Carbon emission evaluation method and system based on multi-source heterogeneous data

Abstract

The invention relates to the technical field of carbon emission evaluation, and relates to a carbon emission evaluation method and system based on multi-source heterogeneous data, and the method comprises the steps: obtaining the space matching data of a target region boundary and parking spaces, and generating the number and distribution information of effective parking spaces; the time sequence dynamic remote sensing data, the traffic density fluctuation data and the demographic data are fused, a parking space activeness factor model is constructed, and the parking space use intensity is output; predicting the vehicle type distribution proportion of the target area by using a neural network, and calculating the annual average driving distance of the vehicle; calculating a dynamic vehicle carbon emission factor and a static vehicle carbon emission factor according to the time dynamic nature and the regional difference of the power grid energy; and calculating the total carbon emission of the target area by integrating the vehicle type distribution proportion, the annual average driving distance and the carbon emission factor based on a collaborative correction mechanism of the parking space resource tensity and the parking space use intensity. The objective of the invention is to realize community-level high-precision carbon emission accounting and improve the spatial resolution and adaptability of carbon emission accounting.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission assessment, and in particular to a carbon emission assessment method and system based on multi-source heterogeneous data. Background Art

[0002] Industry and transportation are the two primary sources of carbon emissions. Transportation emissions account for an increasing proportion of emissions in urban areas, significantly impacting air pollution, energy mix optimization, and the path to carbon peak. Current carbon emissions accounting efforts primarily focus on larger administrative regions, with methods often based on statistical data and macro-level energy consumption models. As carbon peak and carbon neutrality efforts progress towards more refined management units, carbon emissions accounting is gradually being extended to micro-spaces such as streets, communities, and industrial parks.

[0003] Among them, as the functional unit of the city and the main place for residents' activities, the energy consumption and traffic behavior of the community have obvious spatial differences and behavioral regularities. Especially in the context of rapid urbanization and the continuous growth of the number of motor vehicles, carbon emission accounting at the community level has become an important basis for improving the accuracy of low-carbon governance and implementing hierarchical emission reduction management. At present, traditional transportation carbon accounting methods mostly rely on traffic surveys or statistical data, lack high temporal and spatial resolution and dynamics, and therefore cannot meet the needs of community-level carbon emission refined management for high-granularity data. With the rapid development of remote sensing satellite data, AI technology, neural network prediction and big data fusion technology, more accurate community traffic behavior data and carbon emission factors have been provided. Therefore, the present invention proposes a carbon emission assessment method and system based on multi-source heterogeneous data to fill the technical gap of micro-scale carbon accounting and realize community-level carbon emission monitoring, management and decision support. Summary of the Invention

[0004] To achieve high-precision carbon emission accounting at the community level and improve the spatial resolution and adaptability of carbon emission accounting, the present invention provides a carbon emission assessment method and system based on multi-source heterogeneous data. The technical solutions adopted are as follows:

[0005] The technical solution of the first aspect of the present invention provides a carbon emission assessment method based on multi-source heterogeneous data, the method comprising:

[0006] Obtain spatial matching data between the target area boundary and parking spaces, and generate information on the number and distribution of valid parking spaces;

[0007] By integrating time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data, a parking space activity factor model is constructed and the parking space usage intensity is output;

[0008] Use neural networks to predict the distribution ratio of vehicle types in the target area and calculate the average annual distance traveled by vehicles;

[0009] Calculate dynamic vehicle carbon emission factors and static vehicle carbon emission factors based on the temporal dynamics and regional differences of grid energy;

[0010] Based on the coordinated correction mechanism of parking resource tension and parking space usage intensity, the total carbon emissions of the target area are calculated by integrating the vehicle type distribution ratio, average annual driving distance and carbon emission factor.

[0011] Furthermore, spatial matching data between the target area boundary and parking spaces is obtained to generate information on the number and distribution of valid parking spaces, including:

[0012] Determine the spatial coordinate relationship between the vertex coordinate set of the target area boundary and the parking space coordinate point set based on the spatial geometry analysis algorithm;

[0013] Extract the number of valid parking spaces within the target area boundary;

[0014] Generate a visual spatial distribution layer based on the valid parking space coordinate set.

[0015] Furthermore, a parking space activity factor model is constructed and the parking space usage intensity is output, including:

[0016] Extract the brightness change rate of night light remote sensing images in adjacent time periods;

[0017] Calculate the mean road vehicle density per unit time in the traffic heat map;

[0018] Calculate population density based on the total population count and geographic area within the target area;

[0019] Dynamic weight allocation and linear fusion are performed on the brightness change rate, vehicle density mean and population density, and a quantitative value representing the actual usage intensity of the parking space is output.

[0020] Furthermore, a neural network is used to predict the distribution ratio of vehicle types in the target area, including:

[0021] Input the community attribute feature vector in the target area and perform standardization preprocessing;

[0022] Through multi-layer neural network, nonlinear feature mapping and high-dimensional space transformation are performed to output the proportion distribution of fuel vehicles, new energy vehicles and hybrid vehicles;

[0023] The number of each type of vehicle is extracted based on the proportional distribution of vehicle types and the total number of vehicles in the region.

[0024] Furthermore, the calculation of the average annual vehicle distance includes:

[0025] Construct performance influencing factors of vehicle displacement and range;

[0026] The road network density and average commuting distance of the target area are integrated to generate the regional travel support factor;

[0027] The behavioral pattern factor is constructed by correlating daily travel frequency, single trip mileage and weekly travel rate;

[0028] The average annual driving distance is calculated based on the product coupling of performance impact factor, regional support factor and behavior pattern factor.

[0029] Furthermore, based on the temporal dynamics and regional differences of grid energy, dynamic vehicle carbon emission factors are calculated, including:

[0030] Obtain the power generation ratio of different energy types in the power grid during the target period;

[0031] Calculate the dynamic grid emission factor based on the weighted carbon emission factor of the energy type;

[0032] The unit distance emission factor of new energy vehicles is generated by combining the conversion relationship between vehicle charging power and cruising range;

[0033] The emission factors of hybrid vehicles are weighted and combined based on the proportion of fuel driving.

[0034] Furthermore, the calculation of the static vehicle carbon emission factor includes:

[0035] Obtain fuel consumption data per unit mileage of fuel vehicles;

[0036] The carbon emissions per unit distance of a fuel vehicle are calculated based on the fuel carbon content and the carbon dioxide molecular weight conversion coefficient.

[0037] Furthermore, based on the coordinated correction mechanism of parking space resource tension and parking space usage intensity, the total carbon emissions of the target area are calculated by integrating the distribution ratio of vehicle types, average annual driving distance and carbon emission factors, including:

[0038] Generate resource stress parameters based on the total number of vehicles and the number of available parking spaces;

[0039] Use parking space usage intensity as a behavioral dynamic correction weight;

[0040] Adjust the base carbon emissions by using the logarithmic scaling relationship of the resource stress parameter;

[0041] Dynamically correct weights are used to output the final total carbon emissions.

[0042] Furthermore, the expression for calculating the total carbon emissions of the target area is:

[0043]

[0044] Where, Indicates the total carbon emissions in the target area; Ci represents the total carbon emissions of various types of vehicles; log α The logarithmic function with α as the base represents the nonlinear scaling relationship of resource tension; N represents the total number of vehicles in the target area; P i Indicates the number of valid parking spaces in the target area; V i represents the quantitative value of parking space usage intensity; α represents the logarithmic base adjustment coefficient, which is used to control the smoothness of the curve of the resource tension influencing factor; i represents the i-th target area.

[0045] The technical solution of the second aspect of the present invention provides a carbon emission assessment system based on multi-source heterogeneous data, which adopts the carbon emission assessment method based on multi-source heterogeneous data described in the technical solution of the first aspect of the present invention. The system includes:

[0046] a data acquisition module configured to acquire spatial matching data between the target area boundary and the parking spaces and generate valid parking spaces and distribution information;

[0047] The parking space usage intensity calculation module is configured to integrate time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data to construct a parking space activity factor model and output parking space usage intensity;

[0048] a vehicle annual average travel distance calculation module configured to use a neural network to predict the distribution ratio of vehicle types in a target area and calculate the vehicle annual average travel distance;

[0049] A vehicle carbon emission factor calculation module is configured to calculate a dynamic vehicle carbon emission factor and a static vehicle carbon emission factor based on the temporal dynamics and regional differences of grid energy;

[0050] The total carbon emissions calculation module is configured as a coordinated correction mechanism based on parking resource tension and parking space usage intensity, integrating the vehicle type distribution ratio, average annual driving distance and carbon emission factor to calculate the total carbon emissions of the target area.

[0051] The present invention has the following beneficial effects:

[0052] The carbon emission assessment method based on multi-source heterogeneous data provided by the present invention first quantifies the number and distribution information of effective parking spaces in the target area, and uses multi-source heterogeneous data to construct a parking space activity model to quantify the actual usage intensity of parking spaces; then uses a neural network to mine the nonlinear mapping relationship between the characteristics of the target area and the vehicle type, combines vehicle attributes, regional characteristics and travel behavior to calculate the average annual driving distance of vehicles, and designs a dynamic and static coupled carbon emission factor algorithm for vehicles of different energy types; finally, a refined accounting of total carbon emissions is achieved through a collaborative correction mechanism of parking space tension and activity. This method solves the limitations of traditional methods in spatial resolution, dynamic adaptability and multi-factor coupling analysis, constructs a micro-scale carbon emission assessment framework, improves the accuracy, dynamics and scenario adaptability of community-level carbon emission accounting, provides an effective decision-making basis for low-carbon governance, and meets the needs of refined monitoring and governance of community-level carbon emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 A flowchart of a carbon emission assessment method based on multi-source heterogeneous data provided by one embodiment of the present invention;

[0055] Figure 2 A flow chart of a method for spatial geometry analysis provided by one embodiment of the present invention;

[0056] Figure 3 A schematic diagram of a parking space activity factor model provided by one embodiment of the present invention;

[0057] Figure 4 A schematic diagram of a predicted vehicle distribution ratio provided by one embodiment of the present invention;

[0058] Figure 5 A schematic diagram of calculating total carbon emissions provided by one embodiment of the present invention;

[0059] Figure 6 A structural diagram of a carbon emission assessment system based on multi-source heterogeneous data provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0060] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a carbon emissions assessment method and system based on multi-source heterogeneous data proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0061] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0062] The following describes in detail a carbon emission assessment method and system based on multi-source heterogeneous data provided by the present invention with reference to the accompanying drawings.

[0063] See also Figure 1 , which shows a method flow chart of a carbon emission assessment method based on multi-source heterogeneous data provided by an embodiment of the present invention, the method comprising:

[0064] Step S100: Obtain spatial matching data between the target area boundary and parking spaces, and generate information on the number and distribution of valid parking spaces;

[0065] See also Figure 2 , step S100 specifically includes:

[0066] Step S110: Determine the spatial coordinate relationship between the vertex coordinate set of the target area boundary and the parking space coordinate point set based on the spatial geometry analysis algorithm; specifically, basic data such as the community name, specific address, geographical location latitude and longitude, and number of parking spaces can be collected from the webpage content of the real estate data platform, and then obtain community boundaries, administrative divisions and other data on the public GIS data platform or high-precision map platform, and import the urban community AOI boundary layer into the GIS engine. The engine will then perform geospatial matching on the parking spaces, using the spatial geometry analysis algorithm to ensure that each parking space data can correspond to the correct community area, and define the coordinates of the parking space as a point set represented as: P = {p1, p2, ..., p n}, each point p k Expressed as geographic coordinates (x k ,y k ), represents the latitude and longitude coordinates of the kth parking space, and defines the community AOI boundary as the area described by the polygon, expressed as: Where m is the number of polygon vertices, i represents the i-th community;

[0067] Step S120: Extract the number of valid parking spaces within the target area boundary. This embodiment uses ray geometry to determine whether the parking space is within the polygon. The angle and the judgment can be expressed as:

[0068]

[0069] In formula (1), Represents point p k With polygon A i The sum of the angles between the vertex pairs of Then δ(p k ,A i )=1, indicating point p k is within the polygonal area; otherwise, it is outside the area, δ(p k ,A i )=0;

[0070] Then we can determine the number of parking spaces in the community:

[0071]

[0072] In formula (2), P i represents the number of effective parking spaces in the i-th community; δ(p k ,A i ) is the determination result of step S110, 1 indicates that it is within the community, and 0 indicates that it is outside the community; traversing all parking spots, a valid parking space set can be formed;

[0073] Step S130: Generate a visualization spatial distribution layer based on the valid parking space coordinate set; specifically, the points and polygons are visualized in the ArcGIS platform, showing the parking space distribution in each community in the city, and generating a heat map or point map. Step S100 uses multi-source data collection and spatial geometry analysis algorithms to spatially match parking space coordinate points with community boundary polygons, accurately determining whether each parking space is located within the target community, counting the number of valid parking spaces, and generating a visualization distribution layer. This achieves precise geographic positioning and quantification of community parking spaces, thereby improving the spatial resolution and data visualization capabilities of community-level carbon emissions accounting.

[0074] Step S200: Integrate time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data to construct a parking space activity factor model and output parking space usage intensity. Since the effective number of parking spaces cannot truly reflect the number of vehicles traveling, this embodiment introduces a parking space activity factor model to quantify the actual usage frequency of parking spaces in a target area, such as a community. The modeling is mainly based on three dimensions: night light remote sensing map data, traffic heat map data, and population density. Among them, the night light remote sensing map reflects the day and night changes of the parking area, the traffic heat map can identify the peak and trough periods of vehicle travel, and the population density can reflect the travel demand in the area, that is, areas with high population density also have high vehicle activity.

[0075] See also Figure 3 , step S200 specifically includes:

[0076] Step S210: Extract the brightness change rate of night light remote sensing images in adjacent time periods; specifically, obtain night light image data of the target area in adjacent time periods, such as 8 pm and 9 pm on weekdays, from remote sensing satellites (such as Suomi NPP / VIIRS), and extract the average brightness value of the corresponding target area; night light remote sensing images contain the brightness change intensity of night light. By comparing the degree of light change in different time periods, an indicator of human activity intensity can be estimated, and this indicator is defined as the night light change rate NDL i , which can be expressed as:

[0077]

[0078] In formula (3), NL c Indicates the night light index of the current period, that is, the average brightness value of the community in the current period; NL p Indicates the night light index of the previous period; if the night light change rate is high, it means that the traffic flow in the area during that period is large.

[0079] Step S220: Calculate the mean vehicle density per unit time in the traffic heat map. Specifically, the traffic heat map can be realized by collecting real-time road conditions from high-precision maps such as Google and Baidu on a large scale. Traffic density is usually used to display the number of vehicles in the area, that is, the real-time road conditions of the road. For each observation time t, the traffic density TD t The calculation method is:

[0080]

[0081] In formula (4), TF t It refers to the number of vehicles passing through the target area's roads at time t, which can be obtained through monitoring data from traffic management departments or real-time traffic platforms such as online maps. RC refers to the maximum number of vehicles that a road in the target area can accommodate per unit time, that is, the maximum flow rate designed for the road.

[0082] Since high traffic density and long-term congestion will lead to a high traffic heat value in the area, in order to ensure the accuracy of the activity model, this implementation also needs to calculate the traffic heat value TI for a period of time i , calculated as:

[0083]

[0084] In formula (5), T is the time period of traffic condition observation.

[0085] Step S230: Calculate the population density based on the total population count and geographic area within the target area. Population density is a fundamental statistic that influences regional traffic and is an important factor in determining parking space activity. The population density of a region can be estimated using demographic data or satellite remote sensing data from the target area. Calculating population density is easier for smaller areas, so introducing population density into the activity model can broaden its application. The total population TP of a specific area, such as a community or street, can usually be obtained from a city or region's statistical yearbook or census data. The ratio of this to the polygonal area of ​​the target area AOI in step 1 is the population density of the region, which can be expressed as:

[0086]

[0087] In formula (6), PD i Indicates the population density of the target area, quantifying the degree of population concentration within a unit area. High-value areas usually correspond to higher travel demand;

[0088] Step S240: Dynamically weight the brightness change rate, the mean vehicle density, and the population density and linearly fuse them to output a quantitative value representing the actual parking space usage intensity. The activity factor model can be expressed as:

[0089] V i =α·NDL i +β·TI i +γ·PD i (7)

[0090] In formula (7), α, β, and γ represent the empirical weight coefficients of the brightness change rate, the mean vehicle density, and the population density, respectively, which are used to adjust the contribution of different factors to the activity factor. In this model, α + β + γ = 1. It should be noted that for target areas with insufficient historical data and a relatively simple model structure, the specific values ​​can be obtained through experience. For example, in the central area of ​​the city, the traffic heat value β has a greater impact and can be given a higher weight. The night light intensity in the suburbs has a greater impact on the activity, and the population density can be given an appropriate weight based on the differences in different places. In areas with more diversified development, the calculation of the activity factor directly serves the carbon emission management decision. The decision result is highly sensitive to the calculation accuracy. Statistical regression analysis can be used to calculate the weight coefficient to reduce the deviation and combine the historical activity data of the area. Perform linear regression fitting with the established activity model and obtain the optimal weight coefficient through minimum error solution. The minimum error function is as follows:

[0091]

[0092] In formula (8), L(α, β, γ) represents the error function, which is the mean squared deviation between the model prediction value and the true value, and is used to measure the fitting accuracy of the activity factor model; α, β, and γ are the weight coefficients to be optimized, corresponding to the brightness change rate, the mean vehicle density, and the population density, respectively; N is the number of samples used for model training, such as the number of target areas in historical observations; represents the actual activity data of the i-th target area; V i represents the model-predicted activity value of the i-th target area;

[0093] When minimizing the error, The partial derivative equations of the three weight coefficients can be obtained:

[0094]

[0095] Formulas (9), (10), and (11) obtain the necessary conditions for the optimal weight by taking the partial derivatives of the error function L with respect to α, β, and γ and setting them equal to zero; the coefficients Is a constant factor after differentiation, which does not affect the solution of the equation and is only used to standardize the gradient direction; convert equations (9), (10), and (11) into matrix expressions:

[0096] A·w=b(12)

[0097] In matrix (12), w=[α,β,γ] T is the weight coefficient vector to be solved, is a column matrix; b is the dot product vector of the input feature and the true value; A is the quadratic matrix of the input feature matrix representing the correlation between features, and the specific elements are:

[0098]

[0099]

[0100] In formula (13) and formula (14), the diagonal elements, such as is the sum of the squares of each feature, reflecting the variation of the feature itself; non-diagonal elements, such as is the cross product sum between features, reflecting the correlation between features; each element in b represents the real activity The product sum of the corresponding features is used to characterize the degree of linear correlation between the true value and the feature;

[0101] Then, the weight coefficient is solved by the matrix, which is expressed as:

[0102] w=A -1 ·b(15)

[0103] In formula (15), w represents the optimal weight vector; A -1 Indicates the inversion of matrix A;

[0104] This method, driven by real-world data, assigns weights and is applicable to scenarios involving nonlinear relationships between complex factors, dynamic parameter adjustment, and improved model accuracy. In high-precision applications such as transportation planning and carbon emissions forecasting, it can flexibly adjust vehicle activity levels based on the actual conditions of different cities or regions. This method overcomes the subjectivity of empirically assigned values ​​by dynamically calibrating weights through statistical regression. This method is particularly applicable to diverse regions with rich data and complex feature relationships, significantly improving the computational accuracy and decision-making reliability of the activity factor model.

[0105] In summary, this embodiment, by constructing an activity factor model, can more accurately reflect the actual frequency of parking space use. Compared with counting the number of parking spaces, this model can further estimate carbon emissions at the transportation level and can also help government departments rationally assess and plan regional transportation demand and carbon management measures. This method breaks through the limitations of traditional methods that rely solely on the number of parking spaces, accurately capturing the temporal and spatial differences in the actual frequency of parking space use, providing an intensity indicator for carbon emission calculations that is closer to real travel behavior, significantly improving the dynamic and scenario-adaptive nature of community-level carbon emission assessments, and providing a quantitative basis based on real-time data for low-carbon measures such as transportation planning and charging station layout.

[0106] Step S300: Use a neural network to predict the distribution ratio of vehicle types in the target area and calculate the average annual driving distance of vehicles; since different types of vehicles have significantly different energy consumption levels and carbon emission characteristics, in order to refine the calculation of traffic carbon emissions, this embodiment establishes a vehicle carbon emission model by predicting the ratio of different types of vehicles in the target area.

[0107] See also Figure 4 , where the neural network is used to predict the distribution ratio of vehicle types in the target area, including:

[0108] Step S310: Input the community attribute feature vector in the target area and perform standardization preprocessing. In this embodiment, a neural network regression model based on AI analysis is constructed. The model is divided into an input layer, a hidden layer, and an output layer. The input layer represents the main characteristic data of the community, including: community housing price x1 (10,000 yuan / m2), per capita income x2 (10,000 yuan / year), greening rate x3 (%), population density x4 (number of people / km2), and community housing age x5 (years). Among them, housing price, per capita income, and population density reflect the economic development level and purchasing power of residents in the region, and are positively correlated with vehicle type. The greening rate reflects the investment in environmental protection work in the region and affects the carbon emissions of the community to a certain extent. The age of community housing usually affects the consumption awareness of residents. For example, residents in old communities tend to use traditional fuel vehicles, while residents in newer communities have a higher proportion of new energy vehicles. Construct the input feature vector X i =[x1,x2,x3,x4,x5], i.e., the attribute data of the i-th target area. In this embodiment, the usage ratios of fuel vehicles y1, new energy vehicles y2, and hybrid vehicles y3 within the target area are calculated.

[0109] First, the input features of the input layer are standardized. There is a nonlinear relationship between the input features and the vehicle scale, and the magnitude of different input features varies greatly. Directly inputting them into the neural network will cause the features with larger magnitudes to dominate the model training, weakening the influence of other features and reducing the model's output. To eliminate the influence of units and magnitude on the model, the numerical values ​​of different features are scaled and mapped to the range [0,1]. The mapping method is:

[0110]

[0111] Where x i represents the original eigenvalue; min(x i )、max(x i ) represent the minimum and maximum values ​​of the original eigenvalues ​​in the training set respectively; is the standardized eigenvalue;

[0112] Step S311: Perform nonlinear feature mapping and high-dimensional space transformation through a multi-layer neural network to output the proportion distribution of fuel vehicles, new energy vehicles and hybrid vehicles; specifically, set the input layer to X i =[x1,x2,x3,x4,x5], the target variable output layer is Y i=[y1,y2,y3], where y1+y2+y3=1; AI is used to construct a neural network to perform multi-layer structure superposition and map the input layer to the output layer. In the multi-layer structure, the hidden layer is the core part and consists of several nodes. The deep function mapping relationship between the input features and the output layer is learned through layer-by-layer nonlinear transformation. First, a weight matrix is ​​constructed. The value obtained by linearly weighting the output of all nodes in the previous layer through the weight matrix is ​​used as the input value of the next node. The bias term is added to correct the result in the process so that the model can adaptively adjust the mapping relationship. Finally, an activation function is constructed to perform a nonlinear transformation on the output of the hidden layer. The connection weight matrix establishes a linear relationship between the input layer and the hidden layer, where k represents the number of input layer vectors, k∈(1,5),W k is the weight matrix of the input layer:

[0113]

[0114] In formula (17), W k is the k-th layer weight matrix; b k is the bias vector of the kth layer; if the neural network contains n hidden layers, the nth hidden layer is defined as h n , the weight matrix of this layer is W n , the bias vector is b n ; It is the output after linear transformation, and nonlinear mapping is achieved through the activation function ReLU;

[0115] h n =ReLU(W n ·h n-1 +b n )(18)

[0116] In formula (18), h n-1 is the input result of the previous hidden layer, and ReLU is the activation function, which is used to alleviate the vanishing gradient and introduce nonlinearity;

[0117] The output layer maps the features extracted by the hidden layer to the probability distribution of various vehicle types in the community:

[0118]

[0119] In formula (19), the softmax function is a probability mapping function, let the unnormalized output z i =[z1,z2,z3]=W n ·h n-1 +b n Represents the unnormalized output of the vehicle category. The softmax function is used to normalize the model output results, so that the final output layer satisfies the probability distribution:

[0120]

[0121] Finally, the calculation of the vehicle type ratio in the output layer is based on the multi-layer nesting of the input layer features, and the ratio of the three vehicle types in the region can be obtained by solving the following formula.

[0122] Y i =softmax(W n ReLU(…ReLU(W 1 ·X i +b 1 )+…+b n ))(twenty one)

[0123] The neural network continuously adjusts the weight matrix W and bias vector b through training data, optimizes through the gradient descent algorithm, and establishes a minimization loss function. Let θ = [W, b] be the data to be optimized, μ be the learning rate, and establish the gradient descent function:

[0124]

[0125] By iteratively updating the model through machine learning, its value continues to approach 0, making the output results of each layer more accurate.

[0126] Step S312: Extract the number of each type of vehicle based on the vehicle type distribution ratio and the total number of vehicles in the area; the number of three types of vehicles in the community is defined as:

[0127] K i =Y i ·N(23)

[0128] Where N is the total number of vehicles in the target area, K i is the number of different vehicle types, the number of fuel vehicles is K1 = Y1·N, the number of new energy vehicles is K2 = Y2·N, and the number of hybrid vehicles is K3 = Y3·N;

[0129] Step S310 constructs a neural network model consisting of an input layer, a hidden layer, and an output layer, and inputs the attribute feature vectors of community housing prices, per capita income, etc. into the model after standardized preprocessing. After multi-layer nonlinear feature mapping and high-dimensional space transformation, the proportional distribution of fuel vehicles, new energy vehicles, and hybrid vehicles is output, and the number of each type of vehicle is extracted in combination with the total number of vehicles in the area, realizing nonlinear prediction from community attributes to vehicle type distribution, solving the problem that traditional methods are difficult to capture complex correlation relationships, and providing refined vehicle structure data support for community-level carbon emission accounting.

[0130] The calculated annual average vehicle travel distance D iThe average annual mileage refers to the average mileage of a vehicle in a year. The main influencing factors include traffic flow, community travel mode, city size, road conditions, etc. Urban or regional traffic management departments regularly conduct annual traffic surveys in various communities and issue traffic statistics yearbooks. The average annual mileage of vehicles can be determined from these reports. This embodiment provides a method for calculating the average annual mileage of vehicles based on community behavior attributes. It establishes three-dimensional functions, namely, vehicle attribute function f1, regional characteristic function f2, and resident travel behavior function f3. Vehicle attributes are functions based on vehicle displacement and energy efficiency. Regional characteristics are functions of regional road density and commuting patterns. Travel behavior is the travel statistics of community residents.

[0131] Specifically include:

[0132] Step S320: Construct performance influencing factors for vehicle displacement and cruising range. The performance influencing factors are vehicle attribute functions used to analyze the impact of a vehicle's physical performance and technical parameters on its travel use. Displacement size can reflect the feasibility of a vehicle's travel scenarios, and cruising range is compatible with the performance differences between fuel vehicles, hybrid vehicles, and new energy vehicles, providing a unified range expression:

[0133]

[0134] In formula (24), d is the engine displacement of a single vehicle; d avg is the average displacement of vehicles in the target area; M is the actual cruising range of a single vehicle; M max The maximum theoretical cruising range of the vehicle; Reflects the relative size of displacement; Characterizes battery life utilization;

[0135] Step S321: The road network density and average commuting distance of the target area are integrated to generate a regional travel support factor, i.e., a regional characteristic function. The road network density is an indicator of transportation convenience, and the average commuting distance corresponds to the daily car demand of community residents. This is used to construct a driving factor for evaluating the support for car use behavior, achieving quantifiable modeling of travel intensity in multiple scenarios. This factor is transferable in the regional dimension and can be expressed as:

[0136]

[0137] In formula (25), L1 is the total length of roads in the target area; A i is the polygonal area of ​​the target region; c avg is the average commuting distance of residents; τ is the road network density, measured in km / km², which measures the degree of road network development per unit area. A larger value indicates more convenient transportation and supports high-frequency travel. The data in this formula can usually be obtained from the statistical reports of the regional urban planning department.

[0138] Step S322: Associate daily travel frequency, single trip mileage, and weekly travel rate to construct a behavioral pattern factor, i.e., the resident travel behavior function. Due to the intermittent travel patterns of residents in the region, the use of daily average travel frequency cannot accurately express the temporal stability of travel intensity and the coverage of travel days. In this implementation, the "weekly travel rate" temporal adjustment parameter is used to effectively capture the distribution characteristics of travel behavior activity within a week, overcoming the behavioral bias caused by the existing model based only on daily average data. By modifying this factor, the model can accurately distinguish between high-frequency low-activity and medium-frequency high-activity diversified vehicle usage patterns, improving the behavioral stability and individual difference adaptability of vehicle annual mileage estimation, which can be expressed as:

[0139] f3=t′·l·r(26)

[0140] In formula (26), t′ is the average number of trips per day, for example, residents of a community may travel 1.5 times per day; l is the average mileage of a single trip; r is the weekly travel rate, which is the ratio of the actual number of trips per week to the number of trips per 7 days. It is used to correct the difference in activity between weeks and avoid ignoring patterns such as high-frequency travel on weekdays and low activity on weekends by using only daily average data.

[0141] Step S323: Calculate the average annual driving distance based on the product coupling of the performance impact factor, the regional support factor, and the behavior pattern factor; the average annual driving distance D i It can be expressed as:

[0142] D i =f1·f2·f3(27)

[0143] Step S320 constructs a vehicle attribute function, a regional characteristic function, and a resident travel behavior function to quantify the performance impact of vehicle displacement and cruising range, the regional support of road network density and average commuting distance, and the behavioral patterns of daily travel frequency and weekly travel rate. The average annual driving distance is then calculated in a coupled product form, thereby achieving a systematic evaluation of vehicle driving intensity from the three dimensions of vehicle performance, regional demand, and travel willingness. This addresses the limitations of traditional methods that rely on static mileage parameters, thereby improving the accuracy and scenario adaptability of mileage estimation in community-level carbon emission accounting.

[0144] Step S400: Calculate dynamic and static vehicle carbon emission factors based on the temporal dynamics and regional differences in grid energy. Calculate carbon emission factors for fuel vehicles, hybrid vehicles, and new energy vehicles separately. Temporal dynamics refers to the grid energy structure, such as how fluctuations in thermal power, wind power, and photovoltaic power generation over time affect the carbon emissions of new energy vehicle charging. Regional differences refer to the spatial differences in emission factors caused by the grid energy structure in different regions.

[0145] Among them, the dynamic vehicle carbon emission factor is calculated based on the temporal dynamics and regional differences of grid energy, including:

[0146] Step S410: Obtain the power generation ratio of different energy types in the power grid during the target period. Specifically, the power generation structure data of the target community area can be obtained from the power department database or energy statistical report, including the power generation ratio σ of each energy type, such as coal, gas, wind power, photovoltaic power, etc. j′ (t), where j′ represents the energy type index, ∑ j′ σ j′ (t) = 1;

[0147] Step S411: Calculate the dynamic grid emission factor based on the weighted carbon emission factor of the energy type. The grid carbon emission factor in the traditional method generally uses a regional unified average value, ignoring the differences in power sources in different regions and the changes in the grid power generation structure in different power consumption periods. In order to accurately reflect the carbon emissions of electric vehicles in different charging periods and different regions, a dynamic grid carbon emission factor model is proposed in this embodiment, and its expression is:

[0148]

[0149] In formula (28), EF grid (t) is the dynamic grid emission factor at time t; EF j′ is the carbon emission factor per unit of power generation of the energy source in the j′th node, in kgCO2 / kWh. It reflects the carbon intensity difference of the grid energy structure in different periods through weighted average. For example, if the proportion of wind power is high during the night off-peak period, then EF grid (t) lower;

[0150] Step S412: Generate the unit distance emission factor of new energy vehicles by combining the conversion relationship between vehicle charging power and cruising range; the carbon emissions of new energy vehicles are mainly generated during the charging process. Different types of energy generation are transmitted from the power grid to the community charging piles, so the carbon emission factor of new energy vehicles depends on the carbon emission factor of the local power grid. In addition to the power generation structure of the power grid, the carbon emissions of new energy vehicles during the charging process are also related to the specific charging time, charging power and other behaviors. The spatiotemporal carbon emission factor EF of new energy vehicles ev (t) is calculated as follows:

[0151] EF ev (t) = EC charge (t)·EF grid (t)(29)

[0152]

[0153] In formulas (29) and (30), EC charge(t) is the charging behavior factor, which represents the charging energy consumption per unit mileage; P(t) is the charging power at time t; Δt is the charging time; M ev is the maximum range of a new energy vehicle after being fully charged. This formula calculates the spatiotemporal carbon emission factor of new energy vehicles, quantifying the impact of charging time and regional power grid characteristics on emissions.

[0154] Step S413: weighted combination of hybrid vehicle emission factors based on fuel driving ratio; specifically, the hybrid vehicle carbon emission factor EF in this embodiment is hybrid The calculation needs to take into account both fuel-driven and electric-driven vehicles. The carbon emission factor is composed of the fuel emission factor and the charging emission factor in proportion. The calculation method is the same as that of new energy vehicles and can be expressed as:

[0155]

[0156] In formula (31), EF is the fuel-driven ratio of the hybrid vehicle, which is determined by the vehicle's own design parameters and represents the fuel consumption ratio of the hybrid vehicle during normal driving; fuel is the carbon emission factor per unit mileage of fuel vehicles;

[0157] The calculation of the static vehicle carbon emission factor includes:

[0158] Step S420: Obtain the fuel consumption per mileage data of the fuel vehicle; obtain the fuel consumption per mileage FC from the vehicle registration data or fuel consumption test report v , characterizes vehicle fuel efficiency;

[0159] Step S421: Calculate the carbon emissions per unit distance of a fuel vehicle based on the carbon content of the fuel and the carbon dioxide molecular weight conversion coefficient; the carbon emission factor EF of the fuel vehicle fuel It is mainly determined by the vehicle's fuel consumption and the carbon content of the fuel, and is calculated as follows:

[0160]

[0161] In formula (32), FC v CC is the fuel consumption per unit mileage of the vehicle; fuel The carbon content of the fuel used by the vehicle is calculated by converting the carbon element in the fuel into carbon dioxide and expressed as the carbon conversion coefficient, which is the ratio of the relative molecular mass of CO2 to C; is the molecular weight ratio of carbon element converted into carbon dioxide;

[0162] Step S400 constructs a dynamic and static carbon emission factor calculation system by distinguishing the different emission characteristics of fuel vehicles, new energy vehicles and hybrid vehicles. A dynamic factor model based on the grid energy structure and charging behavior is introduced for new energy vehicles to accurately reflect the impact of charging period and regional grid cleanliness. For hybrid vehicles, the fuel and electricity emission factors are weighted and combined according to the fuel drive ratio, which solves the problem that the fixed emission factors in traditional methods cannot capture temporal and spatial differences. At the same time, for fuel vehicles, calculations based on fuel consumption and carbon content molecular weight conversion are used to ensure accounting consistency, ultimately achieving accurate quantification of multi-dimensional carbon emissions from energy production to vehicle use.

[0163] Step S500: Based on a coordinated correction mechanism for parking resource scarcity and parking usage intensity, the total carbon emissions of the target area are calculated by integrating the distribution ratio of vehicle types, average annual travel distance, and carbon emission factors. Specifically, steps S100 to S400 respectively propose an effective parking space quantity method based on multi-source data and spatial geometry analysis, a parking space activity model based on multi-source space, a vehicle type prediction model based on community characteristics, a method for calculating average annual vehicle mileage based on community behavioral attributes, and a process for calculating carbon emission factors for different vehicle types. On this basis, this embodiment further proposes an expanded carbon emissions calculation model based on multi-factor fusion to achieve a comprehensive quantitative assessment of the total carbon emissions generated by residents' travel activities at the community level.

[0164] See also Figure 5 , step S500 specifically includes:

[0165] Step S510: Generate resource tension parameters based on the total number of vehicles and the number of available parking spaces; obtain the target area from step S100, such as the number of available parking spaces P in the community. i ; Get the total number of vehicles N from step S300, that is, the total number of fuel vehicles, new energy vehicles, and hybrid vehicles predicted based on community characteristics; the parking space shortage can be calculated by the ratio To quantify, When the demand for parking spaces exceeds the supply, vehicles need to go around to find parking spaces, resulting in additional carbon emissions;

[0166] Step S520: Use the parking space usage intensity as the behavior dynamic correction weight; extract the parking space usage intensity V from step S200 i As a behavior correction factor to reflect dynamic characteristics such as parking space turnover rate and peak hour occupancy rate;

[0167] Step S530: Adjust the basic carbon emissions by using the logarithmic scaling relationship of the resource intensity parameter, which can be specifically expressed as:

[0168]

[0169] In formulas (33), (34), (35), and (36), C i is the sum of basic carbon emissions of all vehicles in the i-th target area; is the carbon emissions of fuel vehicles in the target area; is the carbon emissions of new energy vehicles; is the carbon emissions of hybrid vehicles; K1 is the number of fuel vehicles in the target area, which can be calculated by the proportion of vehicle types and the total number of vehicles; is the average annual driving distance of the jth fuel vehicle in the target area; is the carbon emission factor per unit mileage of the j-th fuel vehicle; K2 is the number of new energy vehicles in the target area; K3 is the number of hybrid vehicles in the target area; represents the average annual driving distance of the jth new energy vehicle in the target area; represents the charging behavior factor of the j-th new energy vehicle in time period t; is the grid carbon emission factor of the j-th new energy vehicle in period t; represents the average annual driving distance of the jth hybrid vehicle in the target area; represents the unit mileage carbon emission factor of the jth hybrid vehicle;

[0170] Then through A nonlinear correction is introduced, where α is the logarithmic base adjustment coefficient, which is used to control the smoothness of the curve of the resource tension influencing factor;

[0171] Step S540: Use the dynamic correction weight to output the final total carbon emissions. The expression for calculating the total carbon emissions of the target area is:

[0172]

[0173] In formula (37), Indicates the total carbon emissions in the target area; based on resource intensity and parking space usage intensity V i , for the basic emission C i Make corrections; log α The logarithmic function with α as the base represents the nonlinear scaling relationship of resource tension; N represents the total number of vehicles in the target area; P i Indicates the number of valid parking spaces in the target area; V i represents the quantitative value of parking space usage intensity; α represents the logarithmic base adjustment coefficient, which is used to control the smoothness of the curve of the resource tension influencing factor;

[0174] This embodiment integrates accounting data such as vehicle type distribution, average annual driving distance, and carbon emission factors, introduces a collaborative correction mechanism for parking space shortage and parking space usage intensity, and constructs an extended model with multi-factor fusion to achieve a dynamic and accurate assessment of the total carbon emissions of community residents' travel. This avoids the defects of traditional methods that ignore spatial resource constraints and behavioral dynamics, and also improves the adaptability of the accounting model to the community microenvironment through multi-dimensional data coupling.

[0175] In summary, the carbon emission assessment method based on multi-source heterogeneous data provided by the present invention first quantifies the number and distribution information of effective parking spaces in the target area, and uses multi-source heterogeneous data to construct a parking space activity model to quantify the actual usage intensity of parking spaces; then uses a neural network to mine the nonlinear mapping relationship between the characteristics of the target area and the vehicle type, combines vehicle attributes, regional characteristics and travel behavior to calculate the average annual driving distance of vehicles, and designs dynamic and static coupled carbon emission factor algorithms for vehicles of different energy types; finally, a refined calculation of total carbon emissions is achieved through a collaborative correction mechanism of parking space tension and activity. This method solves the limitations of traditional methods in spatial resolution, dynamic adaptability and multi-factor coupling analysis, constructs a micro-scale carbon emission assessment framework, improves the accuracy, dynamism and scenario adaptability of community-level carbon emission accounting, provides an effective decision-making basis for low-carbon governance, and meets the needs of refined monitoring and governance of community-level carbon emissions.

[0176] See also Figure 6 , which shows a schematic structural diagram of a carbon emission assessment system based on multi-source heterogeneous data provided by one embodiment of the present invention. The technical solution of the second aspect of the present invention provides a carbon emission assessment system based on multi-source heterogeneous data, which adopts the carbon emission assessment method based on multi-source heterogeneous data described in the technical solution of the first aspect of the present invention. The system includes:

[0177] a data acquisition module configured to obtain spatial matching data between the target area boundary and the parking spaces and generate valid parking spaces and distribution information;

[0178] The parking space usage intensity calculation module is configured to integrate time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data to build a parking space activity factor model and output parking space usage intensity;

[0179] a vehicle annual average travel distance calculation module configured to use a neural network to predict the distribution ratio of vehicle types in a target area and calculate the vehicle annual average travel distance;

[0180] A vehicle carbon emission factor calculation module is configured to calculate a dynamic vehicle carbon emission factor and a static vehicle carbon emission factor based on the temporal dynamics and regional differences of grid energy;

[0181] The total carbon emissions calculation module is configured as a coordinated correction mechanism based on parking resource tension and parking space usage intensity, integrating the vehicle type distribution ratio, average annual driving distance and carbon emission factor to calculate the total carbon emissions of the target area.

[0182] It should be noted that the present invention involves the fusion and processing of multi-source data in actual application, with large data volumes and high computational complexity. The various functional modules can be modularly developed and flexibly combined according to actual needs to achieve system scalability and deployment flexibility. The specific data types, algorithmic models, spatial analysis methods, and computational tools involved in the aforementioned embodiments can be used independently or in combination in different application scenarios, and can also be replaced or optimized based on technological evolution, without affecting the core ideas and essence of the present invention.

[0183] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0184] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A carbon emission assessment method based on multi-source heterogeneous data, characterized by: The method comprises: Obtain spatial matching data between the target area boundary and parking spaces, and generate information on the number and distribution of valid parking spaces; By integrating time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data, a parking space activity factor model is constructed and the parking space usage intensity is output; Use neural networks to predict the distribution ratio of vehicle types in the target area and calculate the average annual distance traveled by vehicles; Calculate dynamic vehicle carbon emission factors and static vehicle carbon emission factors based on the temporal dynamics and regional differences of grid energy; Based on the coordinated correction mechanism of parking resource tension and parking space usage intensity, the total carbon emissions of the target area are calculated by integrating the vehicle type distribution ratio, average annual driving distance and carbon emission factor.

2. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: Obtain spatial matching data between the target area boundary and parking spaces, and generate information on the number and distribution of valid parking spaces, including: Determine the spatial coordinate relationship between the vertex coordinate set of the target area boundary and the parking space coordinate point set based on the spatial geometry analysis algorithm; Extract the number of valid parking spaces within the target area boundary; Generate a visual spatial distribution layer based on the valid parking space coordinate set.

3. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: Construct a parking space activity factor model and output parking space usage intensity, including: Extract the brightness change rate of night light remote sensing images in adjacent time periods; Calculate the mean road vehicle density per unit time in the traffic heat map; Calculate population density based on the total population count and geographic area within the target area; Dynamic weight allocation and linear fusion are performed on the brightness change rate, vehicle density mean and population density, and a quantitative value representing the actual usage intensity of the parking space is output.

4. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: Use neural networks to predict the distribution ratio of vehicle types in the target area, including: Input the community attribute feature vector in the target area and perform standardization preprocessing; Through multi-layer neural network, nonlinear feature mapping and high-dimensional space transformation are performed to output the proportion distribution of fuel vehicles, new energy vehicles and hybrid vehicles; The number of each type of vehicle is extracted based on the proportional distribution of vehicle types and the total number of vehicles in the region.

5. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: The calculation of the average annual vehicle distance includes: Construct performance influencing factors of vehicle displacement and range; The road network density and average commuting distance of the target area are integrated to generate the regional travel support factor; The behavioral pattern factor is constructed by correlating daily travel frequency, single trip mileage and weekly travel rate; The average annual driving distance is calculated based on the product coupling of performance impact factor, regional support factor and behavior pattern factor.

6. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: Based on the temporal dynamics and regional differences of grid energy, dynamic vehicle carbon emission factors are calculated, including: Obtain the power generation ratio of different energy types in the power grid during the target period; Calculate the dynamic grid emission factor based on the weighted carbon emission factor of the energy type; The unit distance emission factor of new energy vehicles is generated by combining the conversion relationship between vehicle charging power and cruising range; The emission factors of hybrid vehicles are weighted and combined based on the proportion of fuel driving.

7. The carbon emission assessment method based on multi-source heterogeneous data according to claim 1, characterized in that: The calculation of the static vehicle carbon emission factor includes: Obtain fuel consumption data per unit mileage of fuel vehicles; The carbon emissions per unit distance of a fuel vehicle are calculated based on the fuel carbon content and the carbon dioxide molecular weight conversion coefficient.

8. The carbon emission assessment method based on multi-source heterogeneous data according to any one of claims 1 to 7, characterized in that: Based on a coordinated correction mechanism for parking resource scarcity and parking usage intensity, the total carbon emissions of the target area are calculated by integrating the distribution ratio of vehicle types, average annual driving distance, and carbon emission factors, including: Generate resource stress parameters based on the total number of vehicles and the number of available parking spaces; Use parking space usage intensity as a behavioral dynamic correction weight; Adjust the base carbon emissions by using the logarithmic scaling relationship of the resource stress parameter; Dynamically correct weights are used to output the final total carbon emissions.

9. The carbon emission assessment method based on multi-source heterogeneous data according to claim 8, characterized in that: The expression for calculating the total carbon emissions in the target area is: Where, Indicates the total carbon emissions in the target area; C i represents the total carbon emissions of various types of vehicles; log α The logarithmic function with α as the base represents the nonlinear scaling relationship of resource tension; N represents the total number of vehicles in the target area; P i Indicates the number of valid parking spaces in the target area; V i represents the quantitative value of parking space usage intensity; α represents the logarithmic base adjustment coefficient, which is used to control the smoothness of the curve of the resource tension influencing factor; i represents the i-th target area.

10. A carbon emission assessment system based on multi-source heterogeneous data, characterized by: The carbon emission assessment method based on multi-source heterogeneous data according to any one of claims 1 to 9 is adopted, wherein the system comprises: a data acquisition module configured to obtain spatial matching data between the target area boundary and the parking spaces and generate valid parking spaces and distribution information; The parking space usage intensity calculation module is configured to integrate time-series dynamic remote sensing data, traffic density fluctuation data, and demographic data to construct a parking space activity factor model and output parking space usage intensity; a vehicle annual average travel distance calculation module configured to use a neural network to predict the distribution ratio of vehicle types in a target area and calculate the vehicle annual average travel distance; A vehicle carbon emission factor calculation module is configured to calculate a dynamic vehicle carbon emission factor and a static vehicle carbon emission factor based on the temporal dynamics and regional differences of grid energy; The total carbon emissions calculation module is configured as a coordinated correction mechanism based on parking resource tension and parking space usage intensity, integrating the vehicle type distribution ratio, average annual driving distance and carbon emission factor to calculate the total carbon emissions of the target area.

Images