City near-real-time gridding carbon emission measuring and calculating method based on multi-source big data

Abstract

The invention relates to the technical field of carbon emission detection, and discloses an urban near-real-time gridding carbon emission measuring and calculating method based on multi-source big data, which comprises the following steps: constructing a gridding carbon emission measuring and calculating database, historical carbon emission data of all departments and industries and multi-source big data capable of reflecting spatial and temporal characteristics of carbon emission intensity are collected and processed into activity level indication variables and spatial allocation weight variables; constructing a relationship between the activity level indication variable and carbon emissions of different departments or industries, and calculating a nearly real-time total daily carbon emission amount based on a grid carbon emission calculation database; and distributing the total day-by-day carbon emission to the 1km grid by using the spatial distribution weight variable, and generating near-real-time day-by-day carbon emission data of the 1km grid. The urban near-real-time gridding carbon emission measuring and calculating method based on the multi-source big data aims at overcoming the defects of an existing carbon emission list in the aspects of temporal-spatial resolution, data updating speed and urban scale applicability.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission detection, and specifically to a method for calculating near-real-time grid carbon emissions in cities based on multi-source big data. Background Art

[0002] Greenhouse gases are the key factors leading to global warming. As the main force of carbon emissions, cities' emission reduction actions are crucial. With the progress of observation means such as ground-based stations, drones, and satellites, CO2 data is more intensive in space and time, requiring the CO2 source emission inventory to have higher spatial and temporal resolutions to match the scale of the observed data, fully explain local CO2 changes, and accurately quantify sources and sinks. Therefore, near-real-time grid emission inventories have become crucial. They can accurately reflect the spatial and temporal characteristics of carbon emission intensity, improve the accuracy of carbon flux inversion, achieve near-real-time monitoring of urban-level carbon emissions, effectively track anthropogenic emissions, and provide strong support for addressing climate change.

[0003] Existing grid emission inventories mainly allocate the annual emission inventories of cities or countries to corresponding grids through spatial indicator variables such as population, GDP, night light, and industrial point source information. However, the annual emission inventories of cities or countries mainly rely on government official energy consumption statistical data, which often lags behind by more than one year, and the scale is often annual or monthly. In the process of carbon flux inversion, it is difficult to provide accurate emission point source information. In addition, existing grid emission inventories are limited by the resolution of spatial indicator variables and often cannot reach the 1km scale. Even though the existing emission inventory ODIAC can reach 1km, the error is large at low resolutions. The existing near-real-time non-grid emission inventory Carbon Monitor uses near-real-time collectable activity level indicator variables to calculate the daily carbon emissions of each department, that is:

[0004] Emis s =EF s ×AL s

[0005] Where Emis s 、EF s and AL s are the carbon emissions, carbon emission factors, and activity level indicator variables of department s respectively. This method breaks through the limitation of traditional emission inventories lagging behind by more than one year and realizes near-real-time monitoring of anthropogenic carbon emissions. However, the selection of activity level indicator variables in this method is not suitable for the urban scale because many activity level indicator variables are difficult to obtain at the urban scale and are usually applicable to large regions and difficult to apply in urban-level regions. Summary of the Invention

[0006] The objective of the present invention is to address the deficiencies of existing carbon emission inventories in terms of spatio-temporal resolution, data update speed, and applicability at the urban scale, and to propose a method for near-real-time gridded carbon emission measurement in cities based on multi-source big data.

[0007] The technical solution of the present invention to solve the above technical problems is as follows:

[0008] A method for near-real-time gridded carbon emission measurement in cities based on multi-source big data, comprising the following steps:

[0009] S10: Construct a grid carbon emission measurement database, collect historical carbon emission data of various departments and industries, as well as multi-source big data that can reflect the spatio-temporal characteristics of carbon emission intensity, and process them into activity level indicator variables and spatial allocation weight variables;

[0010] S20: Establish the relationship between activity level indicator variables and carbon emissions among different departments or industries, and calculate the near-real-time daily total carbon emissions based on the grid carbon emission measurement database;

[0011] S30: Use the spatial allocation weight variables to allocate the daily total carbon emissions to 1 km grids, and generate near-real-time daily carbon emission data for 1 km grids.

[0012] Based on the above technical solution, the present invention can also be improved as follows.

[0013] Furthermore, in the step S10, the multi-source big data includes: daily electricity consumption of 35 industrial sectors, daily electricity consumption of residents, energy consumption of the power sector, surface site NO2 monitoring data, TROPOMI-NO2 data, traffic road network data, point of interest data, industrial point source data, and environmental covariate data; among them, the environmental covariate data includes meteorological data, land use type, population, and normalized vegetation index; the meteorological data further includes surface atmospheric pressure, total precipitation, surface 2m temperature, surface 2m dew point temperature, boundary layer height, and surface net radiation.

[0014] Furthermore, in the step S10, the data preprocessing includes using an isolated random forest to detect and remove outliers in the multi-source data, using the machine learning model XGBoost for multiple imputation of missing values, and using co-kriging and bilinear interpolation to resample the environmental covariate data to 1 km grids.

[0015] Further, in the step S10, the activity level indicator variable includes the residential heating electricity consumption in the winter heating season extracted based on the daily electricity consumption of residents during the non-heating and non-cooling periods in April and October, and the result after daily reconstruction of the NO2 concentration of the 1km grid on the ground using the XGBoost model combined with the NO2 column concentration, surface NO2 concentration, and environmental covariate data observed by the TROPOMI sensor.

[0016] Further, it also includes using the random forest model to normalize the reconstructed NO2 concentration of the 25km grid meteorologically, establishing the statistical relationship between the normalized NO2 concentration and traffic NO X emissions, and calculating the near-real-time traffic NO X emission.

[0017] Further, in the step S10, the spatial allocation weight variable is determined according to the POI kernel density calculated based on the POI data related to residents' residence and life, and the road density calculated based on the sum of the lengths of specific road types within the 1km grid.

[0018] Further, in the step S20, when constructing the statistical relationship between the activity level indicator variable and the total carbon emissions, the exponential smoothing state space model is used to capture the change trend of the carbon emission factor, estimate the most reasonable near-real-time carbon emission factor, and the calculation formula of the carbon emission factor is:

[0019]

[0020] where EF s,n represents the carbon emission factor of department or industry s in the past nth year; Emis s,n represents the total annual carbon emissions of department or industry s in the past nth year; AL s,n,x represents the activity level indicator variable of department or industry s on the xth day in the past nth year.

[0021] Further, in the step S20, the formula for calculating the near-real-time daily total carbon emissions is:

[0022] C s,T,d = F s,T,est × A s,T,d

[0023] where C s,T,d represents the carbon emissions of department or industry s on the dth day in the current year T, F s,T,est represents the carbon emission factor of department or industry s in the current year T estimated using the ETS model, and A s,T,d represents the activity level indicator variable of department or industry s on the dth day in the current year T.

[0024] Furthermore, the S30 includes the following steps:

[0025] S301: Screen out POI data related to residents' living and life according to keywords, and then calculate the POI kernel density. The basic calculation formula is as follows:

[0026]

[0027] where POId o is the kernel density of grid o, R represents the search radius for kernel density calculation, D o,q is the distance from the POI within the search radius q to grid o, and k is the calculation weight of the POI within the search radius q which is determined by the type of POI;

[0028] S302: Take the sum of the lengths of specific road types within a 1km grid as the road density. The basic calculation formula is as follows:

[0029]

[0030] where RD o is the road density of 1km grid o, and L o,r is the length of road type r in 1km grid o.

[0031] Compared with the prior art, the technical solution of this application has the following beneficial technical effects:

[0032] By constructing a grid carbon emission measurement database, the present invention not only collects historical carbon emission data of various departments and industries, but also integrates multi-source big data that can reflect the spatio-temporal characteristics of carbon emission intensity. Through the fusion of multi-source data, the present invention can more comprehensively capture the dynamic changes of urban carbon emissions, provide more accurate and comprehensive basic data for subsequent carbon emission measurement, and also establish the relationship between activity level indicator variables and carbon emissions among different departments or industries. Based on the grid carbon emission measurement database, the near-real-time daily total carbon emissions are calculated, thus breaking through the limitations of the traditional carbon emission inventory data being lagged and having a rough time scale. Through real-time or near-real-time activity level indicator variables, the real-time status of urban carbon emissions can be dynamically reflected, providing more timely and accurate data support for the monitoring and management of urban carbon emissions. Finally, using the spatial allocation weight variable, the daily total carbon emissions are allocated to 1km grids, generating near-real-time daily carbon emission data for 1km grids, improving the spatial resolution of carbon emission measurement, and enabling the measurement of urban carbon emissions to more precisely reflect specific geographical spaces. Brief Description of the Drawings

[0033] Figure 1 is a schematic diagram of the near-real-time activity level indicator variable of the present invention;

[0034] Figure 2 This is a schematic diagram of the CO2 emission time series of the present invention;

[0035] Figure 3 This is a schematic diagram of the carbon emission data of the present invention. Detailed implementation manners

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] Combined with Figures 1-3 As shown, a method for calculating near-real-time grid carbon emissions in a city based on multi-source big data of the present invention includes the following steps:

[0038] S10: Construct a grid carbon emission calculation database, collect historical carbon emission data of various departments and industries and multi-source big data that can reflect the spatio-temporal characteristics of carbon emission intensity, and process them into activity level indicator variables and spatial allocation weight variables;

[0039] S20: Establish the relationship between the activity level indicator variables and the carbon emissions between different departments or industries, and calculate the near-real-time daily total carbon emissions based on the grid carbon emission calculation database;

[0040] S30: Use the spatial allocation weight variables to allocate the daily total carbon emissions to 1 km grids to generate near-real-time daily carbon emission data for 1 km grids.

[0041] In a preferred embodiment of the present invention, it can be further configured as follows: In step S10, the multi-source big data includes: daily electricity consumption of 35 industrial sectors, daily electricity consumption of residents, energy consumption of the power department, NO2 monitoring data of surface stations, TROPOMI-NO2 data, traffic road network data, point of interest data, industrial point source data, and environmental covariate data; among them, the environmental covariate data includes meteorological data, land use type, population, and normalized vegetation index; the meteorological data further includes surface atmospheric pressure, total precipitation, surface 2m temperature, surface 2m dew point temperature, boundary layer height, and surface net radiation, which clearly defines the specific composition of the multi-source big data, greatly enriching the data basis for carbon emission measurement. By integrating multi-dimensional data such as the daily electricity consumption of 35 industrial sectors, the daily electricity consumption of residents, and the energy consumption of the power department, the present invention can more comprehensively reflect all aspects of urban carbon emissions. At the same time, the addition of environmental monitoring data such as NO2 monitoring data of surface stations and TROPOMI-NO2 data provides a more direct basis for the estimation of carbon emissions, improving the accuracy of the measurement. In addition, the integration of spatial information such as traffic road network data, point of interest data, and industrial point source data enables the measurement of carbon emissions to be more accurately located in specific geographical spaces, providing the possibility for hotspot identification and source tracking of urban carbon emissions. The consideration of environmental covariate data, such as meteorological data, land use type, population, and normalized vegetation index, further refines the influencing factors of carbon emissions, making the measurement of carbon emissions more scientific and comprehensive.

[0042] The daily electricity consumption data of 35 industrial sectors can reflect the production activity level of the industrial sector and is an important basis for estimating industrial carbon emissions; the daily electricity consumption data of residents can reflect the demand for electricity in residents' lives and then estimate the carbon emissions of residents' lives; the energy consumption data of the power sector can comprehensively reflect the carbon emissions in the power production process. The surface station NO2 monitoring data and TROPOMI-NO2 data, as air pollutant monitoring data, have a certain correlation with carbon emissions and can provide auxiliary information for estimating carbon emissions. The traffic road network data can reflect the busyness of urban traffic and then estimate the carbon emissions of the transportation sector; the point-of-interest data can reflect the distribution of urban commercial, service and other activities and provide clues for estimating the carbon emissions of these industries. The industrial point source data directly provides the location and emission information of industrial emission points and is an indispensable part of carbon emission measurement. The consideration of environmental covariate data, such as surface atmospheric pressure, total precipitation, surface 2m temperature, surface 2m dew point temperature, boundary layer height and surface net radiation in meteorological data, can affect the diffusion and transformation process of carbon emissions and is of great significance for accurately estimating carbon emissions; the land use type data can reflect the use of urban land and then affect the estimation of carbon emissions; the population data can reflect the urban population distribution and density and provide a basis for estimating the carbon emissions of residents' lives; the normalized difference vegetation index can reflect the urban greening situation and has a certain auxiliary effect on estimating urban carbon sinks. Through the integration and processing of these multi-source big data, the actual situation of urban carbon emissions can be more comprehensively reflected, providing more accurate and comprehensive data support for the monitoring and management of urban carbon emissions.

[0043] In a preferred embodiment of the present invention, it can be further configured as follows: in step S10, the data preprocessing includes using an isolation random forest to detect and remove outliers in multi-source data, using the machine learning model XGBoost for multiple imputation of missing values, and using co-Kriging and bilinear interpolation to resample the environmental covariate data to a 1km grid. The data preprocessing improves the accuracy and reliability of the carbon emission measurement data. By using an isolation random forest to detect and remove outliers in multi-source data, the interference of abnormal data on the carbon emission measurement results is effectively avoided, ensuring the cleanliness of the data. At the same time, using the machine learning model XGBoost for multiple imputation of missing values not only improves the data integrity but also, through the powerful fitting ability of machine learning, makes the imputation results closer to the true values, reducing the impact of data missing on the measurement results. In addition, using co-Kriging and bilinear interpolation to resample the environmental covariate data to a 1km grid makes the environmental covariate data match the grid requirements of carbon emission measurement, improving the accuracy and consistency of spatial analysis and providing accurate basic data for subsequent carbon emission spatial allocation.

[0044] Isolation Forest is an efficient outlier detection method. It isolates outliers by constructing a random forest, thereby achieving rapid identification and removal of outliers. This method can not only handle high-dimensional data but also effectively deal with outliers in the data, ensuring the accuracy and reliability of the data. Secondly, as an advanced machine learning model, XGBoost (Extreme Gradient Boosting) has powerful fitting ability and generalization performance. During the missing value imputation process, XGBoost can make full use of other information in the data to perform multiple imputations on the missing values, generate multiple possible imputation results, and select the optimal imputation value through model evaluation, thereby improving data integrity and imputation accuracy. Finally, Cokriging and Bilinear Interpolation are two commonly used spatial data interpolation methods. Cokriging considers the correlation between multiple variables and improves the interpolation accuracy by using the information of related variables, while Bilinear Interpolation is a simple and effective interpolation method suitable for interpolating regular grid data. In the present invention, according to the characteristics and requirements of environmental covariate data, Cokriging and Bilinear Interpolation methods are flexibly selected to resample the data, enabling the environmental covariate data to accurately match the 1km grid, providing accurate and consistent basic data for subsequent spatial allocation of carbon emissions.

[0045] In a preferred embodiment of the present invention, it can be further configured as follows: In step S10, the activity level indicator variables include the residential heating electricity consumption in the winter heating season extracted based on the daily electricity consumption of residents during the non-heating and non-cooling periods in April and October, and the result after daily reconstruction of the NO2 concentration of the 1km surface grid by using the XGBoost model combined with the NO2 column concentration, surface NO2 concentration, and environmental covariate data observed by the TROPOMI sensor. By taking the daily electricity consumption of residents during the non-heating and non-cooling periods in April and October as a benchmark and extracting the residential heating electricity consumption in the winter heating season as one of the activity level indicator variables, it effectively reflects the impact of residents' winter heating activities on carbon emissions and improves the pertinence of carbon emission measurement. At the same time, by using the XGBoost model combined with the NO2 column concentration, surface NO2 concentration, and environmental covariate data observed by the TROPOMI sensor to perform daily reconstruction of the NO2 concentration of the 1km surface grid, it not only improves the spatio-temporal resolution of the NO2 concentration data but also makes the reconstructed NO2 concentration data closer to the real situation through the powerful fitting ability of machine learning, providing a more accurate activity level indicator variable for carbon emission measurement, enabling the carbon emission measurement to more precisely reflect the impact of human activities on carbon emissions and improving the accuracy and reliability of the measurement results.

[0046] First, taking the daily electricity consumption of residents during the non-heating and non-cooling periods in April and October as the benchmark, the residential heating electricity consumption during the winter heating season is extracted as an activity level indicator variable. This indicator variable can directly reflect the intensity of residents' winter heating activities, and then estimate the carbon emissions generated by heating activities. To ensure the accuracy and comparability of the data, the present invention has strictly screened and corrected the daily electricity consumption of residents, excluding the influence of outliers and missing values. Secondly, using the XGBoost model combined with the NO2 column concentration, surface NO2 concentration, and environmental covariate data observed by the TROPOMI sensor, the NO2 concentration of the 1km grid on the surface is reconstructed day by day. The TROPOMI sensor can provide high-resolution NO2 column concentration data, but the acquisition of surface NO2 concentration is relatively difficult and there are problems of insufficient spatio-temporal resolution. Therefore, the present invention uses the XGBoost model to fuse the NO2 column concentration observed by the TROPOMI sensor with the surface NO2 concentration and environmental covariate data (such as meteorological data, land use types, etc.), and uses the machine learning algorithm to reconstruct the NO2 concentration of the 1km grid on the surface day by day. In this process, the XGBoost model can fully explore the correlation between data and improve the accuracy of the reconstruction results. At the same time, through day-by-day reconstruction, the present invention obtains high spatio-temporal resolution NO2 concentration data, providing a more accurate activity level indicator variable for carbon emission calculation.

[0047] In a preferred embodiment of the present invention, it can be further configured as: further including using the random forest model to perform meteorological normalization on the reconstructed NO2 concentration of the 25km grid, establishing a statistical relationship between the normalized NO2 concentration and traffic NOx emissions, and calculating the near-real-time traffic NOx emissions. By introducing the random forest model to perform meteorological normalization on the reconstructed NO2 concentration of the 25km grid, the comparability and accuracy of the NO2 concentration data are improved. Through meteorological normalization processing, the present invention effectively eliminates the influence of meteorological factors on the NO2 concentration data, making the normalized NO2 concentration data more representative and comparable. Further, the present invention establishes a statistical relationship between the normalized NO2 concentration and traffic NO X emissions, and calculates the near-real-time traffic NO X emissions accordingly, which not only provides a more accurate method for estimating traffic NO X emissions, but also realizes near-real-time monitoring, greatly improving the timeliness and accuracy of carbon emission calculation, and providing strong support for environmental protection and traffic management.

[0048] The reconstructed NO2 concentration of the 25km grid is meteorologically normalized using a random forest model. As an ensemble learning method, the random forest model has powerful non-linear fitting ability and generalization performance. It improves the accuracy and stability of the normalization process by constructing multiple decision trees and synthesizing their output results. During the meteorological normalization process, this invention considers various meteorological factors such as temperature, humidity, wind speed, wind direction, etc. The impacts of these factors on the NO2 concentration are fully incorporated into the model. Through the training and optimization of the random forest model, this invention obtains a normalization model that can accurately reflect the impact of meteorological factors on the NO2 concentration, traffic NO X emission is one of the main sources of urban air pollution. Accurately estimating its emissions is crucial for formulating effective environmental protection policies. This invention uses statistical analysis methods to explore the correlation between the normalized NO2 concentration and traffic NO X emissions, and establishes a corresponding statistical relationship model. This model can quickly and accurately estimate traffic NO X emissions based on the normalized NO2 concentration data. In addition, this invention also realizes near-real-time calculation of traffic NO X emissions. By real-time monitoring of NO2 concentration data and combining the normalization model and the statistical relationship model, this invention can quickly reflect the changes in traffic NO X emissions, providing timely and accurate data support for environmental protection departments and traffic management departments. It not only improves the timeliness of carbon emission measurement but also provides a more scientific and effective decision-making basis for environmental protection and traffic management.

[0049] In a preferred embodiment of this invention, it can be further configured as follows: In step S10, the spatial allocation weight variable is determined based on the POI kernel density calculated from POI data related to residents' residence and life, and the road density calculated from the sum of the lengths of specific road types within a 1km grid. By introducing the POI kernel density calculated from POI data related to residents' residence and life, and the road density calculated from the sum of the lengths of specific road types within a 1km grid to determine the spatial allocation weight variable, this invention more precisely depicts the spatial distribution of residents' residence and life activities, as well as the spatial differences in traffic flow through these two fine-grained spatial feature indicators of POI kernel density and road density. The POI kernel density reflects the concentration of residents' living facilities and is a direct manifestation of the hotspots of residents' activities; the road density reflects the density of the traffic network and is closely related to traffic flow and emissions. The combination of the two as the spatial allocation weight variable makes the spatial allocation of carbon emissions more in line with reality, providing a more scientific and accurate basis for environmental policy formulation and urban planning.

[0050] The present invention first collects a large amount of POI data related to residents' living and life, including but not limited to residential communities, commercial facilities, educational institutions, medical institutions, etc. Then, using the kernel density estimation method, spatial analysis is performed on these POI data to calculate the POI kernel density within each 1-km grid. The kernel density estimation method can fully consider the spatial proximity and density distribution characteristics of POI points, making the calculated POI kernel density more accurate and reliable. Through the POI kernel density, the present invention can clearly identify the hot spots of residents' activities, providing strong data support for the spatial allocation of carbon emissions. According to the road type and length information, the present invention statistically analyzes the roads within each 1-km grid to calculate the road density. The sum of the lengths of specific road types reflects the degree of development of the transportation network within the grid, which has a close relationship with traffic flow and emissions. Through the road density, the present invention can accurately depict the spatial distribution characteristics of traffic flow, providing another important weighting variable for the spatial allocation of carbon emissions.

[0051] In a preferred embodiment of the present invention, it can be further configured as follows: In step S20, when constructing the statistical relationship between the activity level indicator variable and the total carbon emissions, the exponential smoothing state space model is used to capture the change trend of the carbon emission factor and estimate the most reasonable carbon emission factor in near real-time. The calculation formula of the carbon emission factor is:

[0052]

[0053] where EF s,n represents the carbon emission factor of department or industry s in the nth year in the past; Emis s,n represents the total annual carbon emissions of department or industry s in the nth year in the past; AL s,n,x represents the activity level indicator variable of department or industry s on the xth day in the nth year in the past. In step S20, the present invention constructs the statistical relationship between the activity level indicator variable and the total carbon emissions by using the exponential smoothing state space model, improving the accuracy and timeliness of the carbon emission factor estimation. The introduced exponential smoothing state space model by the present invention can make full use of the internal law of time series data, effectively capture the long-term trend, seasonal variation and random fluctuation of the carbon emission factor, so as to estimate the most reasonable carbon emission factor in near real-time. It not only improves the accuracy of carbon emission measurement, but also makes the estimation of the carbon emission factor more in line with the actual change situation, providing more reliable data support for carbon emission monitoring, reporting and verification.

[0054] The present invention collects historical data on the total carbon emissions and activity level indicator variables of various departments or industries over the past years, which form the basis of time series data. Then, for each department or industry, the present invention constructs an exponential smoothing state space model, which combines the simplicity of the exponential smoothing method and the flexibility of the state space model, and can simultaneously handle the trend, seasonality, and random components of time series data. During the model construction process, the present invention optimally selects the parameters of the exponential smoothing state space model to ensure the fitting effect and prediction ability of the model. Through the training of historical data, the smoothing parameters, trend parameters, and seasonal parameters in the model are determined, enabling the model to accurately reflect the changing trend of carbon emission factors. At the same time, the present invention also adopts a recursive algorithm to update the model parameters to adapt to the addition of new data, thereby realizing near-real-time estimation of carbon emission factors.

[0055] In a preferred embodiment of the present invention, it can be further configured as follows: In step S20, the formula for calculating the near-real-time daily total carbon emissions is:

[0056] C s,T,d =F s,T,est ×A s,T,d

[0057] Wherein, C s,T,d represents the carbon emissions of department or industry s on the d-th day of the current year T, F s,T,est represents the carbon emission factor of department or industry s of the current year T estimated by the ETS model, and A s,T,d represents the activity level indicator variable of department or industry s on the d-th day of the current year T. By introducing the daily activity level indicator variable and the carbon emission factor of the current year, a calculation formula that can dynamically reflect the daily change of carbon emissions is constructed. This formula not only considers the specific carbon emission characteristics of the department or industry, but also realizes the accurate calculation of carbon emissions through the carbon emission factor estimated by the model, making the calculation of carbon emissions more in line with the actual situation, providing strong support for the real-time monitoring, reporting, and verification of carbon emissions, and helping to timely discover and solve carbon emission anomaly problems. The corresponding relationship between the spatial allocation weight variable and the department and industry is as Figure 1 shown: The POI density of residential life is used as the spatial allocation weight variable of residential carbon emissions; the electricity consumption of regional industrial industries is used as the spatial allocation weight variable of industrial carbon emissions. First, the carbon emissions of each industry at the city level are allocated to small regions, and then allocated to each point source according to the enterprise scale; the road density is used as the spatial allocation weight variable of ground traffic carbon emissions to further allocate the ground traffic carbon emissions from a 25 km grid to a 1 km grid. The near-real-time daily total carbon emissions measured are allocated to 1 km using the spatial allocation weight variable, and its allocation formula is:

[0058]

[0059] Among them, SW s,O is the spatial allocation weight variable of department or industry s in the 1km grid o.

[0060] In a preferred embodiment of the present invention, it can be further configured as follows: S30 includes the following steps:

[0061] S301: Screen out POI data related to residents' residence and life according to keywords, and then calculate the POI kernel density. The basic calculation formula is as follows:

[0062]

[0063] Among them, POId o is the kernel density of grid o, R represents the search radius for kernel density calculation, D o,q is the distance from the POI q within the search radius to grid o, and k is the calculation weight of the POI q within the search radius, which is determined by the type of POI;

[0064] S302: Take the sum of the lengths of specific road types within the 1km grid as the road density. The basic calculation formula is as follows:

[0065]

[0066] Among them, RD o is the road density of the 1km grid o, L o,r is the length of road type r in the 1km grid o. By defining and implementing two key calculation steps in detail, namely the POI kernel density calculation in S301 and the road density calculation in S302, the accuracy and fineness of carbon emission spatial allocation and related analysis are improved. By introducing these two fine-grained spatial feature indicators of POI kernel density and road density, this problem is effectively solved. The calculation of POI kernel density takes into account POI data closely related to residents' residence and life. By setting a reasonable search radius and calculation weight, it can accurately reflect the spatial distribution of residents' activity hotspots. This not only provides a more scientific and accurate basis for the spatial allocation of carbon emissions, but also helps to identify high-carbon emission areas, providing strong support for environmental policy formulation and urban planning. At the same time, the calculation of road density quantifies the impact of the traffic network on carbon emissions by statistically summing the lengths of specific road types within the 1km grid, making the spatial allocation of carbon emissions more in line with the actual traffic flow and emission characteristics.

[0067] Filter out POI data related to residents' living and daily life according to keywords. These data may include various types such as residential communities, commercial facilities, educational institutions, medical institutions, etc. Then, set a reasonable search radius, and the size of this radius will directly affect the calculation result of kernel density. Within the search radius, calculate the distance from each POI point to the grid center, and determine the calculation weight according to the type of POI. Different types of POI points have different degrees of impact on carbon emissions. Therefore, reasonable weight values need to be set according to the actual situation. Finally, through the kernel density calculation formula, the kernel density value of each grid can be obtained, and this value reflects the spatial distribution characteristics of residents' activity hotspots. In step S302, the calculation of road density is relatively simple but equally important. First, divide the research area into several 1km grids, and then count the total length of specific road types within each grid. Here, the road types may include main roads, secondary roads, branch roads, etc. Different types of roads also have different impacts on carbon emissions. By calculating the road density, the impact degree of the transportation network on carbon emissions can be quantified, providing a more accurate basis for the spatial allocation of carbon emissions.

[0068] Taking Chengdu from 2019 to 2023 as an example, assuming that the energy consumption data of each official department is only updated to 2022, then the data for 2023 is the estimated data using activity indicator variables.

[0069] Construct a grid carbon emission measurement database for Chengdu from 2019 to 2023, and collect historical carbon emission data of various departments and industries as well as multi-source big data that can reflect the spatio-temporal characteristics of carbon emission intensity.

[0070] Clean and mine the multi-source big data to obtain activity level indicator variables and spatial allocation weight variables that reflect carbon emissions at the daily scale.

[0071] Based on the historical annual carbon emission data and activity indicator variables of each department and industry, calculate their historical carbon emission factors. The carbon emission factors (electric carbon coefficients) of various industries in industry and residential heating will change due to industrial innovation and changes in the residential heating energy structure. Therefore, an exponential smoothing state space model (ETS) is used to capture the change trend of carbon emission factors and reasonably estimate the carbon emission factors in 2023. The result example is shown in Table 1. The change range of the carbon emission factor (carbon-nitrogen ratio) of the transportation department is relatively small. Therefore, it is assumed to be unchanged during the research period, and the carbon-nitrogen ratio of 170.35 (gCO2 / gNO X ) calculated in 2019 is used as the carbon emission factor of this department.

[0072]

[0073]

[0074] Table 1: Electricity carbon coefficients of carbon emissions from 35 industrial sectors and residential heating (gCO2 / kWh)

[0075] Then, the near-real-time activity-level emission variables are used to calculate the daily carbon emissions of Chengdu. Taking July 7 to August 25, 2023 as an example, during this period, the Chengdu Municipal Government took a series of measures to restrict the emission activities of ground motor vehicles and regional industrial sectors. The CO2 emission time series during this period is as follows: Figure 2 As shown, the results show that this method reasonably captures the changes in carbon emissions caused by human control.

[0076] The present invention utilizes the electricity used in many industries, the electricity used for heating of residents, the energy consumption of power plants and the NO X Emissions and other indicators are used as activity level indicator variables to establish a response relationship with the total carbon emissions on a daily scale, and the total carbon emissions of each department are calculated in near real time on a daily scale. The total daily emissions are allocated based on the spatial allocation weight variables constructed based on residential POI data, industrial point source data, road network information, etc., to achieve near real-time 1km grid carbon emission data calculation, which effectively solves the problem of the existing grid emission inventory lagging behind for more than one year and insufficient spatial resolution.

[0077] The spatial allocation weight variable is used to allocate the total carbon emissions on a daily basis, generating near-real-time carbon emissions data for 1km grids in Chengdu. The example results for July 28, 2023 are as follows: Figure 3 As shown, high emission areas such as thermal power plants, cement manufacturing and industrial parks are reasonably displayed. Residential emissions are mainly concentrated in densely built areas and show a radial distribution outward, while traffic emissions show a linear radial distribution along the road network.

[0078] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0079] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A near-real-time grid-based carbon emission estimation method for cities based on multi-source big data, characterized in that: The following steps are involved: S10: Build a grid carbon emission measurement database, collect historical carbon emission data from various departments and industries, and multi-source big data that can reflect the spatiotemporal characteristics of carbon emission intensity, and process them into activity level indicator variables and spatial allocation weight variables; S20: Construct the relationship between activity level indicator variables and carbon emissions between different sectors or industries, and calculate the near-real-time daily total carbon emissions based on the grid carbon emissions measurement database; S30: The total daily carbon emissions are allocated to 1km grids using the spatial allocation weight variable, and near real-time daily carbon emissions data of 1km grids are generated.

2. According to the method of claim 1, a near real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: In the step S10, the multi-source big data includes: daily electricity consumption of 35 industrial sectors, daily electricity consumption of residents, energy consumption of the power sector, surface site NO2 monitoring data, TROPOMI-NO2 data, traffic network data, point of interest data, industrial point source data and environmental covariate data; wherein the environmental covariate data include meteorological data, land use type, population and normalized difference vegetation index; meteorological data further include surface atmospheric pressure, total precipitation, surface temperature at 2m, surface dew point temperature at 2m, boundary layer height and surface net radiation.

3. According to the method of claim 1, a near real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: In the step S10, data pre-processing includes using isolated random forests to detect and remove outliers in multi-source data, using the machine learning model XGBoost to perform multiple interpolations on missing values, and using cokriging and bilinear interpolation to resample environmental covariate data to a 1 km grid.

4. According to the method of claim 1, a near real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: In the step S10, the activity level indicator variables include the residential heating electricity consumption in the winter heating season extracted based on the daily residential electricity consumption in the non-heating and cooling periods in April and October, and the NO2 concentration of the 1 km surface grid reconstructed daily using the XGBoost model combined with the NO2 column concentration, surface NO2 concentration and environmental covariate data observed by the TROPOMI sensor.

5. According to claim 4, a near-real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: It also includes using the random forest model to normalize the reconstructed NO2 concentration of the 25km grid, and establishing the normalized NO2 concentration and traffic NO X emissions and calculates near real-time traffic NO X Emissions.

6. According to the method of claim 1, a near real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: In the step S10, the spatial allocation weight variable is determined based on the POI kernel density calculated based on POI data related to residents' residence and life, and the road density calculated based on the sum of the lengths of specific road types within a 1km grid.

7. According to claim 2, a near-real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: In the step S20, when constructing the statistical relationship between the activity level indicator variable and the total carbon emissions, an exponential smoothing state space model is used to capture the changing trend of the carbon emission factor and estimate the most reasonable carbon emission factor in near real time. The calculation formula of the carbon emission factor is: EF s,n Represents the carbon emission factor of the sector or industry s in the past n years; Emis s,n represents the total annual carbon emissions of the sector or industry s in the past n years; AL s,n,x An indicator variable representing the activity level of sector or industry s on day x in the past n years.

8. The method for calculating urban carbon emissions in near real-time grid based on multi-source big data according to claim 7 is characterized in that: In the step S20, the formula for calculating the near real-time daily total carbon emissions is: C s,T,d =F s,T,est ×A s,T,d Among them, C s,T,d represents the carbon emissions of sector or industry s on day d in the current year T, F s,T,est represents the carbon emission factor of the sector or industry s in the current year T estimated using the ETS model, A s,T,d An indicator variable representing the activity level of sector or industry s on day d in the current year T.

9. According to the method of claim 1, a near-real-time grid-based carbon emission estimation method for cities based on multi-source big data is characterized in that: The S30 includes the following steps: S301: Filter out POI data related to residents’ residence and life according to keywords, and then calculate the POI kernel density. The basic calculation formula is as follows: Among them, POId o is the kernel density of grid o, R represents the search radius for kernel density calculation, and D o,q Search for POI within the radius q The distance to grid o, k is the POI within the search radius q The calculation weight of is determined by the type of POI; S302: The sum of the lengths of the specific road types within the 1km grid is taken as the road density, and the basic calculation formula is as follows: Among them, RD o is the road density of 1km grid o, L o,r is the length of road type r in 1km grid o.

Images