Carbon emission calculation method and device based on dynamic simulation and underway actual measurement

Abstract

The invention discloses a carbon emission calculation method and device based on dynamic simulation and underway actual measurement, and the method comprises the steps: controlling an underway vehicle to collect real-time carbon dioxide data, real-time panoramic image data and real-time meteorological data of a road monitoring point, carrying out the recognition and analysis of the real-time panoramic image data, and obtaining an environment analysis result; inputting the real-time meteorological data and the environment analysis result into a background concentration simulation prediction model to obtain the carbon dioxide prediction background concentration of the monitoring point; calculating the carbon dioxide increment of the monitoring point according to the real-time carbon dioxide data and the carbon dioxide prediction background concentration of the monitoring point; on the basis of a Gaussian plume model, space-time integration is carried out on the carbon dioxide increments of the multiple monitoring points of the road, and the road carbon emission is obtained; the method can effectively improve the accuracy of road carbon emission calculation.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission calculation, and in particular to a carbon emission calculation method and device based on dynamic simulation and actual cruise measurement. Background Art

[0002] Currently, many cities and transportation management agencies rely on emission models based on vehicle type, number, and operating status to estimate transportation carbon emissions. These models typically estimate carbon emissions by statistically analyzing information such as vehicle type, number, and speed, and then combining fuel consumption data with emission factors. A representative model is the Motor Vehicle Emission Simulator (MOVES), which has been widely used to estimate carbon dioxide (CO2) and other greenhouse gases and pollutants in the transportation sector both domestically and internationally.

[0003] The MOVES model calculates carbon emissions from road traffic in four steps: (1) First, researchers collect and compile data on road traffic volume and vehicle types (such as light vehicles, heavy vehicles, trucks, etc.) using methods such as road intersection monitoring equipment, traffic questionnaires, or navigation map software; (2) Second, researchers calculate the driving status of traffic flow, including acceleration, deceleration, idling, and steady driving, using monitoring equipment, navigation software, etc.; (3) Based on the vehicle's operating status and meteorological conditions, the MOVES model is used to calculate the fuel consumption (including different types of fuel, such as #92, #95 gasoline, etc.) of each type of vehicle; (4) Finally, the emission factors for each vehicle type and fuel type are applied to convert fuel consumption into carbon dioxide emissions. Patent document CN119128334A discloses a method for calculating carbon emissions from urban roads under dynamically calibrated working conditions. Referring to the MOVES model, it constructs basic carbon emission factors for working condition units of different vehicle models, calculates second-by-second working conditions based on small sample GPS data, uses an improved NSGA-Ⅱ algorithm to divide the optimal speed range, and obtains the working condition distribution of different speed ranges, and dynamically updates the working condition correction parameters of the carbon emission factor library.

[0004] While the MOVES model and similar models based on traffic flow and emission factors can provide macro-level estimates of transportation carbon emissions, they also have significant shortcomings. Existing energy-based emission estimation models (such as MOVES) derive carbon emissions from statistical calculations of empirical data and emission factors. These methods fail to account for the impact of background CO2 concentrations, resulting in inaccurate carbon emissions. Currently, methods for calculating urban background CO2 concentrations primarily include station observations and mobile observations. Station observations primarily establish urban atmospheric CO2 background monitoring stations, recording CO2 concentrations from sources without carbon emissions as background concentrations. Mobile observations, on the other hand, use the minimum value over a long, mobile route, including a field monitoring section, as the background concentration. However, both methods suffer from temporal variability and spatial non-heterogeneity, resulting in a single background concentration value for the entire city at any given moment, failing to highlight regional variations. Summary of the Invention

[0005] The present invention provides a carbon emission calculation method and device based on dynamic simulation and actual measurement, which can effectively improve the accuracy of road carbon emission calculation.

[0006] A carbon emission calculation method based on dynamic simulation and on-board measurement, including: Control the roving vehicle in advance to collect carbon dioxide concentration data, panoramic image data, and meteorological data at road sampling points and perform multi-dimensional coupling screening to obtain carbon dioxide background concentration data; Training a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; Controlling the navigating vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and performing identification and analysis on the real-time panoramic image data to obtain environmental analysis results; Inputting the real-time meteorological data and environmental analysis results into the background concentration simulation prediction model to obtain the predicted background concentration of carbon dioxide at the monitoring point; Calculating the carbon dioxide increment at the monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; The carbon dioxide increments at multiple monitoring points on the road were calculated based on the Gaussian plume model to obtain the road carbon emissions.

[0007] Furthermore, after collecting carbon dioxide concentration data and meteorological data from road sampling points, the following is also included: The carbon dioxide concentration data and meteorological data are cleaned and standardized.

[0008] Furthermore, multi-dimensional coupling screening is performed on carbon dioxide concentration data, panoramic image data, and meteorological data, including: Performing time matching based on the acquisition time of the carbon dioxide concentration data, panoramic image data, and meteorological data; Using a pre-trained image recognition model to identify the panoramic image data, and determining the non-traffic impact area and its corresponding panoramic image acquisition time and environmental information based on the recognition result; Calculating the turbulence intensity in the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data; Obtaining a road congestion index for the non-traffic affected area at the corresponding panoramic image acquisition time, and calculating a background credibility score for the non-traffic affected area at the corresponding panoramic image acquisition time based on the turbulence intensity and the road congestion index; screening the carbon dioxide concentration data of the non-traffic impact area at the corresponding panoramic image acquisition time according to the background credibility score to obtain screened carbon dioxide concentration data; The screened carbon dioxide concentration data is clustered and analyzed, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data.

[0009] Furthermore, the recognition results include vehicles, vegetation, and building facades; The non-traffic impact area determined based on the recognition results and its corresponding panoramic image acquisition time and environmental information include: The pixel ratios of the vehicles, vegetation, and building facades in the panoramic image are calculated respectively, and the sampling points corresponding to the panoramic images where the pixel ratio of the vehicles is less than a preset value are determined as non-traffic impact areas. The corresponding panoramic image acquisition time, the pixel ratio of the vegetation, and the pixel ratio of the building facade are recorded.

[0010] Furthermore, the meteorological data includes wind speed; Calculating the turbulence intensity of the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data, including: According to the time matching result, the wind speed data of the non-traffic affected area within a preset time period before the corresponding panoramic image acquisition time is obtained; Calculating the wind speed standard deviation within the preset time period based on the wind speed data; The turbulence intensity is obtained by calculation according to the wind speed standard deviation and the wind speed of the non-traffic affected area at the corresponding panoramic image acquisition moment.

[0011] Furthermore, cluster analysis is performed on the screened carbon dioxide data, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data, including: The screened carbon dioxide data were Z-score normalized, and abnormal data points were removed based on the Z-score normalization results; Calculate the coefficient of variation for the remaining screened carbon dioxide data after removing abnormal points, and select candidate carbon dioxide background concentration data based on the coefficient of variation; Using the DBSCAN clustering algorithm to perform density analysis on the candidate carbon dioxide background concentration data, and taking the cluster center value with a density value greater than a preset density value as the clustering result; According to the time matching result, the clustering result is associated with the corresponding meteorological data, the pixel ratio of vegetation, and the pixel ratio of building facades to obtain the carbon dioxide background concentration data.

[0012] Furthermore, the real-time panoramic image data is identified and analyzed to obtain an environmental analysis result, including: The real-time panoramic image data is input into the image recognition model for recognition analysis to obtain the pixel ratio of vegetation at the monitoring point and the pixel ratio of the building facade.

[0013] Furthermore, the carbon dioxide increment at the monitoring point is the difference between the real-time carbon dioxide concentration and the predicted background concentration of carbon dioxide at the monitoring point.

[0014] Furthermore, the real-time meteorological data includes real-time wind speed; Based on the Gaussian plume model, the carbon dioxide increments at multiple monitoring points on the road are integrated in time and space to obtain the road carbon emissions, including: Calculate the monitoring area based on the monitoring points; According to the area of the monitoring area and the real-time wind speed, the Gaussian plume model is used to obtain the carbon dioxide emission rate of the road section corresponding to the monitoring point; The carbon dioxide emission rates calculated at each monitoring point on the same road are accumulated to obtain the total carbon dioxide emission rate of the road; The road carbon emissions are calculated based on the total carbon dioxide emission rate.

[0015] A carbon emission calculation device based on dynamic simulation and cruise measurement, comprising: The sample collection module is used to pre-control the navigation vehicle to collect carbon dioxide concentration data, panoramic image data and meteorological data at road sampling points and perform multi-dimensional coupling screening to obtain carbon dioxide background concentration data; A training module, configured to train a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; A real-time data acquisition module is used to control the navigation vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, identify and analyze the real-time panoramic image data, and obtain environmental analysis results; A prediction module, configured to input the real-time meteorological data and environmental analysis results into the background concentration simulation prediction model to obtain a predicted background concentration of carbon dioxide at a monitoring point; An increment calculation module, configured to calculate the increment of carbon dioxide at a monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; The emission calculation module is used to calculate the carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain the road carbon emissions.

[0016] Furthermore, the sample collection module is also used to: The carbon dioxide concentration data and meteorological data are cleaned and standardized.

[0017] Furthermore, the sample collection module performs multi-dimensional coupled screening on the carbon dioxide concentration data, panoramic image data, and meteorological data, including: Performing time matching based on the acquisition time of the carbon dioxide concentration data, panoramic image data, and meteorological data; Using a pre-trained image recognition model to identify the panoramic image data, and determining the non-traffic impact area and its corresponding panoramic image acquisition time and environmental information based on the recognition result; Calculating the turbulence intensity in the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data; Obtaining a road congestion index for the non-traffic affected area at the corresponding panoramic image acquisition time, and calculating a background credibility score for the non-traffic affected area at the corresponding panoramic image acquisition time based on the turbulence intensity and the road congestion index; screening the carbon dioxide concentration data of the non-traffic impact area at the corresponding panoramic image acquisition time according to the background credibility score to obtain screened carbon dioxide concentration data; The screened carbon dioxide concentration data is clustered and analyzed, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data.

[0018] Furthermore, the recognition results include vehicles, vegetation, and building facades; The sample collection module determines the non-traffic impact area and its corresponding panoramic image collection time according to the recognition result, including: The pixel ratios of the vehicles, vegetation, and building facades in the panoramic image are calculated respectively, and the sampling points corresponding to the panoramic images where the pixel ratio of the vehicles is less than a preset value are determined as non-traffic impact areas. The corresponding panoramic image acquisition time, the pixel ratio of the vegetation, and the pixel ratio of the building facade are recorded.

[0019] Furthermore, the meteorological data includes wind speed; The sample collection module calculates the turbulence intensity of the non-traffic affected area at the corresponding panoramic image collection time based on the meteorological data, including: According to the time matching result, the wind speed data of the non-traffic affected area within a preset time period before the corresponding panoramic image acquisition time is obtained; Calculating the wind speed standard deviation within the preset time period based on the wind speed data; The turbulence intensity is obtained by calculation according to the wind speed standard deviation and the wind speed of the non-traffic affected area at the corresponding panoramic image acquisition moment.

[0020] Furthermore, the sample collection module performs cluster analysis on the screened carbon dioxide data, and associates the clustering results with corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data, including: The screened carbon dioxide data were Z-score normalized, and abnormal data points were removed based on the Z-score normalization results; Calculate the coefficient of variation for the remaining screened carbon dioxide data after removing abnormal points, and select candidate carbon dioxide background concentration data based on the coefficient of variation; Using the DBSCAN clustering algorithm to perform density analysis on the candidate carbon dioxide background concentration data, and taking the cluster center value with a density value greater than a preset density value as the clustering result; According to the time matching result, the clustering result is associated with the corresponding meteorological data, the pixel ratio of vegetation, and the pixel ratio of building facades to obtain the carbon dioxide background concentration data.

[0021] Furthermore, the real-time data acquisition module performs recognition and analysis on the real-time panoramic image data to obtain an environmental analysis result, including: The real-time panoramic image data is input into the image recognition model for recognition analysis to obtain the pixel ratio of vegetation at the monitoring point and the pixel ratio of the building facade.

[0022] Furthermore, the carbon dioxide increment at the monitoring point is the difference between the real-time carbon dioxide concentration and the predicted background concentration of carbon dioxide at the monitoring point.

[0023] Furthermore, the real-time meteorological data includes real-time wind speed; The emission calculation module performs spatiotemporal integration of carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain road carbon emissions, including: Calculate the monitoring area based on the monitoring points; According to the area of the monitoring area and the real-time wind speed, the Gaussian plume model is used to obtain the carbon dioxide emission rate of the road section corresponding to the monitoring point; The carbon dioxide emission rates calculated at each monitoring point on the same road are accumulated to obtain the total carbon dioxide emission rate of the road; The road carbon emissions are calculated based on the total carbon dioxide emission rate.

[0024] The carbon emission calculation method and device based on dynamic simulation and cruise measurement provided by the present invention have at least the following beneficial effects: (1) The background concentration of carbon dioxide is taken into account in the calculation of road carbon emissions. The carbon dioxide increment is calculated based on the background concentration of carbon dioxide, and the carbon emissions are calculated based on the carbon dioxide increment, so that the obtained carbon dioxide emissions are more accurate; (2) The model is trained based on the data obtained from the actual flight measurement to obtain a background concentration simulation prediction model. The background concentration of carbon dioxide is predicted based on this background concentration prediction model. This model is applicable to various road scenes and different traffic conditions. It does not require the establishment of observation stations, effectively reducing costs and having wide applicability. (3) The background concentration simulation prediction model takes into account the impact of wind speed, buildings, and vegetation on the background concentration of carbon dioxide, effectively improving the accuracy of the background concentration prediction of carbon dioxide, and thus improving the accuracy of subsequent carbon dioxide emission calculations; (4) Temporal and spatial matching and screening of carbon dioxide concentration data, panoramic image data, and meteorological data used for model training can effectively improve the accuracy of the background concentration simulation prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of an embodiment of a carbon emission calculation method based on dynamic simulation and cruise measurement provided by the present invention.

[0026] Figure 2 The present invention provides a flowchart of an embodiment of data coupling screening in the carbon emission calculation method based on dynamic simulation and cruise measurement.

[0027] Figure 3 This is a flow chart of an embodiment of turbulence intensity calculation in the carbon emission calculation method based on dynamic simulation and underway measurement provided by the present invention.

[0028] Figure 4The present invention provides a flowchart of an embodiment of cluster analysis in the carbon emission calculation method based on dynamic simulation and cruise measurement.

[0029] Figure 5 The present invention provides a flow chart of an embodiment of carbon emission calculation method based on dynamic simulation and cruise measurement.

[0030] Figure 6 This is a structural schematic diagram of an embodiment of a carbon emission calculation device based on dynamic simulation and cruise measurement provided by the present invention. DETAILED DESCRIPTION

[0031] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0032] refer to Figure 1 In some embodiments, a carbon emission calculation method based on dynamic simulation and cruise measurement is provided, including: S1. Pre-control the roving vehicle to collect carbon dioxide concentration data, panoramic image data, and meteorological data at road sampling points and perform multi-dimensional coupling screening to obtain carbon dioxide background concentration data; S2. Training a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; S3, controlling the navigating vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and performing identification and analysis on the real-time panoramic image data to obtain environmental analysis results; S4, inputting the real-time meteorological data and environmental analysis results into the background concentration simulation prediction model to obtain the predicted background concentration of carbon dioxide at the monitoring point; S5. Calculating the carbon dioxide increment at the monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; S6. Calculate the carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain the road carbon emissions.

[0033] Specifically, in step S1, a navigating vehicle collects CO2 concentration data, panoramic image data, and meteorological data at sampling points along the road. Using an onboard mobile power supply and a high-precision CO2 concentration analyzer (such as the Picarro 2401), CO2 concentration at these sampling points is monitored. This analyzer utilizes advanced cavity ring-down technology, achieving high-precision measurements of 0.1 ppm. A panoramic camera captures panoramic image data. Onboard meteorological instruments record meteorological parameters such as atmospheric temperature, humidity, pressure, wind speed, and direction during the vehicle's journey. This data provides information on the environment, weather, and traffic volume—variables closely related to carbon emissions.

[0034] Furthermore, in step S1, after collecting carbon dioxide concentration data and meteorological data at road sampling points, the method further includes: The carbon dioxide concentration data and meteorological data are cleaned and standardized.

[0035] Specifically, data cleaning removes invalid data, such as sensor failures, lost signals, or anomalous data points (due to vibration). By removing this invalid data, the reliability and accuracy of the calculation results can be improved.

[0036] Standardization of carbon dioxide concentration and meteorological data enables data from different sources to be analyzed and calculated under unified standards.

[0037] Further, refer to Figure 2 In step S1, the carbon dioxide concentration data, panoramic image data and meteorological data are screened in a multi-dimensional coupling manner, including: S11, performing time matching according to the acquisition time of the carbon dioxide concentration data, the panoramic image data, and the meteorological data; S12. Using a pre-trained image recognition model to identify the panoramic image data, and determining a non-traffic impact area and its corresponding panoramic image acquisition time based on the identification result; S13. Calculating the turbulence intensity of the non-traffic-affected area at the corresponding panoramic image acquisition time based on the meteorological data; S14, obtaining a road congestion index of the non-traffic-affected area at the corresponding panoramic image acquisition time, and calculating a background credibility score of the non-traffic-affected area at the corresponding panoramic image acquisition time based on the turbulence intensity and the road congestion index; S15. Filtering the carbon dioxide concentration data of the non-traffic-affected area at the corresponding panoramic image acquisition time according to the background credibility score to obtain filtered carbon dioxide concentration data; S16. Perform cluster analysis on the screened carbon dioxide concentration data, and associate the clustering results with corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data.

[0038] Specifically, in step S11, the carbon dioxide concentration data, panoramic image data, and meteorological data all have time tags, and time matching is performed based on the time tags of each data. That is, each carbon dioxide concentration data corresponds to panoramic image data and meteorological data with the same collection time, ensuring that the time points of each corresponding data remain consistent, providing a basis for subsequent analysis.

[0039] Furthermore, in step S12, the panoramic image data is recognized using a pre-trained image recognition model, wherein the image recognition model may be a DeepLabV3+ model, and the recognition results include vehicles, vegetation, and building facades.

[0040] The non-traffic impact area determined based on the recognition results and its corresponding panoramic image acquisition time and environmental information include: The pixel ratios of the vehicles, vegetation, and building facades in the panoramic image are calculated respectively, and the sampling points corresponding to the panoramic images where the pixel ratio of the vehicles is less than a preset value are determined as non-traffic impact areas. The corresponding panoramic image acquisition time, the pixel ratio of the vegetation, and the pixel ratio of the building facade are recorded.

[0041] Specifically, the pixel ratio is calculated using the following formula: ; (1) in, represents the i-th pixel of the vehicle / vegetation / building facade in the panoramic image, A represents the pixel ratio of the vehicle / vegetation / building facade in the panoramic image, and S represents the total number of pixels in the panoramic image.

[0042] Specifically, vehicles are the main source of carbon emissions, and buildings represent human activities. Therefore, buildings and vegetation can both be factors affecting the background concentration of carbon dioxide. The sampling points corresponding to the panoramic images where the pixel ratio of vehicles is less than the preset value are determined as non-traffic-affected areas. That is, at the corresponding moment, there are fewer or no vehicles at the sampling point. At this time, the obtained carbon dioxide concentration is closer to the background concentration without emission sources.

[0043] Further, refer to Figure 3 , in step S13, the meteorological data includes wind speed; Calculating the turbulence intensity of the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data, including: S13a, based on the time matching result, obtaining wind speed data of the non-traffic-affected area within a preset time period before the corresponding panoramic image acquisition moment; S13b, calculating the wind speed standard deviation within the preset time period based on the wind speed data; S13c. Calculate the turbulence intensity based on the wind speed standard deviation and the wind speed in the non-traffic affected area at the corresponding panoramic image acquisition time.

[0044] Specifically, wind speed is also an important factor affecting the concentration of carbon dioxide in the air. Based on the wind speed data in non-traffic-affected areas, the turbulence intensity is calculated for subsequent data screening.

[0045] The turbulence intensity is calculated using the following formula: ; (2) Where TI represents the turbulence intensity, σ represents the standard deviation of wind speed within a preset time period, and v represents the wind speed in the non-traffic affected area at the corresponding panoramic image acquisition time.

[0046] Furthermore, in step S14, the road congestion index of the non-traffic affected area at the corresponding panoramic image acquisition time is obtained, and the background credibility score of the non-traffic affected area at the corresponding panoramic image acquisition time is calculated based on the turbulence intensity and the road congestion index. The background credibility score is calculated using the following formula: ; (3) Among them, Wtime represents the background credibility score, P represents the road congestion index, TI represents the turbulence intensity, and Q represents the boundary layer stability coefficient of wind speed.

[0047] The boundary layer stability coefficient Q of wind speed can be solved by the Monin-Obukhov similarity theory. In the atmospheric boundary layer, the boundary layer stability coefficient can be expressed by the Monin-Obukhov length.

[0048] Furthermore, in step S15, the carbon dioxide concentration data of the non-traffic impact area at the corresponding panoramic image acquisition time is screened according to the background credibility score, specifically, the carbon dioxide concentration data with a background credibility score greater than a preset score is selected to obtain screened carbon dioxide concentration data.

[0049] Further, refer to Figure 4 In step S16, cluster analysis is performed on the screened carbon dioxide data, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data, including: S16a, performing Z-score normalization on the screened carbon dioxide data, and removing abnormal data points based on the Z-score normalization results; S16b, calculating the coefficient of variation for the remaining screened carbon dioxide data after removing the abnormal points, and selecting candidate carbon dioxide background concentration data based on the coefficient of variation; S16c, using the DBSCAN clustering algorithm to perform density analysis on the candidate carbon dioxide background concentration data, and taking the cluster center value with a density value greater than a preset density value as the clustering result; S16d. According to the time matching result, the clustering result is associated with the corresponding meteorological data, the pixel ratio of vegetation, and the pixel ratio of building facades to obtain the carbon dioxide background concentration data.

[0050] Specifically, in step S16a, firstly, the outliers are eliminated, a time window is set, and the mean of the carbon dioxide data is calculated and filtered according to the preset sliding time step. and standard deviation , and perform Z-score standardization on the data points in the time window, and remove outliers whose Z scores are greater than the preset score value.

[0051] Furthermore, in step S16b, the coefficient of variation CV is calculated for the remaining screened carbon dioxide data after removing the abnormal points. The coefficient of variation CV is the ratio of the mean to the standard deviation. The coefficient of variation CV of the remaining screened carbon dioxide data in each time window is calculated, and the window data with a coefficient of variation CV less than a preset value is retained. A preset percentage of data is selected from the retained window data as the candidate carbon dioxide background concentration data for the window.

[0052] Furthermore, in step S16c, a DBSCAN clustering algorithm is used to perform density analysis on the candidate carbon dioxide background concentration data, specifically including: For each candidate carbon dioxide background concentration data point, all data points in the neighborhood are searched according to the preset neighborhood radius, and the core point is determined based on the number of data points in the neighborhood. Starting from the core point, each data point is recursively added to the corresponding cluster.

[0053] In the final clusters, sparsely distributed clusters (possibly affected by transient emissions) are removed, and the center values of high-density clusters are retained as the final clustering results. The density judgment value can be set according to actual conditions.

[0054] Finally, in step S16d, the clustering result obtained, that is, the carbon dioxide concentration data identified as the background concentration, is associated with the pixel ratio of vegetation in the corresponding panoramic image, the pixel ratio of building facades, and the wind speed to obtain carbon dioxide background concentration data, that is, the carbon dioxide background concentration data includes the carbon dioxide concentration data identified as the background concentration, the pixel ratio of vegetation in the corresponding panoramic image, the pixel ratio of building facades, and the wind speed. The carbon dioxide background concentration data will be used as a sample set to train the deep learning model.

[0055] Furthermore, in step S2, a pre-established deep learning model is trained based on the carbon dioxide background concentration data, wherein the deep learning model may include an LSTM network and a random forest spatial model, which are used to model time series dependencies and process high-dimensional nonlinear features, respectively. The LSTM network adopts a network structure comprising an LSTM layer with 128 hidden units, a Dropout layer with a dropout rate of 0.2, and a fully connected layer with 64 neurons. Time series features are extracted through a 24-hour sliding window, and the model is trained using the Adam optimizer. The random forest model sets 200 decision trees, with a maximum tree depth of 12 layers, a minimum number of leaf node samples of 10, a feature sampling ratio of 0.8, and the importance of feature variables is evaluated by the Gini index. A 5-fold cross-validation method is used for parameter optimization during model training.

[0056] The LSTM network adaptively weights meta-features through an attention mechanism, achieving dynamic fusion of different feature information. The dynamic weight adjustment mechanism employs a multi-dimensional adaptive strategy: calculating the predictive power of each model at different time scales based on temporal correlation; assessing the applicability of models in different regions based on spatial distance; and dynamically adjusting model weights based on the magnitude of the forecast error. When significant changes in meteorological factors are detected (for example, a 24-hour change exceeding twice the historical standard deviation), the attention weight is adjusted, automatically reducing the LSTM model's contribution to 70% of its original value while simultaneously increasing the influence of the random forest model in the forecast.

[0057] Furthermore, in step S3, the navigating vehicle is controlled to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and the real-time panoramic image data is identified and analyzed to obtain environmental analysis results, including: The real-time panoramic image data is input into the image recognition model for recognition analysis to obtain the pixel ratio of vegetation at the monitoring point and the pixel ratio of the building facade.

[0058] In step S4, the real-time meteorological data and environmental analysis results are input into the background concentration simulation prediction model. The background concentration simulation prediction model makes predictions based on the pixel ratio of vegetation in the real-time panoramic image data, the pixel ratio of building facades and meteorological data to obtain the predicted background concentration of carbon dioxide at the monitoring point, where the meteorological data is wind speed.

[0059] Furthermore, in step S5, the carbon dioxide increment at the monitoring point is calculated based on the real-time carbon dioxide data and the predicted carbon dioxide background concentration at the monitoring point. The carbon dioxide increment at the monitoring point is the difference between the real-time carbon dioxide concentration and the predicted carbon dioxide background concentration at the monitoring point, that is: ; (4) in, is the real-time carbon dioxide concentration value, For the predicted background concentration of carbon dioxide, An increase in carbon dioxide.

[0060] Furthermore, the real-time meteorological data includes real-time wind speed.

[0061] refer to Figure 5 In step S6, the carbon dioxide increments at multiple monitoring points on the road are calculated based on the Gaussian plume model to obtain the road carbon emissions, including: S61. Calculate the monitoring area based on the monitoring points; S62. Obtaining a carbon dioxide emission rate of the road section corresponding to the monitoring point using the Gaussian plume model based on the area of the monitoring region and the real-time wind speed; S63, accumulating the carbon dioxide emission rates calculated at each monitoring point on the same road to obtain a total carbon dioxide emission rate for the road; S64. Calculate and obtain road carbon emissions based on the total carbon dioxide emission rate.

[0062] Specifically, in step S61, half of the distance between every two monitoring points is used as the radius, and the area of the circle with the monitoring point as the center is calculated as the area of the monitoring area.

[0063] Furthermore, in step S62, the Gaussian plume model is used to obtain the carbon dioxide emission rate of the road section corresponding to the monitoring point. The specific calculation formula is: ; (5) Among them, Aj represents the monitoring area of the walking monitoring point, σ z Calculated by the Pasquill-Gifford empirical formula, u is the real-time wind speed, H is the height of the emission source, that is, the height of the vehicle exhaust pipe, the default value is 0.3 m, Q jis the carbon dioxide emission rate of the jth road section corresponding to the monitoring point, An increase in carbon dioxide.

[0064] In step S63, the total carbon dioxide emission rate of the road is calculated using the following formula: ; (6) Among them, the m value represents the number of monitoring points on each road. It represents the length of the road section represented by each monitoring point j, and Qtotal represents the total carbon dioxide emission rate of the road.

[0065] In step S64 , the total carbon dioxide emission rate of the road is multiplied by the time coefficient to obtain the carbon emission of the road within the corresponding time.

[0066] refer to Figure 6 In some embodiments, a carbon emission calculation device based on dynamic simulation and cruise measurement is provided, including: The sample collection module 201 is used to pre-control the navigating vehicle to collect carbon dioxide concentration data, panoramic image data, and meteorological data at road sampling points and perform multi-dimensional coupling screening to obtain carbon dioxide background concentration data; A training module 202 is configured to train a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; The real-time data acquisition module 203 is used to control the navigation vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and to identify and analyze the real-time panoramic image data to obtain environmental analysis results; Prediction module 204, for inputting the real-time meteorological data and environmental analysis results into the background concentration simulation prediction model to obtain the predicted background concentration of carbon dioxide at the monitoring point; An increment calculation module 205 is configured to calculate the increment of carbon dioxide at a monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; The emission calculation module 206 is used to calculate the carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain the carbon emissions of the road.

[0067] Furthermore, the sample collection module 201 is further configured to: The carbon dioxide concentration data and meteorological data are cleaned and standardized.

[0068] Furthermore, the sample collection module 201 performs multi-dimensional coupled screening on the carbon dioxide concentration data, the panoramic image data, and the meteorological data, including: Performing time matching based on the acquisition time of the carbon dioxide concentration data, panoramic image data, and meteorological data; Using a pre-trained image recognition model to identify the panoramic image data, and determining the non-traffic impact area and its corresponding panoramic image acquisition time and environmental information based on the recognition result; Calculating the turbulence intensity in the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data; Obtaining a road congestion index for the non-traffic affected area at the corresponding panoramic image acquisition time, and calculating a background credibility score for the non-traffic affected area at the corresponding panoramic image acquisition time based on the turbulence intensity and the road congestion index; screening the carbon dioxide concentration data of the non-traffic impact area at the corresponding panoramic image acquisition time according to the background credibility score to obtain screened carbon dioxide concentration data; The screened carbon dioxide concentration data is clustered and analyzed, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data.

[0069] Furthermore, the recognition results include vehicles, vegetation, and building facades; The sample collection module 201 determines the non-traffic impact area and its corresponding panoramic image collection time according to the recognition result, including: The pixel ratios of the vehicles, vegetation, and building facades in the panoramic image are calculated respectively, and the sampling points corresponding to the panoramic images where the pixel ratio of the vehicles is less than a preset value are determined as non-traffic impact areas. The corresponding panoramic image acquisition time, the pixel ratio of the vegetation, and the pixel ratio of the building facade are recorded.

[0070] Furthermore, the meteorological data includes wind speed; The sample collection module 201 calculates the turbulence intensity of the non-traffic affected area at the corresponding panoramic image collection time based on the meteorological data, including: According to the time matching result, the wind speed data of the non-traffic affected area within a preset time period before the corresponding panoramic image acquisition time is obtained; Calculating the wind speed standard deviation within the preset time period based on the wind speed data; The turbulence intensity is obtained by calculation according to the wind speed standard deviation and the wind speed of the non-traffic affected area at the corresponding panoramic image acquisition moment.

[0071] Furthermore, the sample collection module 201 performs cluster analysis on the screened carbon dioxide data, and associates the clustering results with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data, including: The screened carbon dioxide data were Z-score normalized, and abnormal data points were removed based on the Z-score normalization results; Calculate the coefficient of variation for the remaining screened carbon dioxide data after removing abnormal points, and select candidate carbon dioxide background concentration data based on the coefficient of variation; Using the DBSCAN clustering algorithm to perform density analysis on the candidate carbon dioxide background concentration data, and taking the cluster center value with a density value greater than a preset density value as the clustering result; According to the time matching result, the clustering result is associated with the corresponding meteorological data, the pixel ratio of vegetation, and the pixel ratio of building facades to obtain the carbon dioxide background concentration data.

[0072] Furthermore, the real-time data acquisition module 203 performs recognition analysis on the real-time panoramic image data to obtain an environmental analysis result, including: The real-time panoramic image data is input into the image recognition model for recognition analysis to obtain the pixel ratio of vegetation at the monitoring point and the pixel ratio of the building facade.

[0073] Furthermore, the carbon dioxide increment at the monitoring point is the difference between the real-time carbon dioxide concentration and the predicted background concentration of carbon dioxide at the monitoring point.

[0074] Furthermore, the real-time meteorological data includes real-time wind speed; The emission calculation module 206 calculates the carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain the road carbon emissions, including: Calculate the monitoring area based on the monitoring points; According to the area of the monitoring area and the real-time wind speed, the Gaussian plume model is used to obtain the carbon dioxide emission rate of the road section corresponding to the monitoring point; The carbon dioxide emission rates calculated at each monitoring point on the same road are accumulated to obtain the total carbon dioxide emission rate of the road; The road carbon emissions are calculated based on the total carbon dioxide emission rate.

[0075] The carbon emission calculation method and device based on dynamic simulation and cruise measurement provided in the above embodiment have at least the following beneficial effects: (1) The background concentration of carbon dioxide is taken into account in the calculation of road carbon emissions. The carbon dioxide increment is calculated based on the background concentration of carbon dioxide, and the carbon emissions are calculated based on the carbon dioxide increment, so that the obtained carbon dioxide emissions are more accurate; (2) The model is trained based on the data obtained from the actual flight measurement to obtain a background concentration simulation prediction model. The background concentration of carbon dioxide is predicted based on this background concentration prediction model. This model is applicable to various road scenes and different traffic conditions. It does not require the establishment of observation stations, effectively reducing costs and having wide applicability. (3) The background concentration simulation prediction model takes into account the impact of wind speed, buildings, and vegetation on the background concentration of carbon dioxide, effectively improving the accuracy of the background concentration prediction of carbon dioxide, and thus improving the accuracy of subsequent carbon dioxide emission calculations; (4) Temporal and spatial matching and screening of carbon dioxide concentration data, panoramic image data, and meteorological data used for model training can effectively improve the accuracy of the background concentration simulation prediction model.

[0076] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A carbon emission calculation method based on dynamic simulation and on-board measurement, characterized in that: include: Control the roving vehicle in advance to collect carbon dioxide concentration data, panoramic image data, and meteorological data at road sampling points for multi-dimensional coupling screening to obtain carbon dioxide background concentration data; Training a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; Controlling the navigating vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and performing identification and analysis on the real-time panoramic image data to obtain environmental analysis results; Inputting the real-time meteorological data and environmental analysis results into the background concentration simulation prediction model to obtain the predicted background concentration of carbon dioxide at the monitoring point; Calculating the carbon dioxide increment at the monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; The carbon dioxide increments at multiple monitoring points on the road were calculated based on the Gaussian plume model to obtain the road carbon emissions.

2. The method according to claim 1, characterized in that After collecting carbon dioxide concentration data and meteorological data at road sampling points, it also includes: The carbon dioxide concentration data and meteorological data are cleaned and standardized.

3. The method according to claim 1, characterized in that Multi-dimensional coupling screening of carbon dioxide concentration data, panoramic image data, and meteorological data, including: Performing time matching based on the acquisition time of the carbon dioxide concentration data, panoramic image data, and meteorological data; Using a pre-trained image recognition model to identify the panoramic image data, and determining the non-traffic impact area and its corresponding panoramic image acquisition time and environmental information based on the recognition result; Calculating the turbulence intensity in the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data; Obtaining a road congestion index for the non-traffic affected area at the corresponding panoramic image acquisition time, and calculating a background credibility score for the non-traffic affected area at the corresponding panoramic image acquisition time based on the turbulence intensity and the road congestion index; screening the carbon dioxide concentration data of the non-traffic impact area at the corresponding panoramic image acquisition time according to the background credibility score to obtain screened carbon dioxide concentration data; The screened carbon dioxide concentration data is clustered and analyzed, and the clustering results are associated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data.

4. The method according to claim 3, characterized in that The recognition results include vehicles, vegetation, and building facades; The non-traffic impact area determined based on the recognition results and its corresponding panoramic image acquisition time and environmental information include: The pixel ratios of the vehicles, vegetation, and building facades in the panoramic image are calculated respectively, and the sampling points corresponding to the panoramic images where the pixel ratio of the vehicles is less than a preset value are determined as non-traffic impact areas. The corresponding panoramic image acquisition time, the pixel ratio of the vegetation, and the pixel ratio of the building facade are recorded.

5. The method according to claim 3, characterized in that The meteorological data includes wind speed; Calculating the turbulence intensity of the non-traffic affected area at the corresponding panoramic image acquisition time based on the meteorological data, including: According to the time matching result, the wind speed data of the non-traffic affected area within a preset time period before the corresponding panoramic image acquisition time is obtained; Calculating the wind speed standard deviation within the preset time period based on the wind speed data; The turbulence intensity is obtained by calculation according to the wind speed standard deviation and the wind speed of the non-traffic affected area at the corresponding panoramic image acquisition moment.

6. The method according to claim 4, characterized in that The screened carbon dioxide data is clustered and analyzed, and the clustering results are correlated with the corresponding environmental information and meteorological data to obtain the carbon dioxide background concentration data, including: The screened carbon dioxide data were Z-score normalized, and abnormal data points were removed based on the Z-score normalization results; Calculate the coefficient of variation for the remaining screened carbon dioxide data after removing abnormal points, and select candidate carbon dioxide background concentration data based on the coefficient of variation; Using the DBSCAN clustering algorithm to perform density analysis on the candidate carbon dioxide background concentration data, and taking the cluster center value with a density value greater than a preset density value as the clustering result; According to the time matching result, the clustering result is associated with the corresponding meteorological data, the pixel ratio of vegetation, and the pixel ratio of building facades to obtain the carbon dioxide background concentration data.

7. The method according to claim 4, characterized in that Performing identification and analysis on the real-time panoramic image data to obtain environmental analysis results includes: The real-time panoramic image data is input into the image recognition model for recognition analysis to obtain the pixel ratio of vegetation at the monitoring point and the pixel ratio of the building facade.

8. The method according to claim 1, characterized in that The carbon dioxide increment at a monitoring point is the difference between the real-time carbon dioxide concentration and the predicted background concentration of carbon dioxide at the monitoring point.

9. The method according to claim 1, characterized in that The real-time meteorological data includes real-time wind speed; Based on the Gaussian plume model, the carbon dioxide increments at multiple monitoring points on the road are integrated in time and space to obtain the road carbon emissions, including: Calculate the monitoring area based on the monitoring points; According to the area of the monitoring area and the real-time wind speed, the Gaussian plume model is used to obtain the carbon dioxide emission rate of the road section corresponding to the monitoring point; The carbon dioxide emission rates calculated at each monitoring point on the same road are accumulated to obtain the total carbon dioxide emission rate of the road; The road carbon emissions are calculated based on the total carbon dioxide emission rate.

10. A carbon emission calculation device based on dynamic simulation and actual measurement, characterized in that: include: The sample collection module is used to pre-control the navigation vehicle to collect carbon dioxide concentration data, panoramic image data and meteorological data at road sampling points and perform multi-dimensional coupling screening to obtain carbon dioxide background concentration data; A training module, configured to train a pre-established deep learning model based on the carbon dioxide background concentration data to obtain a background concentration simulation prediction model; A real-time data acquisition module is used to control the navigating vehicle to collect real-time carbon dioxide data, real-time panoramic image data, and real-time meteorological data from road monitoring points, and input the data into the background concentration simulation prediction model to obtain the predicted background concentration of carbon dioxide at the monitoring points; An increment calculation module, configured to calculate the increment of carbon dioxide at a monitoring point based on the real-time carbon dioxide data and the predicted background concentration of carbon dioxide at the monitoring point; The emission calculation module is used to calculate the carbon dioxide increments at multiple monitoring points on the road based on the Gaussian plume model to obtain the road carbon emissions.

Images