Urban road CO2 concentration hourly simulation prediction method based on multi-source data fusion and integration algorithm driving

Abstract

The invention discloses an urban road CO2 concentration hourly simulation prediction method based on multi-source data fusion and integration algorithm driving, and the method comprises the steps: integrating road underway monitoring CO2 concentration data, road information and traffic data, meteorological data, human geographic data and other multi-source heterogeneous data; a CO2 concentration simulation model based on an integrated algorithm of a random forest (RF) model and a long short-term memory (LSTM) network model is constructed. According to the method, when the CO2 concentration of the urban road is simulated, the problem that a single model can not simulate accurately possibly is considered, an integration algorithm is adopted, the random forest model and the long and short-term memory network model are combined and integrated, and the optimal model is selected to simulate the CO2 concentration of the road, so that the simulation accuracy is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of CO2 concentration monitoring, and in particular to an hourly simulation and prediction method for urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving. Background Art

[0002] In recent years, significant technological progress has been made in the field of urban environmental carbon monitoring. One of the important technical means is the cruise monitoring technology based on a vehicle mobile platform, which is used to monitor the atmospheric CO2 concentration on urban roads in real time. This technical solution can obtain key data on CO2 concentration and vehicle emissions during vehicle driving by installing an atmospheric CO2 concentration monitoring system on the vehicle. This method can obtain accurate CO2 concentration data in real time.

[0003] Although this technology can obtain high-precision CO2 concentration data in real time, it also has some limitations, such as huge costs. In addition, a cruise monitoring vehicle can usually only monitor a single road in a short period of time, which causes the temporal and spatial discontinuity of data collection and cannot fully and completely obtain continuous CO2 concentration information on all roads in the entire city at all times.

[0004] In view of the above problems, the present invention provides an hourly simulation and prediction method of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving, which realizes continuous dynamic simulation and real-time prediction of CO2 concentration in urban road network. Summary of the invention

[0005] The present invention provides an hourly simulation and prediction method for urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving, which realizes continuous dynamic simulation and real-time prediction of urban road network CO2 concentration.

[0006] The purpose of the present invention is to provide an hourly simulation prediction method for urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving, and the specific steps of the method are as follows:

[0007] Step 1: Build a road cruise monitoring system, design a cruise monitoring plan, and obtain CO2 concentration data, vehicle speed, road view images, wind speed, wind direction, temperature, humidity, and monitoring data corresponding time and geographic coordinate information for some urban roads;

[0008] Step 2: Using the CO2 concentration data monitored by navigation on some roads as the basic data, perform data cleaning and preprocessing;

[0009] Step 3: Combined with the corresponding road scene images, use image recognition algorithms such as the YOLO-v8 algorithm to extract road vehicle driving conditions and vehicle type information;

[0010] Step 4: Collect road attribute information, traffic conditions, meteorological data, atmospheric background CO2 concentration observation data, land use data, POI data, and population distribution data, match them with road CO2 concentration monitoring data and vehicle information extraction results according to the spatiotemporal matching algorithm, and construct a multi-source input feature variable data set;

[0011] Step 5: Take the road CO2 concentration data as the true value and other data as input feature variables, standardize the data, divide the data into training set and validation set;

[0012] Step 6: Based on the collected multi-source heterogeneous data, a model for hourly simulation and prediction of urban road CO2 concentration based on the random forest model and long short-term memory network integration algorithm is constructed, and the model is trained using the training set;

[0013] Step 7: Use the validation set to verify and evaluate the accuracy of the model;

[0014] Step 8: Based on the trained urban road CO2 concentration simulation prediction model, simulate the carbon concentration of each road in the city hour by hour.

[0015] Furthermore, the specific steps of step 1 are as follows:

[0016] Step 11: By installing greenhouse gas monitoring equipment, GPS system, panoramic camera, and meteorological sensor system on vehicles, a cruise monitoring system for atmospheric CO2 on urban roads was constructed. Based on the cruise monitoring system, the CO2 concentration on urban roads, road scene images, vehicle speed, wind speed, wind direction, temperature, humidity, and geographic coordinate information of monitoring data were obtained;

[0017] Step 12. Subsequently, taking into account the overall planning information of the urban road network, road attribute characteristics, time attribute characteristics, distribution information of key urban pollution sources, terrain and land use information, a cruise monitoring route plan for typical urban roads is designed. The cruise monitoring route plan includes the cruise monitoring time range, the cruise monitoring route, key monitoring sections, and approximate time information.

[0018] Furthermore, the specific steps of step 2 are as follows:

[0019] Step 21, firstly, the monitoring data of the road CO2 concentration cruise monitoring system is processed. Since different sensors collect different information, the CO2 concentration monitoring data, GPS data, meteorological data, and panoramic image data are temporally and spatially matched according to the data collection time information;

[0020] Step 22, the cruise monitoring CO2 concentration data has a resolution of 2-10 seconds. Affected by the vehicle status and road status, the data fluctuates greatly. At the same time, considering the time-space matching with multiple sensor data, the original monitoring data is time-resampled and averaged to minutes. Specifically, a window extending 30 seconds before and after is set for each minute time node. Within this window, all monitoring data with original time resolution are extracted and the arithmetic average is calculated to obtain the time resampling result at the current moment. The mathematical expression of data resampling processing is:

[0021]

[0022] Among them, CO 2(t) Represents the CO2 concentration value after resampling at time t, CO 2(t-1) , CO 2(t-2) and CO 2(t+n) It represents the n CO2 concentration monitoring values ​​collected at the original time resolution within a time window of 30 seconds before and after time t, where n is the number of monitoring data points in the time window.

[0023] Furthermore, the specific steps of step 3 are:

[0024] Step 31: Based on the collected road scene images, the deep learning model technology is used to identify the vehicles around the cruise monitoring vehicle. A certain number of image frames are randomly selected from the road scene image sequence collected by the cruise monitoring as training samples, and these sample images are manually annotated to mark the vehicle information objects in each frame, and the annotation results are saved as a label file in YOLO format;

[0025] Step 32: Use the labeled sample data to train the YOLOv8 model, learn the feature representation of the vehicle object in the image through the convolutional neural network, and predict the location and category of the vehicle based on prior knowledge. During the training process, optimize the model parameters through the back propagation algorithm so that it can accurately identify the vehicle object in the image;

[0026] Step 33: Apply the trained YOLOv8 model to the panoramic image sequence collected by cruise monitoring, and perform vehicle detection on each frame of the image. The model performs dense sampling on the image by means of a sliding window, and performs feature extraction and vehicle judgment on each window area. By setting a confidence threshold, most of the false detection results are filtered out to obtain the vehicle detection result.

[0027] Step 34: According to the detection results of the YOLOv8 model, count the vehicles identified in each frame of the image to obtain the number of vehicles at the current moment, match the number of vehicles corresponding to each frame of the image with the CO2 concentration monitoring data according to the monitoring time, and obtain a matching data pair of vehicle flow and CO2 concentration.

[0028] Furthermore, the specific steps of step 4 are:

[0029] Step 41: for the urban POI data and high spatial resolution population distribution data, a buffer zone with a radius of 500m is set for each CO2 concentration monitoring point, and the total number of POIs and the total number of people at the monitoring point within the buffer zone are counted as the POI and population data values ​​corresponding to the monitoring point;

[0030] Step 42: for high spatial resolution land use data, for each CO2 concentration monitoring point, a buffer zone with a radius of 500m is set, and the area proportion of industrial land, commercial service land, residential land, forest land, grassland, and water land within the buffer zone is counted. The proportion of each land use type corresponding to the monitoring point is used as the land use type data corresponding to the monitoring point;

[0031] Step 43: Match the collected number of road lanes and road width according to the latitude and longitude coordinates of the road monitoring points with the road section information;

[0032] Step 44: Match the road congestion data and the average vehicle speed data one by one according to the latitude and longitude coordinates and monitoring time of the road monitoring points;

[0033] Step 45: Based on the high-temporal and spatial resolution surface meteorological element data and atmospheric background CO2 concentration data, a grid with the same spatial resolution as the road monitoring data is defined, and the grid position of the road monitoring point is determined by the longitude and latitude coordinates of the road monitoring point. Combined with the road CO2 concentration data collection time, the monitoring point data and meteorological elements are matched in space and time to complete the construction of the multi-source input feature data set.

[0034] Furthermore, the specific steps of step 5 are:

[0035] Step 51, select the maximum and minimum value standardization to standardize the data: subtract the minimum value from the data, and then divide it by the value range of the data, so that the data value is between [0,1]. The formula is:

[0036]

[0037] Among them, x min Represents the minimum value in the original data set, x max Represents the maximum value in the original data set, x i is the original value, y i is the value after standardization;

[0038] Step 52: randomly divide the standardized feature data set and the corresponding true value of CO2 concentration into a training set and a validation set in a ratio of 8:2, wherein the training set is used for parameter learning and optimization of the model, and the validation set is used to evaluate the generalization performance and prediction accuracy of the model.

[0039] Furthermore, the specific steps of step 6 are:

[0040] Step 61: Input the training set data, use the grid search method to tune the key hyperparameters of the random forest model, and determine the hyperparameters that need to be optimized: the number of decision trees, the maximum number of features, the maximum tree depth, the minimum number of split samples, and the minimum number of node samples;

[0041] Step 62, set the value range of the hyperparameter to be optimized: for each hyperparameter to be optimized, set a set of discrete values ​​as the search space, the number of decision trees is set to generate a set of discrete values ​​in the range of [300,800] with a step size of 50; the maximum number of features is set to two options [sqrt,auto]; the maximum tree depth is set to generate a set of discrete values ​​in the range of [10,100] with a step size of 10; the minimum number of split samples is set to three discrete values ​​of [2,5,10]; the minimum number of node samples is set to three discrete values ​​of [1,2,4];

[0042] Step 63, generating hyperparameter combinations: generating all possible hyperparameter combinations according to the set hyperparameter value range;

[0043] Step 64, train and evaluate the model: For each hyperparameter combination, use the training set data and the five-fold cross-validation method to train the random forest model, and use the mean square error as an indicator to evaluate the model performance. The hyperparameter combination with the smallest mean square error is the hyperparameter combination with the best performance. The formula for calculating the mean square error is:

[0044]

[0045] Among them, y i represents the observed value of CO2 concentration, represents the predicted value of CO2 concentration, and n represents the number of matching data pairs;

[0046] Step 65, construct a long short-term memory network model, including an input layer, a hidden layer, and an output layer. The hidden layer includes two LSTM layers and a fully connected layer. The input layer data shape of the LSTM model is in a three-dimensional format. The number of samples represents the number of samples in the training data set. The time step is set to a 1-hour sliding time window, representing the CO2 concentration feature sequence of the previous hour to predict the CO2 concentration of the hour. The number of features represents the input feature vector. In terms of parameter settings, the number of training rounds for each hidden layer is set to 20, the batch size is 64, the optimizer is adaptive moment estimation, the initial learning rate is 0.001, and the loss function is the mean absolute error. After the model is deployed, the training set data is input into the model for model training.

[0047] Furthermore, the specific steps of step 7 are:

[0048] Step 71: Input the validation set data and select the coefficient of determination and root mean square error as the evaluation indicators of the model. The coefficient of determination is used to measure the model's ability to explain the data variance, with a value range of 0 to 1, and R 2 The closer it is to 1, the better the model fits. The root square error is the square root of the mean of the square of the difference between the predicted value and the true value. The smaller the RMSE, the better the predictive ability of the model. 2 And the RMSE calculation formula is:

[0049]

[0050] Among them, p i represents the observed value of CO2 concentration, o i represents the predicted value of CO2 concentration, represents the arithmetic mean of the CO2 concentration observations, and n represents the number of matching data pairs.

[0051] Furthermore, the specific steps of step 8 are:

[0052] Step 81: Based on the trained and verified random forest model and long short-term memory network model for simulating urban road CO2 concentration, the model with the largest determination coefficient and the smallest root mean square error is selected as the final model, and the feature data set is input into the model to realize hourly simulation of CO2 concentration on major urban roads.

[0053] The present invention has the following advantages: the present invention integrates multi-source heterogeneous data such as road navigation monitoring CO2 concentration data, road information and traffic data, meteorological data and human geography data, and constructs a CO2 concentration simulation model based on the integrated algorithm of random forest (RF) model and long short-term memory network (LSTM) model, thereby realizing continuous dynamic simulation and real-time prediction of CO2 concentration in urban road networks.

[0054] The present invention fully utilizes the value of monitoring data and forms an integrated algorithm-driven urban road CO2 concentration simulation prediction method for multi-source data. This method is extensible in both time and space dimensions, so it can more comprehensively simulate the changes in urban road CO2 concentration and deeply explore the dynamic trend of urban road carbon emissions. This innovation not only significantly reduces the cost of urban road carbon emission monitoring, but also provides substantial data support for urban low-carbon development and achieves sustainable development goals.

[0055] When simulating the CO2 concentration on urban roads, the present invention takes into account the problem that a single model may not be able to simulate the CO2 concentration accurately. An integrated algorithm is adopted to combine and integrate the random forest model with the long short-term memory network model, and the optimal model is selected to simulate the CO2 concentration on the road, thereby effectively improving the accuracy of the simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is the technical roadmap of the present invention;

[0057] Figure 2 This is a simulation result diagram of the present invention. DETAILED DESCRIPTION

[0058] The present invention provides an hourly simulation prediction method for urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving, and the specific steps of the method are as follows:

[0059] Step 1: Build a road cruise monitoring system, design a cruise monitoring plan, and obtain CO2 concentration data, vehicle speed, road view images, wind speed, wind direction, temperature, humidity, and monitoring data corresponding time and geographic coordinate information for some urban roads;

[0060] Step 2: Using the CO2 concentration data monitored by navigation on some roads as the basic data, perform data cleaning and preprocessing;

[0061] Step 3: Combined with the corresponding road scene images, use image recognition algorithms such as the YOLO-v8 algorithm to extract road vehicle driving conditions and vehicle type information;

[0062] Step 4: Collect road attribute information, traffic conditions, meteorological data, atmospheric background CO2 concentration observation data, land use data, POI data, and population distribution data, match them with road CO2 concentration monitoring data and vehicle information extraction results according to the spatiotemporal matching algorithm, and construct a multi-source input feature variable data set;

[0063] Step 5: Take the road CO2 concentration data as the true value and other data as input feature variables, standardize the data, divide the data into training set and validation set;

[0064] Step 6: Based on the collected multi-source heterogeneous data, a model for hourly simulation and prediction of urban road CO2 concentration based on the random forest model and long short-term memory network integration algorithm is constructed, and the model is trained using the training set;

[0065] Step 7: Use the validation set to verify and evaluate the accuracy of the model;

[0066] Step 8: Based on the trained urban road CO2 concentration simulation prediction model, simulate the carbon concentration of each road in the city hour by hour.

[0067] In this embodiment, the specific steps of step 1 are as follows:

[0068] Step 11: By installing greenhouse gas monitoring equipment, GPS system, panoramic camera, and meteorological sensor system on vehicles, a cruise monitoring system for atmospheric CO2 on urban roads was constructed. Based on the cruise monitoring system, the CO2 concentration on urban roads, road scene images, vehicle speed, wind speed, wind direction, temperature, humidity, and geographic coordinate information of monitoring data were obtained;

[0069] Step 12. Subsequently, taking into account the overall planning information of the urban road network, road attribute characteristics, time attribute characteristics, distribution information of key urban pollution sources, terrain and land use information, a cruise monitoring route plan for typical urban roads is designed. The cruise monitoring route plan includes the cruise monitoring time range, the cruise monitoring route, key monitoring sections, and approximate time information.

[0070] In this embodiment, the specific steps of step 2 are as follows:

[0071] Step 21, firstly, the monitoring data of the road CO2 concentration cruise monitoring system is processed. Since different sensors collect different information, the CO2 concentration monitoring data, GPS data, meteorological data, and panoramic image data are temporally and spatially matched according to the data collection time information;

[0072] Step 22, the cruise monitoring CO2 concentration data has a resolution of 2-10 seconds. Affected by the vehicle status and road status, the data fluctuates greatly. At the same time, considering the time-space matching with multiple sensor data, the original monitoring data is time-resampled and averaged to minutes. Specifically, a window extending 30 seconds before and after is set for each minute time node. Within this window, all monitoring data with original time resolution are extracted and the arithmetic average is calculated to obtain the time resampling result at the current moment. The mathematical expression of data resampling processing is:

[0073]

[0074] Among them, CO 2(t) Represents the CO2 concentration value after resampling at time t, CO 2(t-1), CO 2(t-2) and CO 2(t+n) It represents the n CO2 concentration monitoring values ​​collected at the original time resolution within a time window of 30 seconds before and after time t, where n is the number of monitoring data points in the time window.

[0075] In this embodiment, the specific steps of step 3 are:

[0076] Step 31: Based on the collected road scene images, the deep learning model technology is used to identify the vehicles around the cruise monitoring vehicle. A certain number of image frames are randomly selected from the road scene image sequence collected by the cruise monitoring as training samples, and these sample images are manually annotated to mark the vehicle information objects in each frame, and the annotation results are saved as a label file in YOLO format;

[0077] Step 32: Use the labeled sample data to train the YOLOv8 model, learn the feature representation of the vehicle object in the image through the convolutional neural network, and predict the location and category of the vehicle based on prior knowledge. During the training process, optimize the model parameters through the back propagation algorithm so that it can accurately identify the vehicle object in the image;

[0078] Step 33: Apply the trained YOLOv8 model to the panoramic image sequence collected by cruise monitoring, and perform vehicle detection on each frame of the image. The model performs dense sampling on the image by means of a sliding window, and performs feature extraction and vehicle judgment on each window area. By setting a confidence threshold, most of the false detection results are filtered out to obtain the vehicle detection result.

[0079] Step 34: According to the detection results of the YOLOv8 model, count the vehicles identified in each frame of the image to obtain the number of vehicles at the current moment, match the number of vehicles corresponding to each frame of the image with the CO2 concentration monitoring data according to the monitoring time, and obtain a matching data pair of vehicle flow and CO2 concentration.

[0080] In this embodiment, the specific steps of step 4 are:

[0081] Step 41: for the urban POI data and high spatial resolution population distribution data, a buffer zone with a radius of 500m is set for each CO2 concentration monitoring point, and the total number of POIs and the total number of people at the monitoring point within the buffer zone are counted as the POI and population data values ​​corresponding to the monitoring point;

[0082] Step 42: for high spatial resolution land use data, for each CO2 concentration monitoring point, a buffer zone with a radius of 500m is set, and the area proportion of industrial land, commercial service land, residential land, forest land, grassland, and water land within the buffer zone is counted. The proportion of each land use type corresponding to the monitoring point is used as the land use type data corresponding to the monitoring point;

[0083] Step 43: Match the collected number of road lanes and road width according to the latitude and longitude coordinates of the road monitoring points with the road section information;

[0084] Step 44: Match the road congestion data and the average vehicle speed data one by one according to the latitude and longitude coordinates and monitoring time of the road monitoring points;

[0085] Step 45: Based on the high-temporal and spatial resolution surface meteorological element data and atmospheric background CO2 concentration data, a grid with the same spatial resolution as the road monitoring data is defined, and the grid position of the road monitoring point is determined by the longitude and latitude coordinates of the road monitoring point. Combined with the road CO2 concentration data collection time, the monitoring point data and meteorological elements are matched in space and time to complete the construction of the multi-source input feature data set.

[0086] In this embodiment, the specific steps of step 5 are:

[0087] Step 51, select the maximum and minimum value standardization to standardize the data: subtract the minimum value from the data, and then divide it by the value range of the data, so that the data value is between [0,1]. The formula is:

[0088]

[0089] Among them, x min Represents the minimum value in the original data set, x max Represents the maximum value in the original data set, x i is the original value, y i is the value after standardization;

[0090] Step 52: randomly divide the standardized feature data set and the corresponding true value of CO2 concentration into a training set and a validation set in a ratio of 8:2, wherein the training set is used for parameter learning and optimization of the model, and the validation set is used to evaluate the generalization performance and prediction accuracy of the model.

[0091] In this embodiment, the specific steps of step 6 are:

[0092] Step 61: Input the training set data, use the grid search method to tune the key hyperparameters of the random forest model, and determine the hyperparameters that need to be optimized: the number of decision trees, the maximum number of features, the maximum tree depth, the minimum number of split samples, and the minimum number of node samples;

[0093] Step 62, set the value range of the hyperparameter to be optimized: for each hyperparameter to be optimized, set a set of discrete values ​​as the search space, the number of decision trees is set to generate a set of discrete values ​​in the range of [300,800] with a step size of 50; the maximum number of features is set to two options [sqrt,auto]; the maximum tree depth is set to generate a set of discrete values ​​in the range of [10,100] with a step size of 10; the minimum number of split samples is set to three discrete values ​​of [2,5,10]; the minimum number of node samples is set to three discrete values ​​of [1,2,4];

[0094] Step 63, generating hyperparameter combinations: generating all possible hyperparameter combinations according to the set hyperparameter value range;

[0095] Step 64, train and evaluate the model: For each hyperparameter combination, use the training set data and the five-fold cross-validation method to train the random forest model, and use the mean square error as an indicator to evaluate the model performance. The hyperparameter combination with the smallest mean square error is the hyperparameter combination with the best performance. The formula for calculating the mean square error is:

[0096]

[0097] Among them, y i represents the observed value of CO2 concentration, represents the predicted value of CO2 concentration, and n represents the number of matching data pairs;

[0098] Step 65, construct a long short-term memory network model, including an input layer, a hidden layer, and an output layer. The hidden layer includes two LSTM layers and a fully connected layer. The input layer data shape of the LSTM model is in a three-dimensional format. The number of samples represents the number of samples in the training data set. The time step is set to a 1-hour sliding time window, representing the CO2 concentration feature sequence of the previous hour to predict the CO2 concentration of the hour. The number of features represents the input feature vector. In terms of parameter settings, the number of training rounds for each hidden layer is set to 20, the batch size is 64, the optimizer is adaptive moment estimation, the initial learning rate is 0.001, and the loss function is the mean absolute error. After the model is deployed, the training set data is input into the model for model training.

[0099] In this embodiment, the specific steps of step 7 are:

[0100] Step 71: Input the validation set data and select the coefficient of determination and root mean square error as the evaluation indicators of the model. The coefficient of determination is used to measure the model's ability to explain the data variance, with a value range of 0 to 1, and R 2 The closer it is to 1, the better the model fits. The root square error is the square root of the mean of the square of the difference between the predicted value and the true value. The smaller the RMSE, the better the predictive ability of the model. 2 And the RMSE calculation formula is:

[0101]

[0102] Among them, p i represents the observed value of CO2 concentration, o i represents the predicted value of CO2 concentration, represents the arithmetic mean of the CO2 concentration observations, and n represents the number of matching data pairs.

[0103] In this embodiment, the specific steps of step 8 are:

[0104] Step 81: Based on the trained and verified random forest model and long short-term memory network model for simulating urban road CO2 concentration, the model with the largest determination coefficient and the smallest root mean square error is selected as the final model, and the feature data set is input into the model to realize hourly simulation of CO2 concentration on major urban roads.

[0105] Although the specific embodiments of the present invention are described in detail in conjunction with the accompanying drawings, it should not be understood as limiting the scope of protection of the present invention. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still belong to the scope of protection of the present invention.

Claims

1. A method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm, characterized by: The specific steps of the method are as follows: Step 1: Build a road cruise monitoring system, design a cruise monitoring plan, and obtain CO2 concentration data, vehicle speed, road view images, wind speed, wind direction, temperature, humidity, and monitoring data corresponding time and geographic coordinate information for some urban roads; Step 2: Using the CO2 concentration data monitored by navigation on some roads as the basic data, perform data cleaning and preprocessing; Step 3: Combined with the corresponding road scene images, use image recognition algorithms such as the YOLO-v8 algorithm to extract road vehicle driving conditions and vehicle type information; Step 4: Collect road attribute information, traffic conditions, meteorological data, atmospheric background CO2 concentration observation data, land use data, POI data, and population distribution data, match them with road CO2 concentration monitoring data and vehicle information extraction results according to the spatiotemporal matching algorithm, and construct a multi-source input feature variable data set; Step 5: Take the road CO2 concentration data as the true value and other data as input feature variables, standardize the data, divide the data into training set and validation set; Step 6: Based on the collected multi-source heterogeneous data, a model for hourly simulation and prediction of urban road CO2 concentration based on the random forest model and long short-term memory network integration algorithm is constructed, and the model is trained using the training set; Step 7: Use the validation set to verify and evaluate the accuracy of the model; Step 8: Based on the trained urban road CO2 concentration simulation prediction model, simulate the carbon concentration of each road in the city hour by hour.

2. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1 is characterized by: The specific steps of step 1 are as follows: Step 11: By installing greenhouse gas monitoring equipment, GPS system, panoramic camera, and meteorological sensor system on vehicles, a cruise monitoring system for atmospheric CO2 on urban roads was constructed. Based on the cruise monitoring system, the CO2 concentration on urban roads, road scene images, vehicle speed, wind speed, wind direction, temperature, humidity, and geographic coordinate information of monitoring data were obtained; Step 12. Subsequently, taking into account the overall planning information of the urban road network, road attribute characteristics, time attribute characteristics, distribution information of key urban pollution sources, terrain and land use information, a cruise monitoring route plan for typical urban roads is designed. The cruise monitoring route plan includes the cruise monitoring time range, the cruise monitoring route, key monitoring sections, and approximate time information.

3. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 2 are as follows: Step 21, firstly, the monitoring data of the road CO2 concentration cruise monitoring system is processed. Since different sensors collect different information, the CO2 concentration monitoring data, GPS data, meteorological data, and panoramic image data are temporally and spatially matched according to the data collection time information; Step 22, the cruise monitoring CO2 concentration data has a resolution of 2-10 seconds. Affected by the vehicle status and road status, the data fluctuates greatly. At the same time, considering the time-space matching with multiple sensor data, the original monitoring data is time-resampled and averaged to minutes. Specifically, a window extending 30 seconds before and after is set for each minute time node. Within this window, all monitoring data with original time resolution are extracted and the arithmetic average is calculated to obtain the time resampling result at the current moment. The mathematical expression of data resampling processing is: in, Represents the CO2 concentration value after resampling at time t, , and It represents the n CO2 concentration monitoring values ​​collected at the original time resolution within a time window of 30 seconds before and after time t, where n is the number of monitoring data points in the time window.

4. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 3 are: Step 31: Based on the collected road scene images, the deep learning model technology is used to identify the vehicles around the cruise monitoring vehicle. A certain number of image frames are randomly selected from the road scene image sequence collected by the cruise monitoring as training samples, and these sample images are manually annotated to mark the vehicle information objects in each frame, and the annotation results are saved as a label file in YOLO format; Step 32: Use the labeled sample data to train the YOLOv8 model, learn the feature representation of the vehicle object in the image through the convolutional neural network, and predict the location and category of the vehicle based on prior knowledge. During the training process, optimize the model parameters through the back propagation algorithm so that it can accurately identify the vehicle object in the image; Step 33: Apply the trained YOLOv8 model to the panoramic image sequence collected by cruise monitoring, and perform vehicle detection on each frame of the image. The model performs dense sampling on the image by means of a sliding window, and performs feature extraction and vehicle judgment on each window area. By setting a confidence threshold, most of the false detection results are filtered out to obtain the vehicle detection result. Step 34: According to the detection results of the YOLOv8 model, count the vehicles identified in each frame of the image to obtain the number of vehicles at the current moment, match the number of vehicles corresponding to each frame of the image with the CO2 concentration monitoring data according to the monitoring time, and obtain a matching data pair of vehicle flow and CO2 concentration.

5. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 4 are: Step 41: for the urban POI data and high spatial resolution population distribution data, a buffer zone with a radius of 500m is set for each CO2 concentration monitoring point, and the total number of POIs and the total number of people at the monitoring point within the buffer zone are counted as the POI and population data values ​​corresponding to the monitoring point; Step 42: for high spatial resolution land use data, for each CO2 concentration monitoring point, a buffer zone with a radius of 500m is set, and the area proportion of industrial land, commercial service land, residential land, forest land, grassland, and water land within the buffer zone is counted. The proportion of each land use type corresponding to the monitoring point is used as the land use type data corresponding to the monitoring point; Step 43: Match the collected number of road lanes and road width according to the latitude and longitude coordinates of the road monitoring points with the road section information; Step 44: Match the road congestion data and the average vehicle speed data one by one according to the latitude and longitude coordinates and monitoring time of the road monitoring points; Step 45: Based on the high-temporal and spatial resolution surface meteorological element data and atmospheric background CO2 concentration data, a grid with the same spatial resolution as the road monitoring data is defined, and the grid position of the road monitoring point is determined by the longitude and latitude coordinates of the road monitoring point. Combined with the road CO2 concentration data collection time, the monitoring point data and meteorological elements are matched in space and time to complete the construction of the multi-source input feature data set.

6. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 5 are: Step 51, select the maximum and minimum value standardization to standardize the data: subtract the minimum value from the data, and then divide it by the value range of the data, so that the data value is between [0,1]. The formula is: in, represents the minimum value in the original data set, represents the maximum value in the original data set, is the original value, is the value after standardization; Step 52: randomly divide the standardized feature data set and the corresponding true value of CO2 concentration into a training set and a validation set in a ratio of 8:2, wherein the training set is used for parameter learning and optimization of the model, and the validation set is used to evaluate the generalization performance and prediction accuracy of the model.

7. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 6 are: Step 61: Input the training set data, use the grid search method to tune the key hyperparameters of the random forest model, and determine the hyperparameters that need to be optimized: the number of decision trees, the maximum number of features, the maximum tree depth, the minimum number of split samples, and the minimum number of node samples; Step 62, set the value range of the hyperparameter to be optimized: for each hyperparameter to be optimized, set a set of discrete values ​​as the search space, the number of decision trees is set to generate a set of discrete values ​​in the range of [300,800] with a step size of 50; the maximum number of features is set to two options [sqrt, auto]; the maximum tree depth is set to generate a set of discrete values ​​in the range of [10,100] with a step size of 10; the minimum number of split samples is set to three discrete values ​​of [2,5,10]; the minimum number of node samples is set to three discrete values ​​of [1,2,4]; Step 63, generating hyperparameter combinations: generating all possible hyperparameter combinations according to the set hyperparameter value range; Step 64, train and evaluate the model: For each hyperparameter combination, use the training set data and the five-fold cross-validation method to train the random forest model, and use the mean square error as an indicator to evaluate the model performance. The hyperparameter combination with the smallest mean square error is the hyperparameter combination with the best performance. The formula for calculating the mean square error is: in, represents the observed value of CO2 concentration, represents the predicted value of CO2 concentration, and n represents the number of matching data pairs; Step 65, construct a long short-term memory network model, including an input layer, a hidden layer, and an output layer. The hidden layer includes two LSTM layers and a fully connected layer. The input layer data shape of the LSTM model is in a three-dimensional format. The number of samples represents the number of samples in the training data set. The time step is set to a 1-hour sliding time window, representing the CO2 concentration feature sequence of the previous hour to predict the CO2 concentration of the hour. The number of features represents the input feature vector. In terms of parameter settings, the number of training rounds for each hidden layer is set to 20, the batch size is 64, the optimizer is adaptive moment estimation, the initial learning rate is 0.001, and the loss function is the mean absolute error. After the model is deployed, the training set data is input into the model for model training.

8. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 7 are: Step 71: Input the validation set data and select the coefficient of determination and root mean square error as the evaluation indicators of the model. The coefficient of determination is used to measure the model's ability to explain the data variance, with a value range of 0 to 1, and R 2 The closer it is to 1, the better the model fits. The root square error is the square root of the mean of the square of the difference between the predicted value and the true value. The smaller the RMSE, the better the predictive ability of the model. 2 And the RMSE calculation formula is: in, represents the observed value of CO2 concentration, represents the predicted value of CO2 concentration, represents the arithmetic mean of the CO2 concentration observations, and n represents the number of matching data pairs.

9. The method for hourly simulation and prediction of urban road CO2 concentration based on multi-source data fusion and integrated algorithm driving as claimed in claim 1, characterized in that: The specific steps of step 8 are: Step 81: Based on the trained and verified random forest model and long short-term memory network model for simulating urban road CO2 concentration, the model with the largest determination coefficient and the smallest root mean square error is selected as the final model, and the feature data set is input into the model to realize hourly simulation of CO2 concentration on major urban roads.

Images