set and a testing set according to a proportion;
step 4, constructing an AGC-GRU neural network model;
step 5, inputting the training set obtained in the step 3 into the AGC-GRU neural network model in the step 4 for training, and simultaneously extracting time sequence characteristics and space characteristics;
step 6, iterative learning is carried out on the AGC-GRU neural network in the step 5 by adopting a back propagation algorithm strategy to obtain optimal model parameters;
and 7, inputting the test set into the AGC-GRU neural network model trained in the step 6 for prediction to obtain a predicted value of the selected pollutant.
Further, the collecting data for air quality prediction specifically includes:
step 1.1: collecting monitoring station data, wherein the monitoring station data is the pollutant concentration historical data;
step 1.2: collecting road network data and extracting;
step 1.3: collecting weather data, wherein the weather data comprises historical weather data and weather forecast data;
step 1.4: collecting coordinate data of a monitoring station;
step 1.5: and collecting PoI data of the interest points.
Further, the pollutants to be predicted include PM2.5, PM10, CO and NO.
Further, the preprocessing is specifically abnormal data, missing data and data standardization processing.
Further, a linear interpolation method is adopted for missing data, that is, a missing value at the current time is estimated according to missing values at the previous time and the next time of the current time, and a specific calculation formula is as follows:
wherein, XtIs the missing value at time t, Xt-1And Xt+1Is a missing value of the preceding and following time points.
Further, the data normalization adopts a maximum and minimum normalization method, and converts the input data into numbers between [0,1] by using the maximum and minimum values in the data set, so as to reduce the data range, and the specific calculation formula is as follows:
wherein X is input data, XMINIs the minimum value of the data set, XMAXIs the maximum value of the data set, XnFor obtaining in a standardised mannerThe new value.
Further, the step 4 specifically includes:
step 4.1: replacing an operator in a gate control cycle unit GRU with an image convolution operator, so that a model can capture the time characteristic and the space characteristic of a data set at the same time;
step 4.2: and (4) inputting the hidden state matrix obtained in the step (4.1) into an additive attention mechanism, and further selecting a key node.
Further, in step 4.2, the intermediate hidden state in the AGC-GRU neural network model includes the temporal dynamic information and the spatial structure information in the data set, so that the key node can be trained.
The invention has the beneficial effects that:
(1) the invention can simultaneously consider the time and space characteristics of the data and realize the spatiotemporal co-occurrence of the data. The time characteristics are extracted by utilizing the GRU, and the spatial characteristics between the monitoring stations are learned by a graph convolution neural network GCN, so that the prediction effect is more accurate.
(2) The invention uses an additive attention mechanism, utilizes a hidden layer containing various information in GRU, and trains out key nodes.
(3) The method and the system can provide prediction data of each monitoring station area for the city, and make up the problem that the air quality monitoring station cannot predict.
Drawings
FIG. 1 is a flow chart of the steps of the method of the present invention;
FIG. 2 is a block diagram of an AGC-GRU neural network model according to the present invention;
FIG. 3 is a schematic diagram of an additive attention mechanism of the present invention;
FIG. 4 is a diagram of a graph of a convolutional neural network GCN structure in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
With reference to fig. 1 to 4, the present invention provides an air quality prediction method based on an attention-enhancing graph convolutional neural network AGC and a gated cyclic unit GRU, which specifically includes the following steps:
step 1: selecting pollutants to be predicted, collecting data for air quality prediction, and establishing a data set;
the data set comprises monitoring station data and road network data of the city to be predicted, weather data and the like. According to the method, air quality data, meteorological data, urban PoI data, road network data and industrial pollution data of 35 air quality monitoring stations in Beijing City for three years are selected as data sets, and experimental data span is 2016, 1 and 1 month and 1 day to 2019, 1 and 1 month and 1 day. The data updating time of the same monitoring station is 1 hour;
the data collected for air quality prediction specifically includes:
step 1.1: collecting monitoring station data, wherein the monitoring station data is the pollutant concentration historical data;
the data in the invention is derived from air quality data in Beijing City from an environmental protection detection center website in Beijing City, data of nearly three years are obtained, specifically, the data comprise pollutant concentration data such as PM2.5, PM10, CO, NO and the like, and the updating interval is 1 hour.
Step 1.2: collecting road network data and extracting;
in the invention, data is downloaded through an OpenSteetMap map to obtain road network data of each level in Beijing, and the data is extracted after processing.
Step 1.3: collecting weather data, wherein the weather data comprises historical weather data and weather forecast data; specifically temperature, wind speed, wind direction, etc.
The weather forecast data is obtained through data of a Chinese weather network, specifically comprises temperature, wind speed, wind direction and the like, the updating interval is 1 day, and the weather forecast data and the weather data have the same category and unit.
Step 1.4: collecting coordinate data of a monitoring station;
in the invention, longitude and latitude information of 35 monitoring stations in Beijing is obtained by crawling KDD data sets.
Step 1.5: and collecting PoI data of the interest points.
The interest point data is obtained by classifying and downloading the Taile map and is extracted by exporting a data format.
Step 2: preprocessing the data acquired in the step 1;
the preprocessing is specifically abnormal data, missing data and data standardization processing.
The preprocessing specifically comprises deleting abnormal data and missing data, and adopting a linear interpolation method for the missing data, namely estimating a missing value of the current time according to the missing values of the previous time and the next time of the current time, wherein a specific calculation formula is as follows:
wherein, XtIs the missing value at time t, Xt-1And Xt+1Is a missing value of the preceding and following time points.
The data standardization adopts a maximum and minimum value standardization method, input data are converted into numbers between [0,1] by utilizing the maximum value and the minimum value in a data set, so that the data range is reduced, and a specific calculation formula is as follows:
wherein X is input data, XMINIs the minimum value of the data set, XMAXIs the maximum value of the data set, XnIs a new value obtained in a standardized way.
Step 3, dividing the preprocessed data obtained in the step 2 into a training set and a testing set according to a proportion;
step 4, constructing an AGC-GRU neural network model;
the step 4 specifically comprises the following steps:
step 4.1: replacing an operator in a gate control cycle unit GRU with an image convolution operator, so that a model can capture the time characteristic and the space characteristic of a data set at the same time;
the graph convolution neural network is a method capable of deep learning of graph data and is a feed-forward neural network with a deep structure. The difference between graph convolutional neural networks and fully-connected neural networks is the weight assignment. Therefore, in the present invention, the Air Quality Index (AQI) between adjacent monitored sites can be modeled using the characteristics of graph convolution to find the internal relationship between them. The data input by the graph convolution neural network is the feature matrix and the adjacency matrix of the graph structure, as shown in fig. 4. { X1, X2, X3, X4.. } input, shared features eventually results in an output layer containing information about the nodes.
In the present invention, gated-loop unit (GRU) design: RNN (recurrent Neural Network, RNN) is an artificial Neural Network with a tree-like hierarchical structure, and nodes of RNN recursively input information in the order in which they are connected. The GRU network is a special RNN, which differs from RNNs in long-term learning dependencies. The duplicated modules in a conventional RNN include only a single layer, which is improved by the GRU network.
The GRU can remove or add information to the hidden state, managed through the gate structure. The network comprises a reset gate and an update gate, and information is selectively passed, removed or added into a hidden state through gate management. The input-output structure of the GRU is the same as that of a normal RNN. With a current input XtAnd the hidden state (hidden state) passed by the previous node, wherein the hidden state contains the related information of the previous node. Therefore, the GRU can well extract time series characteristics without adding a plurality of parameters, and can well extract the characteristics of the time series data of the monitoring station.
In the invention, the AGC-GRU model network not only has the time sequence modeling capability of GRU, but also can extract the spatial characteristics like a convolution network GCN to realize space-time co-occurrence. The network is formed by a GRU variant, the overall structure is similar to that of the GRU, wherein the AGC-GRU model network is different from the common GRU network in that the common GRU internally depends on a similar feedforward neural network for calculation, and the network can be called FC-GRU. And AGC-GRU replaces the calculation mode by graph convolution. The derivation formula is changed, and the new derivation formula is as follows:
rt＝σ(Wxr*graph Xt+Whr*graph Ht-1+br)
zt＝σ(Wxz*graph Xt+Whz*graph Ht-1+bz)
H′t-1＝rt*(Whn*graph Ht-1+bn)
H′＝tanh(Win*graph Xt+H′t-1+bh)
yt＝σ(W0,Ht)
whereingraphRepresenting the convolution operator of the graph, rtValue representing reset gating at time t, ztDenotes the value of the update gate at time t, σ denotes the sigmoid function, WxrFor resetting input X in gatingtWeight matrix of WxzFor updating input X in gatingtWeight matrix of WhrFor resetting gated input Ht-1Weight matrix of WhzFor updating gated input Ht-1Weight matrix of WhnIs rtAnd Ht-1Input H in stitchingt-1Weight matrix of WinIs H't-1And input XtIn the splicing of (2) input XtWeight matrix of brTo reset the deviant vector matrix in gating, bzTo be moreDeviation vector matrix in new gating, bnIs rtAnd Ht-1Deviation vector matrix in stitching, bhIs H't-1And input XtH 'represents H't-1And input XtThe data are scaled to the range of-1 to 1 through a tanh activation function,
for an intermediate hidden state, XtFeature matrix, H, representing the layers of storage cells input to the AGC-GRU model at time ttRepresenting the output of the AGC-GRU network at time t, i.e. the output of the AGC-GRU network
Output by attention mechanism, Ht-1Output of the AGC-GRU network model representing time t-1, W0Is an output H for the last layer of the modeltConversion to the finally predicted parameter, ytRepresenting the parameters of the final prediction and att represents the soft attention mechanism function. The specific model structure of the AGC-GRU is shown in fig. 2.
Step 4.2: and (4) inputting the hidden state matrix obtained in the step (4.1) into an additive attention mechanism, and further selecting a key node.
As shown in fig. 3, the objective of using the additive attention model is to strengthen the key node information without weakening the unfocused node information and maintain the integrity of the spatial information. The intermediate hidden state in the AGC-GRU model contains time dynamic information and space structure information in a data set, and key nodes are trained. According to the mechanism of an additive model, firstly, information of all nodes is aggregated into a query feature, and then the query feature and a feature matrix are trained to obtain the weight of each node:
where W is the learnable parameter matrix. q. q.stRepresenting query features into which information for all nodes is aggregated,ReLU denotes the activation function, N denotes the number of nodes,
the eigenvector representing each node in the hidden state, i.e., each row of the hidden matrix, the attention score of all nodes can be calculated as:
a of formula (II a)t＝(ɑt1，ɑt2，…ɑtN) And VS,Wh,WqAre all learnable parameter matrices. bhAnd btAre all deviation vector matrices. The present invention employs an s-shaped nonlinear function, taking into account the possibility that multiple key nodes exist. Node VtiHidden state of (H)tiMay also be expressed as (1+ alpha)ti)·Hti。
In the experiment, the graph structure in the AGC-GRU model is simulated by adopting the distance between detection stations, the distance between the monitoring stations exceeds 20km, edges exist between nodes, and otherwise, no correlation exists between the two nodes.
In step 4.2, the intermediate hidden state in the AGC-GRU neural network model contains the temporal dynamic information and the spatial structure information in the data set, and the key node can be trained.
Step 5, inputting the training set obtained in the step 3 into the AGC-GRU neural network model in the step 4 for training, and simultaneously extracting time sequence characteristics and space characteristics;
step 6, iterative learning is carried out on the AGC-GRU neural network in the step 5 by adopting a back propagation algorithm strategy to obtain optimal model parameters;
the error term of each neuron is reversely calculated, and the method is mainly divided into two parts: one is backward propagation along the time direction, and the other is propagation of an error term to the upper layer; finally, calculating the gradient of each weight according to the corresponding error term; the parameters learned by the AGC-GRU are a weight matrix among the internal network layers and an offset vector matrix.
And 7, inputting the test set into the AGC-GRU neural network model trained in the step 6 for prediction to obtain a predicted value of the selected pollutant.
The air quality prediction method based on the attention-enhancing graph convolutional neural network AGC and the gated loop unit GRU, which is proposed by the present invention, is described in detail above, and the principle and the implementation of the present invention are explained herein by applying specific examples, and the description of the above examples is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (8)
1. An air quality prediction method based on an attention enhancement graph convolutional neural network AGC and a gated cyclic unit GRU is characterized by comprising the following steps: the method specifically comprises the following steps:
step 1: selecting pollutants to be predicted, collecting data for air quality prediction, and establishing a data set;
step 2: preprocessing the data acquired in the step 1;
step 3, dividing the preprocessed data obtained in the step 2 into a training set and a testing set according to a proportion;
step 4, constructing an AGC-GRU neural network model;
step 5, inputting the training set obtained in the step 3 into the AGC-GRU neural network model in the step 4 for training, and simultaneously extracting time sequence characteristics and space characteristics;
step 6, iterative learning is carried out on the AGC-GRU neural network in the step 5 by adopting a back propagation algorithm strategy to obtain optimal model parameters;
and 7, inputting the test set into the AGC-GRU neural network model trained in the step 6 for prediction to obtain a predicted value of the selected pollutant.
2. The method of claim 1, wherein: the data collected for air quality prediction specifically includes:
step 1.1: collecting monitoring station data, wherein the monitoring station data is the pollutant concentration historical data;
step 1.2: collecting road network data and extracting;
step 1.3: collecting weather data, wherein the weather data comprises historical weather data and weather forecast data;
step 1.4: collecting coordinate data of a monitoring station;
step 1.5: and collecting PoI data of the interest points.
3. The method of claim 1, wherein: the pollutants to be predicted include PM2.5, PM10, CO, and NO.
4. The method of claim 1, wherein: the preprocessing is specifically abnormal data, missing data and data standardization processing.
5. The method of claim 4, wherein: for missing data, a linear interpolation method is adopted, that is, the missing value of the current time is estimated according to the missing values of the previous time and the next time of the current time, and the specific calculation formula is as follows:
wherein, XtIs the missing value at time t, Xt-1And Xt+1Is a missing value of the preceding and following time points.
6. The method of claim 4, wherein: the data standardization adopts a maximum and minimum value standardization method, input data are converted into numbers between [0,1] by utilizing the maximum value and the minimum value in a data set, so that the data range is reduced, and a specific calculation formula is as follows:
wherein X is input data, XMINIs the minimum value of the data set, XMAXIs the maximum value of the data set, XnIs a new value obtained in a standardized way.
7. The method of claim 1, wherein: the step 4 specifically comprises the following steps:
step 4.1: replacing an operator in a gate control cycle unit GRU with an image convolution operator, so that a model can capture the time characteristic and the space characteristic of a data set at the same time;
step 4.2: and (4) inputting the hidden state matrix obtained in the step (4.1) into an additive attention mechanism, and further selecting a key node.
8. The method of claim 7, wherein: in step 4.2, the intermediate hidden state in the AGC-GRU neural network model contains the temporal dynamic information and the spatial structure information in the data set, and the key node can be trained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010870423.4A CN112085163A (en) | 2020-08-26 | 2020-08-26 | Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010870423.4A CN112085163A (en) | 2020-08-26 | 2020-08-26 | Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112085163A true CN112085163A (en) | 2020-12-15 |
Family
ID=73728678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010870423.4A Pending CN112085163A (en) | 2020-08-26 | 2020-08-26 | Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112085163A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541302A (en) * | 2020-12-23 | 2021-03-23 | 北京百度网讯科技有限公司 | Air quality prediction model training method, air quality prediction method and device |
| CN112561191A (en) * | 2020-12-22 | 2021-03-26 | 北京百度网讯科技有限公司 | Prediction model training method, prediction method, device, apparatus, program, and medium |
| CN112668775A (en) * | 2020-12-25 | 2021-04-16 | 西安翔迅科技有限责任公司 | Air quality prediction method based on time sequence convolution network algorithm |
| CN112734017A (en) * | 2021-04-01 | 2021-04-30 | 成都信息工程大学 | Method for establishing precipitation prediction model |
| CN112883522A (en) * | 2021-01-14 | 2021-06-01 | 沈阳工业大学 | Micro-grid dynamic equivalent modeling method based on GRU (generalized regression Unit) recurrent neural network |
| CN112946187A (en) * | 2021-01-22 | 2021-06-11 | 西安科技大学 | Refuge chamber real-time state monitoring method based on neural network |
| CN113487856A (en) * | 2021-06-04 | 2021-10-08 | 兰州理工大学 | Traffic flow combination prediction model based on graph convolution network and attention mechanism |
| CN113496310A (en) * | 2021-06-16 | 2021-10-12 | 国家超级计算深圳中心(深圳云计算中心) | Atmospheric pollutant prediction method and system based on deep learning model |
| CN113505536A (en) * | 2021-07-09 | 2021-10-15 | 兰州理工大学 | Optimized traffic flow prediction model based on space-time diagram convolution network |
| CN113610286A (en) * | 2021-07-27 | 2021-11-05 | 中国地质大学（武汉） | PM accounting for spatio-temporal correlations and meteorological factors2.5Concentration prediction method and device |
| CN113762351A (en) * | 2021-08-12 | 2021-12-07 | 吉林大学 | Air quality prediction method based on deep transition network |
| CN114298445A (en) * | 2022-03-10 | 2022-04-08 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Site quantitative precipitation forecasting method and system based on graph convolution neural network |
| CN116050583A (en) * | 2022-12-15 | 2023-05-02 | 生态环境部华南环境科学研究所(生态环境部生态环境应急研究所) | Water environment quality deep learning prediction method coupled with space-time context information |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850913A (en) * | 2015-05-28 | 2015-08-19 | 深圳先进技术研究院 | Air quality PM2.5 predicating method and air quality PM2.5 predicating system |
| CN105893352A (en) * | 2016-04-15 | 2016-08-24 | 苏州爱诺信信息科技有限公司 | Air quality early-warning and monitoring analysis system based on big data of social network |
| CN109492822A (en) * | 2018-11-24 | 2019-03-19 | 上海师范大学 | Air pollutant concentration time-space domain interaction prediction method |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN110147877A (en) * | 2019-05-24 | 2019-08-20 | 福州大学 | PM2.5 prediction technique based on the long network in short-term of convolution |
| 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 |
| CN110598953A (en) * | 2019-09-23 | 2019-12-20 | 哈尔滨工程大学 | Space-time correlation air quality prediction method |
| CN110866631A (en) * | 2019-10-09 | 2020-03-06 | 重庆大学 | Method for predicting atmospheric pollution condition based on integrated gate recursion unit neural network GRU |
| CN111540199A (en) * | 2020-04-21 | 2020-08-14 | 浙江省交通规划设计研究院有限公司 | High-speed traffic flow prediction method based on multi-mode fusion and graph attention machine mechanism |
| CN111563611A (en) * | 2020-04-13 | 2020-08-21 | 北京工业大学 | Cloud data center renewable energy space-time prediction method for orientation graph convolutional network |
-
2020
- 2020-08-26 CN CN202010870423.4A patent/CN112085163A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850913A (en) * | 2015-05-28 | 2015-08-19 | 深圳先进技术研究院 | Air quality PM2.5 predicating method and air quality PM2.5 predicating system |
| CN105893352A (en) * | 2016-04-15 | 2016-08-24 | 苏州爱诺信信息科技有限公司 | Air quality early-warning and monitoring analysis system based on big data of social network |
| CN109492822A (en) * | 2018-11-24 | 2019-03-19 | 上海师范大学 | Air pollutant concentration time-space domain interaction prediction method |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN110147877A (en) * | 2019-05-24 | 2019-08-20 | 福州大学 | PM2.5 prediction technique based on the long network in short-term of convolution |
| 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 |
| CN110598953A (en) * | 2019-09-23 | 2019-12-20 | 哈尔滨工程大学 | Space-time correlation air quality prediction method |
| CN110866631A (en) * | 2019-10-09 | 2020-03-06 | 重庆大学 | Method for predicting atmospheric pollution condition based on integrated gate recursion unit neural network GRU |
| CN111563611A (en) * | 2020-04-13 | 2020-08-21 | 北京工业大学 | Cloud data center renewable energy space-time prediction method for orientation graph convolutional network |
| CN111540199A (en) * | 2020-04-21 | 2020-08-14 | 浙江省交通规划设计研究院有限公司 | High-speed traffic flow prediction method based on multi-mode fusion and graph attention machine mechanism |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112561191A (en) * | 2020-12-22 | 2021-03-26 | 北京百度网讯科技有限公司 | Prediction model training method, prediction method, device, apparatus, program, and medium |
| CN112561191B (en) * | 2020-12-22 | 2024-02-27 | 北京百度网讯科技有限公司 | Prediction model training method, prediction device, prediction apparatus, prediction program, and program |
| CN112541302B (en) * | 2020-12-23 | 2024-02-06 | 北京百度网讯科技有限公司 | Air quality prediction model training method, air quality prediction method and device |
| CN112541302A (en) * | 2020-12-23 | 2021-03-23 | 北京百度网讯科技有限公司 | Air quality prediction model training method, air quality prediction method and device |
| CN112668775A (en) * | 2020-12-25 | 2021-04-16 | 西安翔迅科技有限责任公司 | Air quality prediction method based on time sequence convolution network algorithm |
| CN112883522A (en) * | 2021-01-14 | 2021-06-01 | 沈阳工业大学 | Micro-grid dynamic equivalent modeling method based on GRU (generalized regression Unit) recurrent neural network |
| CN112946187A (en) * | 2021-01-22 | 2021-06-11 | 西安科技大学 | Refuge chamber real-time state monitoring method based on neural network |
| CN112734017A (en) * | 2021-04-01 | 2021-04-30 | 成都信息工程大学 | Method for establishing precipitation prediction model |
| CN113487856A (en) * | 2021-06-04 | 2021-10-08 | 兰州理工大学 | Traffic flow combination prediction model based on graph convolution network and attention mechanism |
| CN113496310A (en) * | 2021-06-16 | 2021-10-12 | 国家超级计算深圳中心(深圳云计算中心) | Atmospheric pollutant prediction method and system based on deep learning model |
| CN113505536A (en) * | 2021-07-09 | 2021-10-15 | 兰州理工大学 | Optimized traffic flow prediction model based on space-time diagram convolution network |
| CN113610286A (en) * | 2021-07-27 | 2021-11-05 | 中国地质大学（武汉） | PM accounting for spatio-temporal correlations and meteorological factors2.5Concentration prediction method and device |
| CN113610286B (en) * | 2021-07-27 | 2024-03-29 | 中国地质大学（武汉） | PM taking into account space-time correlation and meteorological factors 2.5 Concentration prediction method and device |
| CN113762351B (en) * | 2021-08-12 | 2023-12-05 | 吉林大学 | Air quality prediction method based on deep transition network |
| CN113762351A (en) * | 2021-08-12 | 2021-12-07 | 吉林大学 | Air quality prediction method based on deep transition network |
| CN114298445A (en) * | 2022-03-10 | 2022-04-08 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Site quantitative precipitation forecasting method and system based on graph convolution neural network |
| CN114298445B (en) * | 2022-03-10 | 2022-06-03 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Site quantitative precipitation forecast method and system based on graph convolution neural network |
| CN116050583A (en) * | 2022-12-15 | 2023-05-02 | 生态环境部华南环境科学研究所(生态环境部生态环境应急研究所) | Water environment quality deep learning prediction method coupled with space-time context information |
| CN116050583B (en) * | 2022-12-15 | 2024-05-28 | 生态环境部华南环境科学研究所(生态环境部生态环境应急研究所) | Water environment quality deep learning prediction method coupled with space-time context information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112085163A (en) | Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU | |
| CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
| CN111899510B (en) | Intelligent traffic system flow short-term prediction method and system based on divergent convolution and GAT | |
| CN109492830B (en) | Mobile pollution source emission concentration prediction method based on time-space deep learning | |
| CN109142171B (en) | Urban PM10 concentration prediction method based on feature expansion and fusing with neural network | |
| CN114626512B (en) | High-temperature disaster forecasting method based on directed graph neural network | |
| Wang et al. | Modeling inter-station relationships with attentive temporal graph convolutional network for air quality prediction | |
| CN107909206B (en) | PM2.5 prediction method based on deep structure recurrent neural network | |
| CN114220271A (en) | Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network | |
| CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN109492760B (en) | Road traffic interpretable prediction method based on deep fuzzy cognitive map model | |
| CN109523021A (en) | A kind of dynamic network Structure Prediction Methods based on long memory network in short-term | |
| Zhang et al. | Short-term power load forecasting using integrated methods based on long short-term memory | |
| CN110555551A (en) | air quality big data management method and system for smart city | |
| CN114492922A (en) | Medium-and-long-term power generation capacity prediction method | |
| CN113011455B (en) | Air quality prediction SVM model construction method | |
| CN115860286B (en) | Air quality prediction method and system based on time sequence gate mechanism | |
| Dai et al. | Spatio-temporal deep learning framework for traffic speed forecasting in IoT | |
| Ouyang et al. | Spatial-temporal dynamic graph convolution neural network for air quality prediction | |
| CN116168548A (en) | Traffic flow prediction method of space-time attention pattern convolution network based on multi-feature fusion | |
| CN117494871A (en) | Ship track prediction method considering ship interaction influence | |
| CN110289987B (en) | Multi-agent system network anti-attack capability assessment method based on characterization learning | |
| CN111141879A (en) | Deep learning air quality monitoring method, device and equipment | |
| Jovova et al. | Air pollution forecasting using CNN-LSTM deep learning model |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201215 |
|
| RJ01 | Rejection of invention patent application after publication |