Method for siting co2 monitoring stations

Abstract

The application provides a CO2 monitoring station site selection method, which comprises the following steps: dividing a target area into multiple grids; marking the grids based on existing environmental data information to form grids in a marked state, wherein the environmental data information comprises CO2 concentration and other factors, the other factors comprising CO2 emission of a carbon source in the grid, CO2 absorption capacity of a carbon sink in the grid, meteorological conditions and the month; inferring CO2 concentration of all grids in the target area based on the grids in the marked state; determining the number of CO2 monitoring stations to be built in the target area; obtaining satellite remote sensing CO2 monitoring data corresponding to the grids in the target area; and evaluating the satellite remote sensing CO2 monitoring data, the number of CO2 monitoring stations and the CO2 concentration of all grids in the target area to obtain an optimal site selection scheme. The method can ensure the reliability of the site selection scheme while controlling the cost.

Description

Technical Field

[0001] This invention relates to the field of CO2 monitoring, and specifically to a method for selecting the location of a CO2 monitoring station. Background Technology

[0002] Currently, to better understand carbon emissions in different regions, cities are increasingly looking to build ground-based CO2 monitoring stations. Since the number of CO2 monitoring stations that can be installed in a given area is limited, generally not exceeding a dozen, the selection of station locations is crucial. A good site selection strategy can lead to more accurate estimates of CO2 concentration distribution, and consequently, more accurate estimates of total carbon emissions for the entire region.

[0003] Although there are currently atmospheric environment monitoring stations (used to monitor PM2.5) 2.5 Patents related to location selection algorithms (etc.) exist, but the technical problem solved by existing patents is how to build several new sites when some sites have already been built. Summary of the Invention

[0004] The purpose of this invention is to provide a method for selecting the location of CO2 monitoring stations, which is based on satellite remote sensing data and solves one or more of the above-mentioned problems.

[0005] This invention proposes a method for selecting the location of a CO2 monitoring station, comprising the following steps: Divide the target area into multiple grids; The grid is labeled based on existing environmental data to form a labeled grid. The environmental data includes CO2 concentration and other factors, including CO2 emissions from carbon sources within the grid, CO2 absorption capacity of carbon sinks within the grid, meteorological conditions, and the month in question. Based on the labeled grid, the CO2 concentration of all grids within the target area is inferred; Determine the number of CO2 monitoring stations that need to be built within the target area; Acquire satellite remote sensing CO2 monitoring data corresponding to the grid within the target area; The optimal site selection scheme is obtained by evaluating the CO2 monitoring data from satellite remote sensing, the number of CO2 monitoring stations, and the CO2 concentration of all grids in the target area.

[0006] In some implementations... The CO2 emissions from carbon sources within the grid include the calculation of total carbon emissions. The CO2 absorption capacity of the carbon sink within the grid includes the calculation of the total carbon sink volume. The meteorological conditions include temperature, humidity, wind speed, weather, and air pressure.

[0007] In some implementations, the specific process of inferring the CO2 concentration of all grids within a target area based on the labeled grid includes the following steps: S101. Divide the grid of the target area into a set of grids with labeled states and a set of grids with unlabeled states; S102. Establish a neural network model and train it based on the environmental data information corresponding to the labeled grid. S103. Apply the trained neural network model to predict the CO2 concentration of each unlabeled grid. S104. For each unlabeled state, the grid is temporarily treated as a labeled state and added to the original set of labeled state grids for retraining. The unlabeled state grids corresponding to the model with the smallest root mean square error are transformed into labeled states, and the unlabeled state grids are moved from the unlabeled state grid set to the labeled state grid set. S105. Determine if the set of unlabeled grids is empty. If it is empty, the inference is complete; otherwise, execute S101.

[0008] In some implementations, the neural network model described in S102 includes an input layer, a hidden layer, and an output layer. The environmental data information corresponding to the labeled grid is used as the input data of the neural network model, and the CO2 concentration of the unlabeled grid is used as the output data of the neural network model.

[0009] In some implementations, a semi-supervised learning model is used to infer the CO2 concentration of all grids in the target area based on the labeled grid.

[0010] In some implementations... Assuming the target area is divided into k grids, and n CO2 monitoring stations need to be built within the target area, The process of evaluating and obtaining the optimal addressing scheme includes the following steps: Select any n grids as the site selection scheme, and train the SEMI-CARBON model using satellite remote sensing CO2 monitoring data; The root mean square error (RMSE) of the model is evaluated using the remaining (kn) grids within the region combined with satellite remote sensing CO2 monitoring data. conduct The optimal site selection scheme is the SEMI-CARBON model that minimizes the root mean square error (RMSE) among different site selection schemes. Each site selection scheme includes n CO2 monitoring stations.

[0011] In some implementations, no CO2 monitoring station is located in the target area.

[0012] The advantages of the CO2 monitoring station site selection method described in this invention are: It can effectively control costs while selecting CO2 monitoring station sites based on remote sensing data. A good site selection plan can make the estimation of CO2 concentration distribution more accurate, thereby making the estimation of total carbon emissions in the entire region more accurate. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the process of inferring the CO2 concentration of all grids within a target area based on labeled grids in some embodiments of the present invention. Figure 2 This is a flowchart illustrating the process of evaluating and obtaining the optimal addressing scheme in some embodiments of the present invention; Figure 3 This is a schematic diagram of a target region divided into multiple grids in some embodiments of the present invention; Figure 4 This is a model diagram of a neural network model in some embodiments of the present invention. Detailed Implementation

[0014] This embodiment proposes a method for selecting a CO2 monitoring station site, which is used to select a site for a target area where no CO2 monitoring station is currently located. The CO2 monitoring station site selection method includes the following steps: S1. Divide the target area into x grids, where x is an integer greater than 1. The CO2 concentration of a grid is typically related to several factors, such as: CO2 emissions from carbon sources within the grid, CO2 absorption capacity of carbon sinks within the grid, meteorological conditions, and the month in question. Figure 3 The content shown is as follows: If grid 1 and grid 2 in a region have similar total carbon emissions, total carbon sinks, meteorological conditions and the month they belong to, then the CO2 concentrations of these two grids will also be similar. S2. Based on existing environmental data, the grid is labeled to form a labeled grid. The environmental data includes CO2 concentration and other factors, such as CO2 emissions from carbon sources within the grid, CO2 absorption capacity of carbon sinks within the grid, meteorological conditions, and the month in question. The CO2 emissions from carbon sources within the grid include the calculation of total carbon emissions. The CO2 absorption capacity of the carbon sink within the grid includes the calculation of the total carbon sink volume. Meteorological conditions include temperature, humidity, wind speed, weather conditions, and air pressure; S3. Based on the labeled grid, infer the CO2 concentration of all grids within the target area, and combine... Figure 1 The specific reasoning process for the content shown is as follows: S3101. Divide the grid of the target area into a set of grids with labeled states and a set of grids with unlabeled states; S302. A neural network model is established and trained based on the environmental data information corresponding to the labeled grid states. A semi-supervised learning model can be used for training, which is existing technology and will not be elaborated upon here. Combined with... Figure 4 The content shown indicates that the neural network model in this step includes an input layer (the input data is...). , , , , This represents the total carbon emissions calculated for the k-th grid. It is the total carbon emissions calculated using standard methods. An increase in the total carbon emissions will directly lead to an increase in CO2 concentration. The total carbon emissions can be obtained from the local greenhouse gas emissions inventory. This represents the total carbon sink of the k-th grid, which is the total carbon sink calculated using standard methods. An increase in the total carbon sink will directly lead to a decrease in CO2 concentration. The total carbon sink can be obtained from the local greenhouse gas carbon inventory. This represents the meteorological conditions of the k-th grid. Meteorological conditions and CO2 concentration have an indirect influence. Meteorological conditions can be obtained from ECMWF. This represents the month of the k-th grid. From both a human economic activity and ecoclimatic perspective, carbon emissions and carbon sinks vary across different months, especially different seasons, which indirectly affects CO2 concentration. It also includes a hidden layer and an output layer (outputting data as...). , (This represents the CO2 concentration of the k-th grid). The environmental data corresponding to the labeled grids are used as the input data for the neural network model, and the CO2 concentration of the unlabeled grids is used as the output data for the neural network model. Input to the neural network model: a set of labeled grid cells Unlabeled set of meshes All feature sets ( ); Neural network model output: CO2 concentration in the grid .

[0015] The neural network model can use the following algorithm: 1. Do 2. SEMI-CARBON ← nn.learning( , , , , ) 3. Model ← [] 4. For k ← 1 tolength[ ] 5. ← [k] 6. ← SEMI-CARBON.infer( , , , , ) 7. Model[k] ←nn.learning( , , , , [ , { }]) 8. Remove the unlabeled mesh corresponding to the model with the smallest RMSE from the Model set. Move to middle 9. Until It is an empty set; S303. Apply the trained neural network model to predict the CO2 concentration of each unlabeled grid. S304. For each unlabeled state, the grid is temporarily treated as a labeled state and added to the original set of labeled state grids for retraining. The unlabeled state grids corresponding to the model with the smallest root mean square error are transformed into labeled states, and the unlabeled state grids are moved from the set of unlabeled state grids to the set of labeled state grids. S305. Determine if the set of unlabeled grids is empty. If it is empty, the inference is complete; otherwise, execute S301. S4. Determine the number of CO2 monitoring stations that need to be built in the target area. Assume that n CO2 monitoring stations need to be built in the target area. S5. Obtain satellite remote sensing CO2 monitoring data corresponding to the grid within the target area; S6. Based on satellite remote sensing CO2 monitoring data, the number of CO2 monitoring stations, and the CO2 concentration of all grids within the target area, an optimal site selection scheme is obtained through evaluation, combined with... Figure 2 The process of evaluating and obtaining the optimal addressing scheme, as shown, includes the following steps: S601. Select any n grids as the site selection scheme, and train the SEMI-CARBON model by combining satellite remote sensing CO2 monitoring data; S602. Use the remaining (xn) grids in the region combined with satellite remote sensing CO2 monitoring data to evaluate the root mean square error (RMSE) of the model. S603, proceed The optimal site selection scheme is the SEMI-CARBON model that minimizes the root mean square error (RMSE) among different site selection schemes. Each site selection scheme includes n CO2 monitoring stations.

[0016] Inputs to the evaluation model: the planned number of monitoring stations n, the total number of grids k in the region, the set of all grids G within each grid, and the feature set ( ), satellite remote sensing CO2 column concentration monitoring data S; Evaluation model output: The set of grids W selected in the optimal site selection scheme.

[0017] The optimal addressing scheme described in S6 can be obtained using the following algorithm: 1. Find the set P of all permutations and combinations of n grids in G. 2.RMSE ← [] 3. for i ← 1 to length[P] 4. Select the i-th grid combination P[i]=[r1i,r2i,…,rni] from P. 5. Set P[i] as the labeled set. And set the remote sensing concentration data of the corresponding grid in S as the label. 6. ← G - 7. Model ← SEMI-CARBON.learn( , , , , , ) 8. RMSE[i] ← Model.infer( ) 9.W ← P[i] corresponding to the minimum value RMSE[i] in the RMSE set.

[0018] Matters not covered in this invention can be directly implemented using existing technologies, and therefore will not be elaborated upon here.

[0019] The above description is merely a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several similar modifications and improvements can be made without departing from the inventive concept of the present invention, and these should also be considered within the scope of protection of the present invention.

Claims

1. A method for selecting the location of a CO2 monitoring station, characterized in that, Includes the following steps: Divide the target area into multiple grids; The grid is labeled based on existing environmental data to form a labeled grid. The environmental data includes CO2 concentration and other factors, including CO2 emissions from carbon sources within the grid, CO2 absorption capacity of carbon sinks within the grid, meteorological conditions, and the month in question. The CO2 concentration of all grid cells within the target area is inferred based on the labeled grid. The specific inference process includes the following steps: S101. Divide the grid of the target area into a set of grids with labeled states and a set of grids with unlabeled states; S102. Establish a neural network model and train it based on the environmental data information corresponding to the labeled grid. S103. Apply the trained neural network model to predict the CO2 concentration of each unlabeled grid. S104. For each unlabeled state, the grid is temporarily treated as a labeled state and added to the original set of labeled state grids for retraining. The unlabeled state grids corresponding to the model with the smallest root mean square error are transformed into labeled states, and the unlabeled state grids are moved from the unlabeled state grid set to the labeled state grid set. S105. Determine if the set of unlabeled grids is empty. If it is empty, the inference is complete; otherwise, execute S101. Determine the number of CO2 monitoring stations that need to be built within the target area; Acquire satellite remote sensing CO2 monitoring data corresponding to the grid within the target area; The optimal site selection scheme is obtained by evaluating the CO2 monitoring data from satellite remote sensing, the number of CO2 monitoring stations, and the CO2 concentration of all grids in the target area. Assuming the target area is divided into k grids, and n CO2 monitoring stations need to be built within the target area, The process of evaluating and obtaining the optimal addressing scheme includes the following steps: Select any n grids as the site selection scheme, and train the SEMI-CARBON model using satellite remote sensing CO2 monitoring data; The root mean square error (RMSE) of the model is evaluated using the remaining (kn) grids within the region combined with satellite remote sensing CO2 monitoring data. conduct The optimal site selection scheme is the SEMI-CARBON model that minimizes the root mean square error (RMSE) among different site selection schemes. Each site selection scheme includes n CO2 monitoring stations.

2. The CO2 monitoring station site selection method according to claim 1, wherein, The CO2 emissions from carbon sources within the grid include the calculation of total carbon emissions. The CO2 absorption capacity of the carbon sink within the grid includes the calculation of the total carbon sink volume. The meteorological conditions include temperature, humidity, wind speed, weather, and air pressure.

3. The CO2 monitoring station site selection method according to claim 1, wherein, The neural network model described in S102 includes an input layer, a hidden layer, and an output layer. The environmental data information corresponding to the labeled grid is used as the input data of the neural network model, and the CO2 concentration of the unlabeled grid is used as the output data of the neural network model.

4. The CO2 monitoring station site selection method according to claim 1, wherein, A semi-supervised learning model is used to infer the CO2 concentration of all grids in the target area based on the labeled grid.

5. The CO2 monitoring station site selection method according to claim 1, wherein, No CO2 monitoring stations are located in the target area.

Images