CN112819218A - High-resolution urban mobile source pollution space-time prediction method, system and storage medium - Google Patents
High-resolution urban mobile source pollution space-time prediction method, system and storage medium Download PDFInfo
- Publication number
- CN112819218A CN112819218A CN202110110653.5A CN202110110653A CN112819218A CN 112819218 A CN112819218 A CN 112819218A CN 202110110653 A CN202110110653 A CN 202110110653A CN 112819218 A CN112819218 A CN 112819218A
- Authority
- CN
- China
- Prior art keywords
- resolution
- time
- data
- source pollution
- space
- 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 33
- 238000003860 storage Methods 0.000 title claims abstract description 10
- 238000009826 distribution Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000012544 monitoring process Methods 0.000 claims abstract description 24
- 230000007613 environmental effect Effects 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 23
- 230000004927 fusion Effects 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000004069 differentiation Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 4
- MWUXSHHQAYIFBG-UHFFFAOYSA-N Nitric oxide Chemical compound O=[N] MWUXSHHQAYIFBG-UHFFFAOYSA-N 0.000 description 12
- 238000010586 diagram Methods 0.000 description 10
- 230000007246 mechanism Effects 0.000 description 7
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 4
- 229910002091 carbon monoxide Inorganic materials 0.000 description 4
- 238000011109 contamination Methods 0.000 description 4
- 239000003344 environmental pollutant Substances 0.000 description 3
- 231100000719 pollutant Toxicity 0.000 description 3
- 238000003915 air pollution Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004043 dyeing Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Biophysics (AREA)
- Educational Administration (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a high-resolution urban mobile source pollution space-time prediction method, a system and a storage medium, which comprises the steps of processing historical mobile source pollution monitoring data and external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution; carrying out standardization processing on the above; dividing the standardized low-resolution historical mobile source pollution data according to the length of a specified time interval, and combining the low-resolution historical mobile source pollution data with the external environmental factor data and the high-resolution mobile source pollution data at the current moment to construct a historical mobile source pollution data sample set; and constructing a high-resolution urban mobile source pollution space-time prediction model, training the model, and predicting the high-resolution mobile source pollution space-time distribution in a future time period by using the trained prediction model. The invention can effectively process the space structure constraint of the emission list under different space scales and realize the high-precision emission list space-time distribution prediction under the sparse monitoring.
Description
Technical Field
The invention relates to the technical field of environmental monitoring, in particular to a high-resolution urban mobile source pollution space-time prediction method, a high-resolution urban mobile source pollution space-time prediction system and a storage medium.
Background
With the rapid development of the urbanization process and the social economy in China, the quantity of motor vehicles kept in China is rapidly increased, and the motor vehicles become the first major country for the production and sale of the motor vehicles in the world in eleven years continuously. In 2019, the number of motor vehicles in China reaches 3.48 hundred million, which is 6.4% higher than that in 2018, and the problem of air pollution caused by motor vehicle emission is increasingly severe. The annual environmental management report of China Mobile Source (2020) shows that the total emission of four pollutants of motor vehicles in China is preliminarily accounted for 1603.8 ten thousand tons, wherein gasoline vehicles are main contributors to the total emission of the pollutants, and the emission of four main pollutants, such as carbon monoxide (CO), hydrocarbon (THC), nitrogen oxide (NOx) and Particulate Matters (PM), exceeds 90 percent. The mobile source pollution becomes an important source of air pollution in big and medium cities in China, and is an important reason for causing pollution of fine particulate matters and photochemical smog.
The objective of the analysis of the spatiotemporal characteristics of the mobile source emission list is to research the spatiotemporal distribution characteristics of the mobile source emission by using historical detection data of the actual mobile source emission, and the current research mainly focuses on single-road-section mobile source emission list prediction and urban area-based emission list prediction. The cyclic neural network model is mainly adopted for predicting the moving source emission list on the single road section, so that the time sequence dependency of the emission sequence on the single road section can be better learned, but the time-space correlation of the emission time-space sequence in the local region of the city cannot be processed. A space-time network model is mainly adopted in the aspect of urban area emission list prediction, but the method only considers the space-time change distribution prediction of the emission of the mobile source under the same spatial scale, and cannot construct space-time characteristic mapping relations of emission lists with different spatial resolutions.
Therefore, it is necessary to develop a high spatial-temporal resolution dynamic emission inventory prediction technique for mobile source pollution, which is significant for urban mobile source pollution accounting and emission control measure evaluation.
Disclosure of Invention
The invention provides a high-resolution urban mobile source pollution space-time prediction method, a high-resolution urban mobile source pollution space-time prediction system and a storage medium, which can solve the technical problems.
In order to achieve the purpose, the invention adopts the following technical scheme:
a high-resolution urban mobile source pollution space-time prediction method comprises the following steps:
step S1: processing historical mobile source pollution monitoring data and external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
step S2: carrying out standardized processing on low-resolution mobile source pollution data, external environment factor data and high-resolution mobile source pollution data;
step S3: dividing the standardized low-resolution historical mobile source pollution data according to the length of a specified time interval, and combining the low-resolution historical mobile source pollution data with the external environmental factor data and the high-resolution mobile source pollution data at the current moment to construct a historical mobile source pollution data sample set;
step S4: constructing a high-resolution urban mobile source pollution space-time prediction model, which comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
step S5: using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training, and performing model test by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
step S6: and predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
Further, the step S1 includes:
s1.1: accumulating the pollution monitoring data in the appointed time interval of each mobile source pollution emission observation point respectively to obtain the pollution monitoring data of the appointed time interval of each observation point, and generating low-resolution mobile source pollution data and high-resolution mobile source pollution data by utilizing reverse distance weighted interpolation according to different spatial resolution scales;
s1.2: the external environmental factor data is obtained by collecting weather information and holiday information corresponding to the high-resolution moving source pollution data time.
Further, the step S2 includes:
s2.1: selecting a maximum pollution monitoring value of a specified time interval, and normalizing the ratio of the low-resolution and high-resolution mobile source pollution data of the specified time interval to the maximum pollution monitoring value to be 0, 1]Interval standardized low-resolution and high-resolution moving source pollution data XC,XF;
S2.2: selecting each maximum weather factor of the appointed time interval, and normalizing the ratio of each weather factor data of the appointed time interval to each maximum weather factor to [0, 1%]Normalized weather data E of intervalsweatherFor holiday information, processing by single-hot coding EholidayCombining the two to form a standardized external factor EC={Eweather,Eholiday}。
Further, the step S3 includes:
s3.1: determining the wait for each time step tPredicting high resolution sample label XF(t) and corresponding time external environmental factor EC(t);
S3.2: at each time step t, the low-resolution pollution sequence is divided into an adjacent time segment, a short-time period time segment and a long-time trend time segment according to time sequence characteristics, and the adjacent time segment, the short-time period time segment and the long-time trend time segment are respectively recorded as c, p, s are three types of time slice intervals, respectively, lc,lp,lsAre three types of time slice lengths;
s3.3: a sample corresponding to time t is recorded asIs the input of the model, a total data sample set D is constructed for an acquirable time span T, while any time T is within the range of [ max { l [ ]c+1,p*lp+1,s*ls+1},min{T+c,T+p,T+s}]And taking the last 24 hours of data of the data sample set D as a test set and the others as a training set.
Further, the step S4 includes:
s4.1, constructing a low-resolution pollution space-time feature extraction module, wherein the module comprises a two-dimensional convolution layer, a ReLU activation function layer, a residual error unit and a fusion layer;
the spatio-temporal feature extraction is expressed as
Where Conv2D is a 2D convolution operation, l2DIs the number of 2D convolutional layers, are respectively provided withIs the input and output unit of the l-th layer residual error unit, k belongs to { c, p, s }; f is a residual mapping function; thetalIs a learning parameter;
s4.2: constructing an external influence factor fusion module, including feature dimension transformation and feature fusion;
will standardize weather data EweatherAnd holiday data E processed by single-hot codingholidayVia the full connection layer and remapped toThe same tensor shape size is expressed as
Then, the three extracted space-time change characteristics are fused by adopting a weight matrix, and external factors and pollution space-time characteristics are fused by adopting threshold activation, as shown in formula (2)
Wherein, Wc,Wp,WsLow resolution contaminated spatio-temporal features of near time segments, short time period time segments and long time trend time segments, respectivelyThe weight matrix of (a) is determined,representing the Hadamard product, HfcIs a fused low-resolution polluted spatio-temporal feature;
s4.3: constructing a high-resolution distributed upsampling feature mapping module which comprises a nearest neighbor interpolation unit and a distributed upsampling unit;
determining the expansion multiple N of low-resolution data according to the spatial scale of the low-resolution and high-resolution polluted data in a nearest neighbor interpolation unit, wherein N is the nth power of 4, and fusing the low-resolution polluted data by adopting nearest neighbor interpolation operationSpace-time characteristic of dyeing HfcUpsampling to high resolution scale Hff;
Hff=NearestNeighborUp(Hfc,N) (3)
In a distributed upsampling unit, N sub-pixel blocks, convolutional layers and N are used2Normalization layer N2Performing up-sampling reconstruction on the low-resolution features by Norm;
each sub-pixel block SubPixel enlarges the input characteristic size by 4 times, reduces the number of characteristic channels by 4 times, and N2The normalization layer normalizes each channel feature map according to the spatial structure constraint relationship of low-resolution and high-resolution pollution data, and the distributed upsampling operation is specifically expressed as follows,
wherein Hfo(i,j)=Ho(i,j)/∑i′,j′Ho(i′,j′),HfoNamely high resolution pollution space-time characteristic probability distribution, and high resolution scale pollution data H by combining nearest neighbor interpolationffThe prediction of the space-time feature distribution of high-resolution pollution is obtained as follows,
further, the step S5 includes,
s5.1: for the training of a high-resolution pollution space-time special distribution prediction model, firstly, the mean square error loss between a high-resolution pollution space-time special distribution prediction value and a real label is calculated,
and the weight of the network is optimized and adjusted by a gradient descent algorithm to obtain high-resolution pollution space-time special distribution prediction model parameters,
in the formula, epsilon is the learning rate,for the differentiation of the weight θ, n is the size of minimatch, D (i) is the sample in the training sample set, L (HRST (D (i); θ), xf(i) Is a loss function, xf(i) Is the corresponding label.
S5.2: and for the high-resolution pollution space-time distribution prediction model, the last 24-hour data of the data sample set D is used as a test set sample as the input of the trained high-resolution pollution space-time special distribution prediction model, and the high-resolution pollution prediction output on the corresponding 24-hour test set is obtained.
On the other hand, the invention also discloses a space-time prediction system for urban mobile source pollution, which comprises the following units,
the data processing unit is used for processing the historical mobile source pollution monitoring data and the external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
the standardization processing unit is used for carrying out standardization processing on the low-resolution mobile source pollution data, the external environment factor data and the high-resolution mobile source pollution data;
the sample set construction unit is used for dividing the standardized low-resolution historical mobile source pollution data into external environment factor data and high-resolution mobile source pollution data at the current moment according to the length of a specified time interval to construct a historical mobile source pollution data sample set;
the prediction model construction unit is used for constructing a high-resolution urban mobile source pollution space-time prediction model and comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
the model training unit is used for using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training and performing model testing by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
and the prediction unit is used for predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
According to the technical scheme, the high-resolution urban mobile source pollution space-time prediction method is realized by seeking for constructing the high-resolution mobile source pollution space-time prediction method and combining traffic network information with different spatial scales and external environment information. The deep residual error network structure can extract deep space-time distribution characteristics of a large-scale low-resolution pollution emission list, set distributed upsampling is carried out to construct different space-scale emission characteristic mapping relations, and the high-resolution mobile source emission list space-time distribution prediction is realized by adopting a deep embedded network to fuse external influence factors. Compared with the existing method, the spatial structure constraint of the emission list under different spatial scales can be effectively processed, and the high-precision emission list space-time distribution prediction under sparse monitoring is realized.
Compared with the prior art, the method can effectively process the space structure constraint of the emission list under different space scales, and can carry out space-time feature extraction and smoothing processing on the low-resolution emission list to realize high-precision and high-resolution emission list space-time distribution prediction under sparse monitoring.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a block diagram of a high resolution prediction model of spatio-temporal contamination distribution in accordance with the present invention;
FIG. 3 is a diagram of the results of the estimation of the spatial-temporal variation distribution of CO pollution (the left graph is predicted using the upsampling mechanism of the present invention; the right graph does not employ the upsampling mechanism);
FIG. 4 is a spatio-temporal variation distribution estimation of NOx pollution (left graph predicted using the upsampling method of the present invention; right graph not using the upsampling mechanism);
fig. 5 is a spatiotemporal variation distribution estimation of HC contamination (left graph predicted using the upsampling mechanism of the present invention; right graph not using the upsampling mechanism).
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention.
As shown in fig. 1, the high-resolution urban mobile source pollution space-time prediction method according to the embodiment includes the following steps:
the method comprises the following steps:
step S1: processing historical mobile source pollution monitoring data and external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
step S2: carrying out standardized processing on low-resolution mobile source pollution data, external environment factor data and high-resolution mobile source pollution data;
step S3: dividing the standardized low-resolution historical mobile source pollution data according to the length of a specified time interval, and combining the low-resolution historical mobile source pollution data with the external environmental factor data and the high-resolution mobile source pollution data at the current moment to construct a historical mobile source pollution data sample set;
step S4: constructing a high-resolution urban mobile source pollution space-time prediction model, which comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
step S5: using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training, and performing model test by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
step S6: and predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
According to an embodiment of the present invention, the step S1 includes:
s1.1: and respectively accumulating the pollution monitoring data in the appointed time interval of each mobile source pollution emission observation point to obtain the pollution monitoring data of the appointed time interval of each observation point. Respectively generating low-resolution moving source pollution data and high-resolution moving source pollution data by utilizing reverse distance weighted interpolation according to different spatial resolution scales;
s1.2: the external environmental factor data is obtained by collecting weather information and holiday information corresponding to the high-resolution moving source pollution data time.
According to an embodiment of the present invention, the step S2 includes:
s2.1: selecting a maximum pollution monitoring value of a specified time interval, and normalizing the ratio of the low-resolution and high-resolution mobile source pollution data of the specified time interval to the maximum pollution monitoring value to be 0, 1]Interval standardized low-resolution and high-resolution moving source pollution data XC,XF。
S2.2: selecting each maximum weather factor of the appointed time interval, and normalizing the ratio of each weather factor data of the appointed time interval to each maximum weather factor to [0, 1%]Normalized weather data E of intervalsweather. Processing the holiday information by adopting single-hot coding Eholiday. Combining the two to form a standardized external factor EC={Eweather,Eholiday}
According to an embodiment of the present invention, the step S3 includes:
s3.1: determining a high resolution sample label X to be predicted for each time step tF(t) and corresponding time external environmental factor EC(t);
S3.2: at each time step t, the low-resolution pollution sequence is divided into adjacent time segments according to the time sequence characteristic,The short-term cycle time segment and the long-term trend time segment are respectively recorded as c, p, s are three types of time slice intervals, respectively, lc,lp,lsAre three types of time slice lengths;
s3.3: a sample corresponding to time t is recorded asIs the input of the model, a total data sample set D is constructed for an acquirable time span T, while any time T is within the range of [ max { l [ ]c+1,p*lp+1,s*ls+1},min{T+c,T+p,T+s}]And taking the last 24 hours of data of the data sample set D as a test set and the others as a training set.
According to the embodiment of the present invention, the step S4 is to construct a high resolution urban mobile source pollution spatio-temporal prediction model (HRST), which includes a low resolution pollution spatio-temporal feature extraction module, an external influence factor fusion module, and a high resolution distributed upsampling feature mapping module, and includes:
s4.1, constructing a low-resolution pollution space-time feature extraction module which comprises a 2-dimensional convolution layer, a ReLU activation function layer, a residual error unit and a fusion layer.
The spatiotemporal feature extraction may be expressed as
Where Conv2D is a 2D convolution operation, l2DIs the number of 2D convolutional layers, respectively, the l-th layer residual error sheetThe element input and output unit, k belongs to { c, p, s }; f is a residual mapping function; thetalAre learning parameters.
S4.2: and constructing an external influence factor fusion module, including feature dimension transformation and feature fusion.
Will standardize weather data EweatherAnd holiday data E processed by single-hot codingholidayVia the full connection layer and remapped toThe same tensor shape size is expressed as
Then, the three extracted space-time change characteristics are fused by adopting a weight matrix, and external factors and pollution space-time characteristics are fused by adopting threshold activation, as shown in a formula (7)
Wherein, Wc,Wp,WsLow resolution contaminated spatio-temporal features of near time segments, short time period time segments and long time trend time segments, respectivelyThe weight matrix of (a) is determined,representing the Hadamard product, HfcIs a fused low-resolution contaminated spatio-temporal feature.
S4.3: and constructing a high-resolution distributed upsampling feature mapping module which comprises a nearest neighbor interpolation unit and a distributed upsampling unit.
Determining low-resolution data expansion multiple N (N is the nth power of 4) in a nearest neighbor interpolation unit according to the spatial scale of low-resolution and high-resolution polluted data, and fusing the fused low-resolution polluted space-time feature H by adopting nearest neighbor interpolation operationfcUpsampling to high resolution scale Hff。
Hff=NearestNeighborUp(Hfc,N) (3)
In a distributed upsampling unit, N sub-pixel blocks, convolutional layers and N are used2Standardization layer (N)2Norm) performs an upsampled reconstruction of the low resolution features. Each sub-pixel block (SubPixel) can enlarge the input characteristic size by 4 times, reduce the number of characteristic channels by 4 times, and reduce N2The normalization layer normalizes each channel feature map according to the spatial structure constraint relationship of low-resolution and high-resolution pollution data, and the distributed upsampling operation is specifically expressed as follows,
wherein Hfo(i,j)=Ho(i,j)/∑i′,j′Ho(i′,j′),HfoNamely high resolution pollution space-time characteristic probability distribution, and high resolution scale pollution data H by combining nearest neighbor interpolationffThe prediction of the space-time feature distribution of high-resolution pollution is obtained as follows,
according to an embodiment of the present invention, the step S5 includes training a high-resolution pollution spatio-temporal feature distribution prediction model and testing the prediction model:
s5.1: for the training of a high-resolution pollution space-time special distribution prediction model, firstly, the mean square error loss between a high-resolution pollution space-time special distribution prediction value and a real label is calculated,
and the weight of the network is optimized and adjusted by a gradient descent algorithm to obtain high-resolution pollution space-time special distribution prediction model parameters,
in the formula, epsilon is the learning rate,for the differentiation of the weight θ, n is the size of minimatch, D (i) is the sample in the training sample set, L (HRST (D (i); θ), xf(i) Is a loss function, xf(i) Is the corresponding label.
S5.2: and for the high-resolution pollution space-time distribution prediction model, the last 24-hour data of the data sample set D is used as a test set sample as the input of the trained high-resolution pollution space-time special distribution prediction model, and the high-resolution pollution prediction output on the corresponding 24-hour test set is obtained.
For example, fig. 3-5 are test results of the present invention, which respectively show the high-resolution spatio-temporal variation distribution estimation results of CO, NOx, and HC contamination, compared with the low-resolution prediction results without using the upsampling mechanism, and show that the spatio-temporal upsampling mechanism used in the present application has a good generalization capability in the multi-scale mobile-source contamination prediction task.
According to the technical scheme, the deep residual error network structure can extract deep space-time distribution characteristics of a large-scale low-resolution pollution emission list, the distributed upsampling is set to construct different space-scale emission characteristic mapping relations, and the deep embedded network is adopted to fuse external influence factors to realize high-resolution mobile source emission list space-time distribution prediction. Compared with the existing method, the spatial structure constraint of the emission list under different spatial scales can be effectively processed, and the high-precision emission list space-time distribution prediction under sparse monitoring is realized.
On the other hand, the embodiment of the invention also discloses an urban mobile source pollution space-time prediction system, which comprises the following units,
the data processing unit is used for processing the historical mobile source pollution monitoring data and the external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
the standardization processing unit is used for carrying out standardization processing on the low-resolution mobile source pollution data, the external environment factor data and the high-resolution mobile source pollution data;
the sample set construction unit is used for dividing the standardized low-resolution historical mobile source pollution data into external environment factor data and high-resolution mobile source pollution data at the current moment according to the length of a specified time interval to construct a historical mobile source pollution data sample set;
the prediction model construction unit is used for constructing a high-resolution urban mobile source pollution space-time prediction model and comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
the model training unit is used for using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training and performing model testing by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
and the prediction unit is used for predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
In a third aspect, the present invention also discloses a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method as described above.
It is understood that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, the example and the beneficial effects of the related contents can refer to the corresponding parts in the method.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (8)
1. A high-resolution urban mobile source pollution space-time prediction method is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
step S1: processing historical mobile source pollution monitoring data and external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
step S2: carrying out standardized processing on low-resolution mobile source pollution data, external environment factor data and high-resolution mobile source pollution data;
step S3: dividing the standardized low-resolution historical mobile source pollution data according to the length of a specified time interval, and combining the low-resolution historical mobile source pollution data with the external environmental factor data and the high-resolution mobile source pollution data at the current moment to construct a historical mobile source pollution data sample set;
step S4: constructing a high-resolution urban mobile source pollution space-time prediction model, which comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
step S5: using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training, and performing model test by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
step S6: and predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
2. The high-resolution urban mobile source pollution space-time prediction method according to claim 1, characterized in that: the step S1 includes:
s1.1: accumulating the pollution monitoring data in the appointed time interval of each mobile source pollution emission observation point respectively to obtain the pollution monitoring data of the appointed time interval of each observation point, and generating low-resolution mobile source pollution data and high-resolution mobile source pollution data by utilizing reverse distance weighted interpolation according to different spatial resolution scales;
s1.2: the external environmental factor data is obtained by collecting weather information and holiday information corresponding to the high-resolution moving source pollution data time.
3. The high-resolution urban mobile source pollution space-time prediction method according to claim 2, characterized in that:
the step S2 includes:
s2.1: selecting a maximum pollution monitoring value of a specified time interval, and normalizing the ratio of the low-resolution and high-resolution mobile source pollution data of the specified time interval to the maximum pollution monitoring value to be 0, 1]Interval standardized low-resolution and high-resolution moving source pollution data XC,XF;
S2.2: selecting each maximum weather factor of the appointed time interval, and normalizing the ratio of each weather factor data of the appointed time interval to each maximum weather factor to [0, 1%]Normalized weather data E of intervalsweatherFor holiday information, processing by single-hot coding EholidavCombining the two to form a standardized external factor EC={Eweather,Eholidav}。
4. The high-resolution urban mobile source pollution space-time prediction method according to claim 3, characterized in that: the step S3 includes:
s3.1: determining a high resolution sample label X to be predicted for each time step tF(t) and corresponding time external environmental factor EC(t);
S3.2: contaminating low resolution at each time step tThe sequence is divided into an adjacent time segment, a short-time period time segment and a long-time trend time segment according to time sequence characteristics, and the adjacent time segment, the short-time period time segment and the long-time trend time segment are respectively recorded as XpC(t)=XpCt-p*lp,...,XpCt-p,XsC(t)=[XsCt-s*ls,...,XsCt-s]C, p, s are three types of time slice intervals, lc,lp,lSAre three types of time slice lengths;
s3.3: a sample corresponding to time t is recorded as Is the input of the model, a total data sample set D is constructed for an acquirable time span T, while any time T is within the range of [ max { l [ ]c+1,p*lp+1,s*ls+1},min{T+c,T+p,T+s}]And taking the last 24 hours of data of the data sample set D as a test set and the others as a training set.
5. The high-resolution urban mobile source pollution space-time prediction method according to claim 4, wherein: the step S4 includes:
s4.1: constructing a low-resolution pollution space-time feature extraction module, wherein the module comprises a two-dimensional convolution layer, a ReLU activation function layer, a residual error unit and a fusion layer;
the spatio-temporal feature extraction is expressed as
Where Conv2D is a 2D convolution operation, l2DIs the number of 2D convolutional layers,the residual error unit of the l-th layer is an input unit and an output unit respectively, and k belongs to { c, p, s }; f is a residual mapping function; thetalIs a learning parameter;
s4.2: constructing an external influence factor fusion module, including feature dimension transformation and feature fusion;
will standardize weather data EweatherAnd holiday data E processed by single-hot codingholidayVia the full connection layer and remapped toThe same tensor shape size is expressed as
Then, the three extracted space-time change characteristics are fused by adopting a weight matrix, and external factors and pollution space-time characteristics are fused by adopting threshold activation, as shown in formula (2)
Wherein, Wc,Wp,WsLow resolution contaminated spatio-temporal features of near time segments, short time period time segments and long time trend time segments, respectivelyThe weight matrix of (a) is determined,representing the Hadamard product, HfcIs a fused low-resolution polluted spatio-temporal feature;
s4.3: constructing a high-resolution distributed upsampling feature mapping module which comprises a nearest neighbor interpolation unit and a distributed upsampling unit;
nearest neighbor interpolationDetermining low-resolution data expansion multiple N in the unit according to the space scale of low-resolution and high-resolution polluted data, wherein N is the nth power of 4, and fusing the low-resolution polluted space-time characteristics H by adopting nearest neighbor interpolation operationfcUpsampling to high resolution scale Hff;
Hff=NearestNeighborUp(Hfc,N) (3)
In a distributed upsampling unit, N sub-pixel blocks, convolutional layers and N are used2Normalization layer N2Performing up-sampling reconstruction on the low-resolution features by Norm;
each sub-pixel block SubPixel enlarges the input characteristic size by 4 times, reduces the number of characteristic channels by 4 times, and N2The normalization layer normalizes each channel feature map according to the spatial structure constraint relationship of low-resolution and high-resolution pollution data, and the distributed upsampling operation is specifically expressed as follows,
wherein Hfo(i,j)=Ho(i,j)/∑i′,j′Ho(i′,j′),HfoNamely high resolution pollution space-time characteristic probability distribution, and high resolution scale pollution data H by combining nearest neighbor interpolationffThe prediction of the space-time feature distribution of high-resolution pollution is obtained as follows,
6. the high-resolution urban mobile source pollution space-time prediction method according to claim 5, characterized in that: the step S5 includes the steps of,
s5.1: for the training of a high-resolution pollution space-time special distribution prediction model, firstly, the mean square error loss between a high-resolution pollution space-time special distribution prediction value and a real label is calculated,
and the weight of the network is optimized and adjusted by a gradient descent algorithm to obtain high-resolution pollution space-time special distribution prediction model parameters,
in the formula, epsilon is the learning rate,for the differentiation of the weight θ, n is the size of minimatch, D (i) is the sample in the training sample set, L (HRST (D (i); θ), xf(i) Is a loss function, xf(i) Is the corresponding label.
S5.2: and for the high-resolution pollution space-time distribution prediction model, the last 24-hour data of the data sample set D is used as a test set sample as the input of the trained high-resolution pollution space-time special distribution prediction model, and the high-resolution pollution prediction output on the corresponding 24-hour test set is obtained.
7. A city mobile source pollution space-time prediction system is characterized in that: comprises the following units of a first unit, a second unit,
the data processing unit is used for processing the historical mobile source pollution monitoring data and the external environment factor data into a low-resolution mobile source pollution and external environment factor observation sequence and a high-resolution mobile source pollution observation sequence according to a specified spatial resolution;
the standardization processing unit is used for carrying out standardization processing on the low-resolution mobile source pollution data, the external environment factor data and the high-resolution mobile source pollution data;
the sample set construction unit is used for dividing the standardized low-resolution historical mobile source pollution data into external environment factor data and high-resolution mobile source pollution data at the current moment according to the length of a specified time interval to construct a historical mobile source pollution data sample set;
the prediction model construction unit is used for constructing a high-resolution urban mobile source pollution space-time prediction model and comprises a low-resolution pollution space-time feature extraction module, an external influence factor fusion module and a high-resolution distributed upsampling feature mapping module;
the model training unit is used for using a training set of a moving source historical pollution data sample set for high-resolution urban moving source pollution space-time prediction model training and performing model testing by using a test set to obtain a high-resolution urban moving source pollution space-time prediction model with stable performance;
and the prediction unit is used for predicting the high-resolution moving source pollution space-time distribution in the future time period by using the trained high-resolution urban moving source pollution space-time prediction model.
8. A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to carry out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110110653.5A CN112819218B (en) | 2021-01-27 | 2021-01-27 | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110110653.5A CN112819218B (en) | 2021-01-27 | 2021-01-27 | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112819218A true CN112819218A (en) | 2021-05-18 |
CN112819218B CN112819218B (en) | 2022-05-06 |
Family
ID=75859641
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110110653.5A Active CN112819218B (en) | 2021-01-27 | 2021-01-27 | High-resolution urban mobile source pollution space-time prediction method, system and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112819218B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113344759A (en) * | 2021-06-30 | 2021-09-03 | 北京建筑大学 | Analysis method for mobile source pollution emission |
CN115619008A (en) * | 2022-09-23 | 2023-01-17 | 中科三清科技有限公司 | Air quality prediction method, device, equipment and computer readable storage medium |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102768738A (en) * | 2012-07-31 | 2012-11-07 | 华锐风电科技(集团)股份有限公司 | Method and device for power predication of wind power station |
CN105117493A (en) * | 2015-09-23 | 2015-12-02 | 环境保护部环境工程评估中心 | AERSURFACE integrated system based on high-resolution soil utilization data |
CN109376969A (en) * | 2018-12-14 | 2019-02-22 | 中南大学 | City fining population distribution dynamic prediction method and device based on deep learning |
CN109492830A (en) * | 2018-12-17 | 2019-03-19 | 杭州电子科技大学 | A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning |
CN109492822A (en) * | 2018-11-24 | 2019-03-19 | 上海师范大学 | Air pollutant concentration time-space domain interaction prediction method |
CN109862299A (en) * | 2017-11-30 | 2019-06-07 | 北京大学 | The processing method and processing device of resolution ratio |
CN110503139A (en) * | 2019-08-07 | 2019-11-26 | 中国科学技术大学 | A kind of urban area tail gas pollution prediction technique |
CN111461439A (en) * | 2020-04-02 | 2020-07-28 | 浙江大学 | Urban atmospheric diffusion process-based simulation prediction multi-heat-source heat supply system load scheduling distribution method and system |
CN111505738A (en) * | 2020-03-17 | 2020-08-07 | 上海眼控科技股份有限公司 | Method and equipment for predicting meteorological factors in numerical weather forecast |
CN111639748A (en) * | 2020-05-15 | 2020-09-08 | 武汉大学 | Watershed pollutant flux prediction method based on LSTM-BP space-time combination model |
CN111896680A (en) * | 2020-07-08 | 2020-11-06 | 天津师范大学 | Greenhouse gas emission analysis method and system based on satellite remote sensing data |
CN112132264A (en) * | 2020-09-11 | 2020-12-25 | 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) | Regional exhaust emission prediction method and system based on space-time residual perception network |
-
2021
- 2021-01-27 CN CN202110110653.5A patent/CN112819218B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102768738A (en) * | 2012-07-31 | 2012-11-07 | 华锐风电科技(集团)股份有限公司 | Method and device for power predication of wind power station |
CN105117493A (en) * | 2015-09-23 | 2015-12-02 | 环境保护部环境工程评估中心 | AERSURFACE integrated system based on high-resolution soil utilization data |
CN109862299A (en) * | 2017-11-30 | 2019-06-07 | 北京大学 | The processing method and processing device of resolution ratio |
CN109492822A (en) * | 2018-11-24 | 2019-03-19 | 上海师范大学 | Air pollutant concentration time-space domain interaction prediction method |
CN109376969A (en) * | 2018-12-14 | 2019-02-22 | 中南大学 | City fining population distribution dynamic prediction method and device based on deep learning |
CN109492830A (en) * | 2018-12-17 | 2019-03-19 | 杭州电子科技大学 | A kind of mobile pollution source concentration of emission prediction technique based on space-time deep learning |
CN110503139A (en) * | 2019-08-07 | 2019-11-26 | 中国科学技术大学 | A kind of urban area tail gas pollution prediction technique |
CN111505738A (en) * | 2020-03-17 | 2020-08-07 | 上海眼控科技股份有限公司 | Method and equipment for predicting meteorological factors in numerical weather forecast |
CN111461439A (en) * | 2020-04-02 | 2020-07-28 | 浙江大学 | Urban atmospheric diffusion process-based simulation prediction multi-heat-source heat supply system load scheduling distribution method and system |
CN111639748A (en) * | 2020-05-15 | 2020-09-08 | 武汉大学 | Watershed pollutant flux prediction method based on LSTM-BP space-time combination model |
CN111896680A (en) * | 2020-07-08 | 2020-11-06 | 天津师范大学 | Greenhouse gas emission analysis method and system based on satellite remote sensing data |
CN112132264A (en) * | 2020-09-11 | 2020-12-25 | 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) | Regional exhaust emission prediction method and system based on space-time residual perception network |
Non-Patent Citations (8)
Title |
---|
LEE M 等: "Spatiotemporal prediction of fine particulate matter using high-resolution satellite images in the Southeastern US 2003-2011", 《JOURNAL OF EXPOSURE SCIENCE & ENVIRONMENTAL EPIDEMIOLOGY》 * |
ZAMMIT-MANGION 等: "FRK:An R package for spatial and spatio-temporal prediction with large datasets", 《ARXIV PREPRINT ARXIV》 * |
刘庚等: "天津市2017年移动源高时空分辨率排放清单", 《环境科学》 * |
李大成等: "一种单时相高分辨率遥感影像时空融合算法", 《计算机工程与应用》 * |
毛红梅等: "成都市大气污染物排放清单高分辨率的时空分配", 《环境科学学报》 * |
程哲等: "使用深层卷积神经网络提高Hi-C数据分辨率", 《计算机科学》 * |
陆秋琴等: "陕西省VOCs人为源高分辨率排放清单及时空分布", 《福建师范大学学报(自然科学版)》 * |
马义超 等: "基于ENVI的高分辨率遥感影像城市生态环境研究", 《软件》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113344759A (en) * | 2021-06-30 | 2021-09-03 | 北京建筑大学 | Analysis method for mobile source pollution emission |
CN115619008A (en) * | 2022-09-23 | 2023-01-17 | 中科三清科技有限公司 | Air quality prediction method, device, equipment and computer readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN112819218B (en) | 2022-05-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kim et al. | Surface crack detection using deep learning with shallow CNN architecture for enhanced computation | |
CN113487066B (en) | Long-time-sequence freight volume prediction method based on multi-attribute enhanced graph convolution-Informer model | |
CN109859163A (en) | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks | |
CN112819218B (en) | High-resolution urban mobile source pollution space-time prediction method, system and storage medium | |
CN108986453A (en) | A kind of traffic movement prediction method based on contextual information, system and device | |
CN111915081B (en) | Peak sensitive travel demand prediction method based on deep learning | |
CN111428558A (en) | Vehicle detection method based on improved YO L Ov3 method | |
CN111275171A (en) | Small target detection method based on parameter sharing and multi-scale super-resolution reconstruction | |
CN113239753A (en) | Improved traffic sign detection and identification method based on YOLOv4 | |
CN110991447A (en) | Train number accurate positioning and identification method based on deep learning | |
Zhao et al. | Unsupervised anomaly detection based method of risk evaluation for road traffic accident | |
Lin et al. | Path travel time estimation using attribute-related hybrid trajectories network | |
Hou et al. | CANet: Contextual information and spatial attention based network for detecting small defects in manufacturing industry | |
Zhang et al. | Integrated APC-GAN and AttuNet framework for automated pavement crack pixel-level segmentation: A new solution to small training datasets | |
Bui et al. | Big data analytics-based urban traffic prediction using deep learning in ITS | |
Ramhormozi et al. | Multi-task graph neural network for truck speed prediction under extreme weather conditions | |
Joseph et al. | Investigation of deep learning methodologies in intelligent green transportation system | |
Rahman et al. | MDLpark: available parking prediction for smart parking through mobile deep learning | |
Filali et al. | Real-time parking availability classification on a large-area scale | |
Jia et al. | The hierarchical clustering of human mobility behaviors | |
CN113496305A (en) | Data processing method and device | |
Sun et al. | Alleviating data sparsity problems in estimated time of arrival via auxiliary metric learning | |
Xia et al. | An ASM-CF model for anomalous trajectory detection with mobile trajectory big data | |
Salinas et al. | Cityhub: A library for urban data integration | |
Kong et al. | A Visual Analytics Approach for Traffic Flow Prediction Ensembles. |
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 |