CN114861882A - CO (carbon monoxide) 2 Space-time distribution reconstruction method and system - Google Patents
CO (carbon monoxide) 2 Space-time distribution reconstruction method and system Download PDFInfo
- Publication number
- CN114861882A CN114861882A CN202210490951.6A CN202210490951A CN114861882A CN 114861882 A CN114861882 A CN 114861882A CN 202210490951 A CN202210490951 A CN 202210490951A CN 114861882 A CN114861882 A CN 114861882A
- Authority
- CN
- China
- Prior art keywords
- neural network
- deep neural
- network model
- data
- environment
- 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
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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/211—Schema design and management
- G06F16/212—Schema design and management with details for data modelling support
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses CO 2 The space-time distribution reconstruction method and system comprise the following steps: s1: establishing an environment database, wherein the environment database and a multi-output deep neural network model are established; s2: by means of NO 2 Initially training a multi-output deep neural network model by satellite remote sensing data and environment data; s3: by using CO 2 Carrying out secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the environment data; s4: CO pair using environmental data and trained multi-output deep neural network model 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution. The invention reconstructs NO with higher accuracy 2 Simultaneous generation of NO with high spatial and temporal resolution data set of satellite data 2 Represented by satellite dataThe information of fossil fuel combustion is given to the model to realize CO 2 Reconstruction of high spatial-temporal resolution spatial-temporal distributions.
Description
Technical Field
The invention relates to the technical field of environmental monitoring, in particular to CO 2 A space-time distribution reconstruction method and system.
Background
At present, CO is reconstituted 2 The machine learning method of space-time distribution mainly comprises A Neural Network (ANN), an extreme gradient propeller (XGBOST), an optical gradient propeller (Light-GBM) and the like. CO reconstruction Using machine learning 2 In the technical modeling of space-time distribution, the dependent variable is CO 2 Satellite remote sensing data, independent variable mainly includes: land use type, normalized vegetation index, weather conditions, population, altitude, road information, and CO 2 An emissions manifest, and the like. The prior art is on CO 2 Lack of support of relevant data information when spatial-temporal distribution is reconstructed, resulting in CO 2 A large deviation occurs in the reconstruction process; at the same time, due to CO 2 The satellite is easy to be interfered by external conditions when acquiring data, the amount of the acquired satellite remote sensing data sample is small, and sampling deviation exists, so that CO is generated 2 The satellite data scarce area has certain underestimation problem, and the method is suitable for CO in high vegetation coverage and high altitude industrial area 2 The concentration has a certain high value underestimation problem.
In response to the above problems, there have been some prior art on NO 2 Satellite remote sensing data (TROPOMI-NO) 2 ) Sampling and spatio-temporal interpolationValue and adding it as an independent variable to the model to perform CO 2 And (4) reconstructing the space-time distribution. However, TROPOMI-NO 2 The coverage and spatial-temporal resolution of the raw data are difficult to satisfy the CO 2 High resolution spatiotemporal distribution reconstruction; and reconstructed TROPOMI-NO by space-time kriging interpolation 2 The results are more biased and will also reduce CO 2 And (5) accuracy of a space-time reconstruction result.
In view of this, the present application is specifically made.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: CO by means of the prior art 2 The accuracy of the result obtained by space-time distribution reconstruction is lower, and the purpose is to provide a CO 2 The space-time distribution reconstruction method and system utilize satellite data of TROPOMI-NO2 as relevant information of fossil fuel combustion and utilize a transfer learning method based on shared parameters to realize CO (carbon monoxide) 2 And (5) reconstructing the spatial-temporal distribution of the concentration comprehensive domain.
The invention is realized by the following technical scheme:
in one aspect, the invention provides a CO 2 The space-time distribution reconstruction method comprises the following steps:
s1: establishing an environment database, the environment database comprising: TROPOMI-NO 2 Satellite remote sensing data, CO 2 Satellite remote sensing data and base data related to the environment;
s2: establishing a multi-output deep neural network model;
s3: using said TROPOMI-NO 2 Initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment;
s4: utilizing the CO 2 Performing secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment;
s5: utilizing the basic data related to the environment and the multi-output deep neural network model after the secondary training to the CO 2 Predicting the space-time distribution to obtain CO 2 Space-time divisionAnd (5) reconstructing the result.
As a further description of the present invention,
the context-related base data includes: CO2 2 Emission data, population density data, altitude elevation data, land use data, normalized vegetation index and meteorological data; the meteorological data includes: surface temperature, surface air pressure, wind speed, wind direction, relative humidity, and planet boundary layer height.
As a further description of the present invention, said S2 includes before:
s11: carrying out 1km gridding processing and standardization processing on all data in the environment database;
s12: subjecting the fraction treated in S11 to TROPOMI-NO treatment 2 Dividing grids of the satellite remote sensing data into an initial training set and an initial testing set, and processing the grid containing CO through S11 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
As a further description of the present invention, the S2 includes the following steps:
establishing a first deep neural network model and a second deep neural network model, wherein the first deep neural network model and the second deep neural network model have the same model structure, and the method comprises the following steps: a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer;
establishing a corresponding data transmission chain between each layer of the first deep neural network model and the second deep neural network model, wherein the direction of the data transmission chain is from the first deep neural network model to the second deep neural network model;
connecting the input end of the first deep neural network model and the input end of the second deep neural network model with the output end of a self-encoder.
As a further description of the present invention, the S3 includes the following steps:
s31: inputting the initial training set serving as a dependent variable into the multi-output deep neural network model by taking the basic data related to the environment as independent variables;
s32: performing deep learning on the basic data related to the environment and the initial training set by using the first deep neural network model, and fitting the basic data related to the environment and the TROPOMI-NO 2 Initial nonlinear relation between satellite remote sensing data, and sending the fitting result of each layer of the first deep neural network to the corresponding layer of the second deep neural network;
s33: performing deep learning on the environment-related basic data, the initial training set and the data from the first deep neural network model by using the second deep neural network model, and fitting the environment-related basic data and the TROPOMI-NO 2 And (4) final nonlinear relation between satellite remote sensing data.
As a further description of the present invention,
the S32 previously includes the following steps:
performing dimensionality reduction on the environment-related basic data and the initial training set by using the self-encoder;
performing cross validation on the initial training set;
the step S33 includes the following steps:
and testing the multi-output deep neural network model after the initial training by using the initial test set.
As a further description of the present invention, the S4 includes:
s41: inputting the basic data related to the environment as independent variables and the secondary training set as dependent variables into the multi-output deep neural network model;
s42: performing deep learning on the environment-related basic data and the secondary training set only by using the second deep neural network model trained in the S33, and fitting the environment-related basic data and the CO 2 A non-linear relationship between the satellite telemetry data.
As a further description of the present invention,
before S42, the method includes the following steps: performing dimensionality reduction on the environment-related basic data and the secondary training set by using the self-encoder;
after the step S42, the method includes the following steps: and testing the multi-output deep neural network model after the secondary training by using the secondary test set.
In another aspect, the present invention provides a CO 2 A spatio-temporal distribution reconstruction system, comprising:
a database creation module for creating a database containing TROPOMI-NO 2 Satellite remote sensing data, CO 2 An environment database of satellite remote sensing data and base data related to the environment;
the model establishing module is used for establishing a multi-output deep neural network model;
an initial training module for utilizing the TROPOMI-NO 2 Initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment;
a secondary training module for utilizing the CO 2 Performing secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment;
a model prediction module for using the environment-related basic data and the secondarily trained multi-output deep neural network model to predict CO 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution.
As a further description of the present invention, the system further comprises:
a data processing module for performing 1km gridding processing and standardization processing on all data in the environment database and processing the processed data containing CO 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
The model creation module includes:
the first model creating unit is used for creating a first deep neural network model comprising a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer, and connecting the input end of the first deep neural network model with the output end of the self-encoder;
a second model creating unit for creating a second deep neural network model having the same model structure as the first deep neural network model and connecting an input end of the second deep neural network model with an output end of the self-encoder;
and the data transmission link creation unit is used for creating a corresponding data transmission link between each layer of the first deep neural network model and the second deep neural network model, and the direction of the data transmission link is from the first deep neural network model to the second deep neural network model.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. one aspect of the invention utilizes TROPOMI-NO 2 The satellite remote sensing data has the advantages of large quantity and can be used as related information of fossil fuel combustion, and the satellite remote sensing data is introduced into CO 2 In the space-time distribution reconstruction, human activities are reflected laterally, and the problem of CO existing in the prior art can be solved 2 Lack of support of relevant data information and CO when spatial-temporal distribution is reconstructed 2 The problem that the reconstruction result has larger deviation due to small sample amount of satellite remote sensing data acquired by a satellite; on the other hand, a multi-output deep neural network model is established, and initial training and secondary training are carried out on the model in sequence, and TROPOMI-NO is utilized 2 Satellite data reflecting fossil fuel combustion information and NO 2 With CO 2 Emitting a homologous character, TROPOMI-NO 2 The represented fossil fuel combustion information is endowed to the model to realize CO 2 Reconstruction of the spatial-temporal distribution at high spatial-temporal resolution.
2. The invention carries out dimension reduction processing on the initial independent variable dimension through the self-encoder part, thereby improving the learnability of data;
3. the invention effectively combines TROPOMI-NO2 and CO2 related information on space-time distribution by using a transfer learning method, and TROPOMI-NO 2 Data as auxiliary data, solveCO 2 Data sparseness problem and CO increase 2 Accuracy of concentration prediction results.
Drawings
In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are required in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and that those skilled in the art may also derive other related drawings based on these drawings without inventive effort.
FIG. 1 shows a CO according to an embodiment of the present invention 2 A flow diagram of a space-time distribution reconstruction method;
FIG. 2 is a schematic diagram of a model architecture of a multiple-output deep neural network according to an embodiment of the present invention;
FIG. 3 shows a CO according to an embodiment of the present invention 2 And the space-time distribution reconstructs a system structural diagram.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not meant to limit the present invention.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that: it is not necessary to employ these specific details to practice the present invention. In other instances, well-known structures, circuits, materials, or methods have not been described in detail so as not to obscure the present invention.
Throughout the specification, reference to "one embodiment," "an embodiment," "one example," or "an example" means: the particular features, structures, or characteristics described in connection with the embodiment or example are included in at least one embodiment of the invention. Thus, the appearances of the phrases "one embodiment," "an embodiment," "one example" or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combination and/or sub-combination in one or more embodiments or examples. Further, those of ordinary skill in the art will appreciate that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
In the description of the present invention, the terms "front", "rear", "left", "right", "upper", "lower", "vertical", "horizontal", "upper", "lower", "inner", "outer", etc. indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplicity of description, but do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed in a particular orientation, and be operated, and therefore, should not be construed as limiting the scope of the present invention.
Example 1
Due to the prior art on CO 2 Lack of support of relevant data information when spatio-temporal distributions are reconstructed, and CO 2 The satellite is easily interfered by external conditions when acquiring data, resulting in CO 2 The problem of large deviation in the reconstruction process, this embodiment provides a CO 2 The space-time distribution reconstruction method has the method flow as shown in figure 1, and utilizes TROPOMI-NO 2 The satellite data is used as the relevant information of fossil fuel combustion, and the CO is migrated and learned based on shared parameters 2 And (5) reconstructing the spatial-temporal distribution of the concentration comprehensive domain. The main modeling process is TROPOMI-NO 2 Pre-training the deep learning model as a dependent variable to fit the model to the independent variable and TROPOMI-NO 2 Non-linear relationship between them, and further with CO 2 And finally training the model after pre-training as a dependent variable. By the method, the TROPOMI-NO with higher accuracy is reconstructed by the model 2 Simultaneous high spatial and temporal resolution of the data set with TROPOMI-NO 2 Information of the represented fossil fuel combustion is endowed to the model to realize CO 2 Reconstruction of high spatial-temporal resolution spatial-temporal distributions. The implementation steps are as follows:
s1: establishing an environment database, the environment database comprising: TROPOMI-NO 2 Satellite remote sensing data, CO 2 Satellite remote sensing data and environment-related base data. Wherein the context-related base data includes: CO2 2 Emission data, population density data, altitude elevation data, land use data, normalized vegetation index and meteorological data; the meteorological data includes: surface temperature, surface air pressure, wind speed, wind direction, relative humidity, and planet boundary layer height.
Using TROPOMI-NO 2 The satellite remote sensing data has the advantages of large quantity and can be used as related information of fossil fuel combustion, and the satellite remote sensing data is introduced into CO 2 In the space-time distribution reconstruction, human activities are reflected laterally, and the problem of CO existing in the prior art can be solved 2 Lack of support of relevant data information and CO when spatial-temporal distribution is reconstructed 2 The problem of large deviation of a reconstruction result caused by small sample amount of satellite remote sensing data acquired by a satellite is solved, and accurate full coverage of CO is realized 2 The concentration prediction result can provide basis for carbon emission statistical accounting and provide data support for carbon reduction policy establishment.
S2: and performing 1km gridding processing and standardization processing on all data in the environment database. The method specifically comprises the following steps:
s21: adopting an area weighted average method for population density data, altitude elevation data, land utilization data and normalized vegetation index variables to convert the spatial resolution into a 1km grid;
s22: and (4) resampling the CO2 emission list and meteorological data to a 1km grid by adopting elevation Cokriging interpolation.
S23: after the data are processed to the same spatial scale, the data are respectively standardized, so that the independent variables are in the same data scale, and the training of a deep learning model is facilitated.
S24: for TROPOMI-NO 2 Satellite remote sensing data and CO 2 Carrying out 1km gridding processing on the satellite remote sensing data, and carrying out TROPOMI-containing processing after the steps (1) to (3) are spoken-NO 2 Dividing grids of the satellite remote sensing data into an initial training set and an initial testing set; the CO-containing substances treated in the steps (1) to (3) are introduced 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
S3: and establishing a multi-output deep neural network model. The method comprises the following steps:
s31: establishing a first deep neural network model and a second deep neural network model, wherein the first deep neural network model and the second deep neural network model have the same model structure, and the method comprises the following steps: a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer;
s32: establishing a corresponding data transmission chain between each layer of the first deep neural network model and the second deep neural network model, wherein the direction of the data transmission chain is from the first deep neural network model to the second deep neural network model;
s33: connecting the input end of the first deep neural network model and the input end of the second deep neural network model with the output end of a self-encoder.
The deep learning model used in the method is a multi-output deep neural network based on an auto-encoder, and as shown in fig. 2, the model is composed of three modules including an auto-encoder and two sub-deep learning modules.
The model utilizes an autoencoder to reduce the dimension of the independent variable quantity, and the first deep neural network model and the second deep neural network model are respectively constructed by utilizing the reduced-dimension variables. The fully-connected layer in the first deep neural network model and the second deep neural network model comprises a most basic Dense layer of a neural network, and a Batch Normalization, an activation function and a Dropout are used. And the two deep neural network models are connected in a one-way mode, namely, the output of each layer of the first deep neural network model is added with the output of each layer of the second deep neural network model to serve as the input data of the next layer of the second deep neural network model.
S34: the invention optimizes the model by using a joint loss function, the loss function corresponding to each output is based on Mean Square Error (MSE), and the specific function expression is as follows:
outputting one:
Loss min ＝min(Loss 2 ，Loss 3 )
outputting two:
outputting three:
and (4) integral output: loss all ＝λ min Loss 1 +(1-λ)Loss 2 +λLoss 3
λ min ＝min(λ，1-λ)
Wherein:
respectively outputting a first model predicted value, a second model predicted value and a third model predicted value; y is 1 、y 2 、y 3 Respectively outputting true values corresponding to the first output, the second output and the third output; lambda, (1-lambda) and lambda min The range is (0, 1) for the weight applied to each loss function, and ω is the weight value that outputs a final layer of neural network nodes. The emphasis point of model training is set by adjusting the value of lambda, and when lambda is less than 0.5, the Loss in the model 2 The weight occupied in the training process is larger; loss when λ > 0.5 3 And the weight of the training process is more. Using lambda min For Loss 1 Is adjusted to lambda min Is a small value between lambda and 1-lambda and is in Loss 1 Intermediate binding Loss 2 And Loss 3 Less value of between Loss min . When Loss min At larger values, Loss 1 Has higher weight in the gradient descending process, thereby enabling the self-encoder module to quickly approach to the global optimumA value; when Loss min Gradually decreasing value, Loss 1 The weighted value is correspondingly reduced, so that the internal parameter change of the self-encoder formed by the input data and the output I is smaller, a more stable independent variable dimension reduction result is provided for the sub-deep learning module I and the sub-deep learning module II, and meanwhile, the two norm value of omega is added to apply L2 regularization to the output I result, so that the self-encoder structure in the model has stronger generalization capability. The invention utilizes the Loss function Loss of the whole all And optimizing the model to update the internal parameters of the model.
S4: using said TROPOMI-NO 2 And initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment. The method comprises the following steps:
s41: inputting the initial training set serving as a dependent variable into the multi-output deep neural network model by taking the basic data related to the environment as independent variables;
s42: performing dimensionality reduction on the environment-related basic data and the initial training set by using the self-encoder;
s43: performing deep learning on the environment-related basic data and the initial training set by using the first deep neural network model, and fitting the environment-related basic data and the TROPOMI-NO 2 Initial nonlinear relation between satellite remote sensing data, and sending the fitting result of each layer of the first deep neural network to the corresponding layer of the second deep neural network;
s44: performing deep learning on the environment-related basic data, the initial training set and the data from the first deep neural network model by using the second deep neural network model, and fitting the environment-related basic data and the TROPOMI-NO 2 A final nonlinear relationship between the satellite remote sensing data;
s45: and testing the multi-output deep neural network model after the initial training by using the initial test set.
Has better generalization for modelChemical capacity, in this example TROPOMI-NO will be included 2 The method comprises the following steps that grids of satellite remote sensing data are divided into an initial training set and an initial testing set, and cross validation is adopted aiming at the initial training set and used for determining the optimal parameters of a model; the initial test set is only used for evaluating the generalization capability of the model and does not participate in the pre-training process of the model.
In the initial training, the CO2 emission list, population density data, elevation data, land utilization data, normalized vegetation index and meteorological data are used as input data of the model. The output one is input data, and the outputs of the first deep neural network model and the second deep neural network model are both TROPOMI-NO 2 Remote sensing of data from satellites, thereby using TROPOMI-NO 2 And the initial optimization of parameters in the model is completed due to the advantage of more data.
Through initial training of the model, two different environment-related integrated data (independent variables for short) and TROPOMI-NO are constructed by the first deep neural network model and the second deep neural network model respectively 2 A non-linear relationship between the satellite telemetry data. The decoder part and the first deep neural network model part in the model reflect independent variables and TROPOMI-NO 2 The second deep neural network model is combined with TROPOMI-NO in the first deep neural network model 2 Construction of independent variable and TROPOMI-NO by using relevant information of satellite remote sensing data 2 And (4) nonlinear relation of satellite remote sensing data.
In addition, the dimension of the initial independent variable dimension is reduced through the self-encoder part, and the data learnability is improved.
Independent variable and TROPOMI-NO 2 The relation between the satellite remote sensing data is as follows: y is TROPOMI-NO2 ＝f(x 1 ，x 2 …，x n )。
S5: utilizing the CO 2 And carrying out secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment. The method comprises the following steps:
s51: inputting the basic data related to the environment as independent variables and the secondary training set as dependent variables into the multi-output deep neural network model;
s52: performing dimensionality reduction on the environment-related basic data and the secondary training set by using the self-encoder;
s53: performing deep learning on the environment-related basic data and the secondary training set only by using the second deep neural network model trained in the S33, and fitting the environment-related basic data and the CO 2 A non-linear relationship between the satellite remote sensing data;
s54: and testing the multi-output deep neural network model after the secondary training by using the secondary test set.
This example will contain CO 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set, the secondary training set is used for determining the optimal parameters of the model, and the secondary testing set is only used for evaluating the generalization capability of the model.
When the model is trained secondarily, parameters of an autocoder and a first deep neural network model in the model are frozen, so that the parameters of the autocoder and the first deep neural network model are not changed in the secondary training process. The variable type of the input data of the model is the same as that of the initial training, and only the second deep neural network model participates in the training to construct independent variable and CO 2 A non-linear relationship therebetween.
Because the first deep neural network model can provide TROPOMI-NO during the initial training process 2 The satellite remote sensing data related information, so the second deep neural network model can be combined with TROPOMI-NO 2 Realization of satellite remote sensing data on CO 2 The prediction is effective. The second depth is the independent variable and CO constructed by the neural network model 2 Relationship between satellite remote sensing data: y is CO2 ＝f(x 1 ，x 2 …，x n ，Y TROPOMI-NO2 )。
S6: utilizing the context-dependent base data and passing through the twoMulti-output deep neural network model pair CO after sub-training 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution.
After the initial training and the secondary training of the model are finished, the independent variables in the region range are input into the trained multi-output deep neural network model, and the TROPOMI-NO with the spatial resolution of 1km and the time resolution of day is output through the first deep neural network model 2 The reconstructed result of the space-time distribution outputs CO with the spatial resolution of 1km and the time resolution of days through a second deep neural network model 2 Reconstructing the result of space-time distribution to achieve TROPOMI-NO in the region 2 And CO 2 And (5) reconstructing high-resolution space-time distribution.
In summary, the present embodiment provides a CO 2 Spatio-temporal distribution reconstruction method, using NO in the absence of data directly related to human activity 2 Satellite remote sensing data TROPOMI-NO 2 Laterally reflect human activities and reconstruct CO 2 High resolution spatial and temporal distribution, accurate full coverage of CO 2 The concentration prediction result can provide basis for carbon emission statistical accounting and provide data support for carbon reduction policy establishment. In addition, the dimension reduction is carried out on the initial independent variable dimension through the self-encoder part in the method, and the data learnability is improved. Using transfer learning method to convert TROPOMI-NO 2 And CO 2 Effectively combining related information on space-time distribution to obtain TROPOMI-NO 2 Data as auxiliary data to increase CO 2 Accuracy of concentration prediction results.
Example 2
This example provides a CO as described in example 1 2 The system corresponding to the space-time distribution reconstruction method, as shown in fig. 3, comprises:
a database creation module for creating a database containing TROPOMI-NO 2 Satellite remote sensing data, CO 2 An environment database of satellite remote sensing data and base data related to the environment;
the model establishing module is used for establishing a multi-output deep neural network model;
an initial training module for training the training device,for using said TROPOMI-NO 2 Initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment;
a secondary training module for utilizing the CO 2 Performing secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment;
a model prediction module for using the environment-related basic data and the secondarily trained multi-output deep neural network model to predict CO 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution.
A data processing module for performing 1km gridding processing and standardization processing on all data in the environment database and processing the processed data containing CO 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
Wherein the model creation module comprises:
the first model creating unit is used for creating a first deep neural network model comprising a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer, and connecting the input end of the first deep neural network model with the output end of the self-encoder;
a second model creating unit for creating a second deep neural network model having the same model structure as the first deep neural network model and connecting an input end of the second deep neural network model with an output end of the self-encoder;
a data transmission link creation unit, configured to create a data transmission link corresponding to each layer of the first deep neural network model and the second deep neural network model, where the direction of the data transmission link is from the first deep neural network model to the second deep neural network model
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. CO (carbon monoxide) 2 The spatial-temporal distribution reconstruction method is characterized by comprising the following steps of:
s1: establishing an environment database, the environment database comprising: TROPOMI-NO 2 Satellite remote sensing data, CO 2 Satellite remote sensing data and environment-related basic data;
s2: establishing a multi-output deep neural network model;
s3: using said TROPOMI-NO 2 Initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment;
s4: utilizing the CO 2 Performing secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment;
s5: utilizing the basic data related to the environment and the multi-output deep neural network model after the secondary training to the CO 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution.
2. CO according to claim 1 2 A method for reconstructing a spatial-temporal distribution, characterized in that,
the context-related base data includes: CO2 2 Emission data, population density data, altitude elevation data, land use data, normalized vegetation index and meteorological data; the meteorological data includes: surface temperature, surface air pressure, wind speed, wind direction, relative humidity, and planet boundary layer height.
3. CO according to claim 2 2 The spatio-temporal distribution reconstruction method, wherein S2 is preceded by:
s11: carrying out 1km gridding processing and standardization processing on all data in the environment database;
s12: subjecting the fraction treated in S11 to TROPOMI-NO treatment 2 Dividing the grid of the satellite remote sensing data into an initial training set and an initial testing set, and processing the grid containing CO after S11 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
4. CO according to claim 3 2 The method for reconstructing spatio-temporal distribution, wherein S2 comprises the steps of:
establishing a first deep neural network model and a second deep neural network model, wherein the first deep neural network model and the second deep neural network model have the same model structure, and the method comprises the following steps: a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer;
establishing a corresponding data transmission chain between each layer of the first deep neural network model and the second deep neural network model, wherein the direction of the data transmission chain is from the first deep neural network model to the second deep neural network model;
connecting the input end of the first deep neural network model and the input end of the second deep neural network model with the output end of a self-encoder.
5. CO according to claim 4 2 The method for reconstructing spatio-temporal distribution, wherein S3 comprises the steps of:
s31: inputting the initial training set serving as a dependent variable into the multi-output deep neural network model by taking the basic data related to the environment as independent variables;
s32: performing deep learning on the basic data related to the environment and the initial training set by using the first deep neural network model, and fitting the basic data related to the environment and the TROPOMI-NO 2 Initial non-linear relationship between satellite remote sensing data and the first deep neural networkThe fitting result of each layer is sent to the corresponding layer of the second deep neural network;
s33: performing deep learning on the environment-related basic data, the initial training set and the data from the first deep neural network model by using the second deep neural network model, and fitting the environment-related basic data and the TROPOMI-NO 2 And (4) final nonlinear relation between satellite remote sensing data.
6. CO according to claim 5 2 A method for reconstructing a spatial-temporal distribution, characterized in that,
the S32 previously includes the following steps:
performing dimensionality reduction on the environment-related basic data and the initial training set by using the self-encoder;
performing cross validation on the initial training set;
the step S33 includes the following steps:
and testing the multi-output deep neural network model after the initial training by using the initial test set.
7. A CO according to claim 5 or 6 2 The method for reconstructing spatio-temporal distribution, wherein S4 includes:
s41: inputting the basic data related to the environment as independent variables and the secondary training set as dependent variables into the multi-output deep neural network model;
s42: performing deep learning on the environment-related basic data and the secondary training set only by using the second deep neural network model trained in the S33, and fitting the environment-related basic data and the CO 2 A non-linear relationship between the satellite telemetry data.
8. CO according to claim 7 2 A method for reconstructing a spatial-temporal distribution, characterized in that,
before S42, the method includes the following steps: performing dimensionality reduction on the environment-related basic data and the secondary training set by using the self-encoder;
after the step S42, the method includes the following steps: and testing the multi-output deep neural network model after the secondary training by using the secondary test set.
9. CO (carbon monoxide) 2 A system for spatio-temporal distribution reconstruction, comprising:
a database creation module for creating a database containing TROPOMI-NO 2 Satellite remote sensing data, CO 2 An environment database of satellite remote sensing data and base data related to the environment;
the model establishing module is used for establishing a multi-output deep neural network model;
an initial training module for utilizing the TROPOMI-NO 2 Initially training the multi-output deep neural network model by using satellite remote sensing data and the basic data related to the environment;
a secondary training module for utilizing the CO 2 Performing secondary training on the multi-output deep neural network model after the initial training by using the satellite remote sensing data and the basic data related to the environment;
a model prediction module for using the environment-related basic data and the secondarily trained multi-output deep neural network model to predict CO 2 Predicting the space-time distribution to obtain CO 2 And reconstructing a result by space-time distribution.
10. CO according to claim 9 2 A system for spatio-temporal distribution reconstruction, the system further comprising:
a data processing module for performing 1km gridding processing and standardization processing on all data in the environment database and processing the processed data containing CO 2 The grid of the satellite remote sensing data is divided into a secondary training set and a secondary testing set.
The model creation module includes:
the first model creating unit is used for creating a first deep neural network model comprising a Dense layer, a Batch Normalization layer, an activation function and a Dropout layer, and connecting the input end of the first deep neural network model with the output end of the self-encoder;
a second model creating unit for creating a second deep neural network model having the same model structure as the first deep neural network model and connecting an input end of the second deep neural network model with an output end of the self-encoder;
and the data transmission link creation unit is used for creating a corresponding data transmission link between each layer of the first deep neural network model and the second deep neural network model, and the direction of the data transmission link is from the first deep neural network model to the second deep neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490951.6A CN114861882B (en) | 2022-05-07 | 2022-05-07 | CO (carbon monoxide) 2 Space-time distribution reconstruction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490951.6A CN114861882B (en) | 2022-05-07 | 2022-05-07 | CO (carbon monoxide) 2 Space-time distribution reconstruction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114861882A true CN114861882A (en) | 2022-08-05 |
| CN114861882B CN114861882B (en) | 2023-05-09 |
Family
ID=82636115
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210490951.6A Active CN114861882B (en) | 2022-05-07 | 2022-05-07 | CO (carbon monoxide) 2 Space-time distribution reconstruction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114861882B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115310550A (en) * | 2022-08-16 | 2022-11-08 | 国网四川省电力公司电力科学研究院 | Method and system for calculating concentration of atmospheric carbon dioxide dry air column |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529081A (en) * | 2016-12-03 | 2017-03-22 | 安徽新华学院 | PM2.5 real-time level prediction method and system based on neural net |
| US20210209424A1 (en) * | 2020-01-06 | 2021-07-08 | Quantela Inc | Computer-based method and system for predicting and generating land use land cover (lulc) classification |
| CN113128780A (en) * | 2021-04-29 | 2021-07-16 | 平安普惠企业管理有限公司 | PM2.5 concentration prediction method and device, computer equipment and storage medium |
| CN113297528A (en) * | 2021-06-10 | 2021-08-24 | 四川大学 | NO based on multi-source big data2High-resolution space-time distribution calculation method |
| CN114266200A (en) * | 2022-02-24 | 2022-04-01 | 山东大学 | Nitrogen dioxide concentration prediction method and system |
-
2022
- 2022-05-07 CN CN202210490951.6A patent/CN114861882B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529081A (en) * | 2016-12-03 | 2017-03-22 | 安徽新华学院 | PM2.5 real-time level prediction method and system based on neural net |
| US20210209424A1 (en) * | 2020-01-06 | 2021-07-08 | Quantela Inc | Computer-based method and system for predicting and generating land use land cover (lulc) classification |
| CN113128780A (en) * | 2021-04-29 | 2021-07-16 | 平安普惠企业管理有限公司 | PM2.5 concentration prediction method and device, computer equipment and storage medium |
| CN113297528A (en) * | 2021-06-10 | 2021-08-24 | 四川大学 | NO based on multi-source big data2High-resolution space-time distribution calculation method |
| CN114266200A (en) * | 2022-02-24 | 2022-04-01 | 山东大学 | Nitrogen dioxide concentration prediction method and system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115310550A (en) * | 2022-08-16 | 2022-11-08 | 国网四川省电力公司电力科学研究院 | Method and system for calculating concentration of atmospheric carbon dioxide dry air column |
| CN115310550B (en) * | 2022-08-16 | 2023-07-14 | 国网四川省电力公司电力科学研究院 | Atmospheric carbon dioxide dry air column concentration calculation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114861882B (en) | 2023-05-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Short-term rainfall forecasting using multi-layer perceptron | |
| CN111859800B (en) | Space-time estimation and prediction method for PM2.5 concentration distribution | |
| US11452286B2 (en) | Method of predicting central fishing ground of flying squid family ommastrephidae | |
| CN112561058B (en) | Short-term photovoltaic power prediction method based on Stacking-integrated learning | |
| CN111210082A (en) | Optimized BP neural network algorithm-based precipitation prediction method | |
| Usharani | ILF-LSTM: Enhanced loss function in LSTM to predict the sea surface temperature | |
| CN114239422B (en) | Method for improving marine chlorophyll a concentration prediction accuracy based on machine learning | |
| CN114707688A (en) | Photovoltaic power ultra-short-term prediction method based on satellite cloud chart and space-time neural network | |
| CN116933949B (en) | Water quality prediction method and system integrating hydrodynamic model and numerical model | |
| CN114861882A (en) | CO (carbon monoxide) 2 Space-time distribution reconstruction method and system | |
| CN116523187A (en) | Engineering progress monitoring method and system based on BIM | |
| CN115099497A (en) | CNN-LSTM-based real-time flood forecasting intelligent method | |
| CN116029419A (en) | Deep learning-based long-term new energy daily average generation power prediction method and system | |
| Li et al. | Stochastic deep gaussian processes over graphs | |
| CN114462683A (en) | Cloud edge collaborative multi-residential area load prediction method based on federal learning | |
| Zhang et al. | All‐factor short‐term photovoltaic output power forecast | |
| Zhang et al. | SolarGAN: Synthetic annual solar irradiance time series on urban building facades via Deep Generative Networks | |
| Huang et al. | Short-term PV power forecasting based on CEEMDAN and ensemble DeepTCN | |
| CN112532429B (en) | Multivariable QoS prediction method based on position information | |
| CN109215743A (en) | Prediction method of biogas production process based on novel BP neural network modeling | |
| CN117849903A (en) | Global ocean environment forecasting method based on multi-level feature aggregation | |
| CN117114168A (en) | Long-time-scale sea surface temperature intelligent forecasting method | |
| CN112183721A (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
| CN116843069A (en) | Commuting flow estimation method and system based on crowd activity intensity characteristics | |
| Ye et al. | Efficient calibration of agent-based traffic simulation using variational auto-encoder |
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 |