Personal travel carbon emission calculation method and device based on ABM and deep learning

Abstract

The invention relates to the technical field of carbon emission control, in particular to a personal travel carbon emission calculation method and device based on ABM and deep learning, and the method comprises the steps: collecting related feature data of a reference travel object; wherein the related feature data comprises multi-dimensional travel information; acquiring an ABM agent model based on deep learning based on the related feature data; on the basis of the ABM agent model, acquiring travel spatio-temporal information corresponding to the to-be-tested travel mode information of the reference travel object, so as to construct and train the ABM agent model; and evaluating the trained ABM agent model by using the test set, calculating an evaluation index, and applying the trained ABM agent model to actual individual travel carbon emission calculation and analysis. According to the method, the traffic cost can be reduced, the traffic efficiency can be improved, remarkable economic benefits are brought, making of more scientific and reasonable traffic policies is helped, the urban traffic structure is optimized, low-carbon city construction is promoted, and support is provided for achieving the carbon neutralization target.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission control, and in particular to a method and device for calculating carbon emissions from personal travel based on ABM and deep learning. Background Art

[0002] With the acceleration of urbanization, problems such as traffic congestion and environmental pollution are becoming increasingly serious. As the main component of greenhouse gases, the calculation and control of carbon emissions are crucial to achieving carbon peak and carbon neutrality goals. Personal travel is one of the important sources of carbon emissions, so the calculation of personal travel carbon emissions has become particularly important.

[0003] Traditional calculation methods often rely on fixed carbon emission factors and travel data, which may not accurately reflect the actual situation. In order to accurately calculate the carbon emissions of personal travel, a personal travel carbon emission calculation method based on ABM+ deep learning model is proposed. This method can simulate individual behavior and continuously optimize the prediction model through machine learning to more accurately calculate personal travel carbon emissions. It is of great significance to promote green travel, reduce carbon emissions, and help achieve carbon peak and carbon neutrality goals. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] According to the first aspect of the present invention, the present invention claims protection for a method for calculating carbon emissions from personal travel based on ABM and deep learning, comprising:

[0006] Collect relevant characteristic data of the reference travel object; wherein the relevant characteristic data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of the traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information;

[0007] Based on the relevant feature data, an ABM agent model based on deep learning is collected;

[0008] Based on the ABM intelligent agent model, travel time and space information corresponding to the travel mode information to be tested of the reference travel object is collected for use in ABM intelligent agent model construction and training;

[0009] Use the test set to evaluate the trained ABM agent model, calculate the evaluation indicators, and apply the trained ABM agent model to the actual calculation and analysis of personal travel carbon emissions.

[0010] Furthermore, the relevant characteristic data also includes: at least one of the traveler attribute characteristic data, road network characteristic data, and meteorological characteristic data collected based on the multi-dimensional travel information; wherein,

[0011] The traveler attribute feature data includes at least one of traveler feature data of non-single-person feature sequence, single-person feature age and gender ratio feature data, and single-person feature age and gender variance feature data;

[0012] The road network feature data includes variance feature data of road attribute information in non-single-person features and variance feature data of road attribute information in single-person features;

[0013] The traveler network information includes traveler information formed based on road attribute information, age and gender traveler elements, age and gender, and age and gender curves.

[0014] Furthermore, the relevant characteristic data also includes traveler attribute characteristic data, road network characteristic data, and meteorological characteristic data;

[0015] The collection of relevant characteristic data of the reference travel object includes:

[0016] Collecting the multi-dimensional travel information based on a multi-person travel set of a reference travel object;

[0017] Collecting the traveler attribute characteristic data based on the traveler information, the traffic facility information, and the environmental information;

[0018] Collecting the road network characteristic data based on the traveler attribute characteristic data;

[0019] The meteorological characteristic data is collected based on the road network characteristic data.

[0020] Further, the selected ABM agent model is trained based on the relevant feature data to obtain the ABM agent model, including:

[0021] Acquire an ABM agent model, wherein the ABM agent model includes an input layer, an output layer, and at least one hidden layer and at least one convolutional layer located between the input layer and the output layer;

[0022] The optimized ABM intelligent agent model is trained based on the relevant feature data to obtain the ABM intelligent agent model.

[0023] Further, the at least one hidden layer includes a first hidden layer, a second hidden layer, and a third hidden layer, and the at least one hidden layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer;

[0024] The input layer, the first hidden layer, the second hidden layer, the third hidden layer, the first convolutional layer, the second convolutional layer, the third convolutional layer, and the output layer are connected in sequence;

[0025] The collected information of the input layer includes the travel mode information;

[0026] The collected information of the first hidden layer includes the road network feature data;

[0027] The collected information of the second hidden layer includes the attribute feature data of the travelers;

[0028] The collected information of the third hidden layer includes the traveler network information;

[0029] The collected information of the output layer includes the travel time and space information.

[0030] Furthermore, the collection of relevant characteristic data of reference travel objects includes:

[0031] Collect a multi-person trip set of reference trip objects;

[0032] Collecting the travel mode information and the travel time and space information based on the individual characteristics of the multi-person travel set;

[0033] Collecting the traveler information, the traffic facility information, and the environmental information based on the multi-person travel set;

[0034] Wherein, the traveler information and the traffic facility information are collected based on different tissue information of the sample population at different developmental stages;

[0035] The traveler information includes at least chromatin open region data.

[0036] Further, the multi-traveler attribute feature data includes at least one of the traveler feature data of non-single-person feature sequence, the ratio feature data of single-person feature age and gender, and the variance feature data of single-person feature age and gender; the road network feature data includes the variance feature data of road attribute information in non-single-person feature and the variance feature data of road attribute information in single-person feature; the traveler network information includes the traveler information formed based on road attribute information, age and gender traveler elements, age and gender, and age and gender curves;

[0037] Collecting the traveler attribute feature data based on the traveler information, the traffic facility information, and the environmental information, including: constructing a labeling matrix of the traveler information based on the location information of the peak area on the heat map corresponding to the traveler information, and collecting the traveler feature data of the non-single-person feature sequence based on the labeling matrix; collecting age-sex ratio level information based on the traffic facility information, constructing an age-sex ratio matrix, and collecting the single-person feature age-sex ratio feature data and the single-person feature age-sex variance feature data based on the age-sex ratio matrix and the environmental information;

[0038] The collecting of road network characteristic data based on the traveler attribute characteristic data includes: collecting the variance characteristic data of the road attribute information in the non-single-person characteristic based on the traveler characteristic data of the non-single-person characteristic sequence; collecting the location information of the road attribute information and the single-person characteristic age and gender based on the proportion characteristic data of the single-person characteristic age and gender and the variance characteristic data of the single-person characteristic age and gender; collecting the variance characteristic data of the road attribute information in the single-person characteristic based on the influence degree information of the location information and the road attribute information on the carbon footprint structure and curve;

[0039] The collecting of meteorological characteristic data based on the road network characteristic data includes: collecting the traveler information based on the variance characteristic data of the road attribute information in single-person characteristics and the variance characteristic data of the road attribute information in non-single-person characteristics.

[0040] According to the second aspect of the present invention, the present invention claims protection for a personal travel carbon emission calculation device based on ABM and deep learning, comprising:

[0041] A relevant feature data collection unit, used to collect relevant feature data of a reference travel object; wherein the relevant feature data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information;

[0042] An ABM agent model acquisition unit, used for acquiring an ABM agent model based on deep learning based on the relevant feature data;

[0043] A travel time-space collection unit, used for collecting travel time-space information corresponding to the travel mode information to be tested of the reference travel object based on the ABM intelligent agent model, so as to be used for ABM intelligent agent model construction and training;

[0044] The personal travel carbon emission calculation device based on ABM and deep learning is used to execute the personal travel carbon emission calculation method based on ABM and deep learning.

[0045] The present invention relates to the field of carbon emission control technology, and in particular to a method and device for calculating carbon emissions from personal travel based on ABM and deep learning, collecting relevant feature data of a reference travel object; wherein the relevant feature data includes multi-dimensional travel information; collecting an ABM intelligent agent model based on deep learning based on the relevant feature data; collecting travel time and space information corresponding to the travel mode information to be tested of the reference travel object based on the ABM intelligent agent model for ABM intelligent agent model construction and training; using a test set to evaluate the trained ABM intelligent agent model, calculating evaluation indicators, and applying the trained ABM intelligent agent model to actual personal travel carbon emission calculation and analysis. The present invention can reduce transportation costs, improve transportation efficiency, bring significant economic benefits, help formulate more scientific and reasonable transportation policies, optimize urban transportation structure, promote low-carbon city construction, and provide support for achieving carbon neutrality goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A workflow diagram of a method for calculating carbon emissions from personal travel based on ABM and deep learning as claimed in an embodiment of the present invention;

[0047] Figure 2 A second workflow diagram of a method for calculating carbon emissions from personal travel based on ABM and deep learning as claimed in an embodiment of the present invention;

[0048] Figure 3 A third workflow diagram of a method for calculating carbon emissions from personal travel based on ABM and deep learning as claimed in an embodiment of the present invention;

[0049] Figure 4 A fourth work flow chart of a method for calculating carbon emissions from personal travel based on ABM and deep learning as claimed in an embodiment of the present invention;

[0050] Figure 5 A structural module diagram of a personal travel carbon emission calculation device based on ABM and deep learning as claimed in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0053] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0054] The personal travel carbon emission calculation method of ABM+ deep learning model can accurately calculate the carbon emission of personal travel, improve the public's environmental awareness, and encourage more people to choose low-carbon travel methods, such as public transportation, bicycles or walking, thereby reducing greenhouse gas emissions and environmental pollution. It can also reduce transportation costs, improve transportation efficiency, and bring significant economic benefits. At the same time, accurate carbon emission data can help the government formulate more scientific and reasonable transportation policies, optimize urban transportation structure, promote low-carbon city construction, and provide support for achieving carbon neutrality goals.

[0055] From a technical perspective, artificial intelligence technology represented by deep learning algorithms promotes the in-depth mining of massive multi-dimensional travel information, can filter out redundant noise, and improve the utilization of data information. Therefore, the present invention constructs an ABM intelligent agent model with the support of deep learning algorithms and multi-dimensional travel information. Based on this, Figure 1 As shown, the first aspect of the embodiment of the present invention provides a method for calculating carbon emissions of personal travel based on ABM and deep learning, including the following steps S101 to S104:

[0056] S101. Collect relevant characteristic data of a reference travel object; wherein, the relevant characteristic data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of the traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information.

[0057] The relevant factors for calculating carbon emissions from personal travel include:

[0058] (1) Motor vehicles

[0059] Fuel vehicles: Their carbon emissions are related to factors such as fuel type (gasoline, diesel, etc.), fuel efficiency, and mileage. Different fuels have different carbon emission factors. For example, the carbon emission coefficient of gasoline is about 0.5532t carbon / t standard coal, and diesel is about 0.5913t carbon / t standard coal. In addition, the fuel efficiency of a car is affected by the model, engine technology, driving habits, etc. The more mileage, the greater the carbon emissions.

[0060] Electric vehicles: Although they do not directly produce carbon emissions during driving, the electricity used will produce some carbon emissions during the production process. This is related to the power generation structure of the local power grid. For example, in areas where coal-fired power generation is the main source of power, the indirect carbon emissions of electric vehicles are relatively high.

[0061] (2) Public transportation

[0062] Buses: Their carbon emissions are related to vehicle type (such as fuel buses, electric buses), passenger capacity, operating mileage, etc. If the bus load rate is high, the carbon emissions per passenger will be lower.

[0063] Subway: The carbon emissions of the subway are mainly related to electricity consumption, which is in turn affected by factors such as the scale of subway operations (such as the number of trains, mileage, number of stations, etc.) and passenger flow.

[0064] (3) Non-motorized vehicles and pedestrians

[0065] Bicycling and walking themselves do not produce carbon emissions, but when calculating personal travel carbon emissions, choosing cycling or walking instead of other high-carbon travel modes will have an impact on the overall carbon emissions calculation, which will be reflected in emission reductions.

[0066] S102: Collecting an ABM intelligent agent model based on deep learning based on the relevant feature data.

[0067] Specifically, the relevant feature data can be used as samples, and then a corresponding ABM agent model can be selected to train the ABM agent model, so as to obtain the ABM agent model. The ABM agent model includes the corresponding relationship between the travel mode and travel time and space of the reference travel object.

[0068] It is easy to understand that the ABM intelligent agent model generally includes an input layer, at least one intermediate layer and an output layer. Therefore, the ABM intelligent agent model in this embodiment can be trained in the following manner: the travel mode information is used as the collection information of the input layer, and the travel time and space information is used as the output data of the output layer; other data in the multi-dimensional travel information, such as data formed based on traveler information, traffic facility information, environmental information, etc., are used as the collection information of the intermediate layer of the ABM intelligent agent model, so that the correspondence between the travel mode and travel time and space of the reference travel object can be collected. Based on this correspondence, the ABM intelligent agent model can be obtained. The ABM intelligent agent model here can be determined by those skilled in the art based on actual needs.

[0069] Travel distance is an important factor in calculating carbon emissions. Generally speaking, no matter which mode of transportation is used, the longer the travel distance, the higher the carbon emissions.

[0070] Travel time and period will affect traffic congestion, which will indirectly affect carbon emissions.

[0071] (1) Traffic congestion: During traffic congestion, vehicles travel slowly and start and stop frequently, which increases fuel consumption and carbon emissions. For example, during peak hours in the morning and evening, urban roads are congested, and private cars may need to wait at idle speed or travel at low speed for a long time, which will lead to reduced engine combustion efficiency, increased fuel consumption, and correspondingly increased carbon emissions. For public transportation vehicles, although they may not increase carbon emissions per passenger as much as private cars during congestion, their overall operating efficiency will be reduced, and more vehicles may be required to be put into operation, thereby increasing total carbon emissions.

[0072] (2) Differences in energy consumption at different times: Certain modes of transportation have different energy consumption patterns at different times. For example, the temperature is lower at night, and the vehicle's engine may require more energy to maintain normal operation, which may lead to an increase in carbon emissions when driving at night compared to during the day (for fuel vehicles).

[0073] S103. Based on the ABM intelligent agent model, travel time and space information corresponding to the travel mode information to be tested of the reference travel object is collected for use in ABM intelligent agent model construction and training.

[0074] Specifically, after the ABM agent model described in step S102 is collected, the travel mode information to be tested can be input into the ABM agent model, and the ABM agent model can output the corresponding travel time and space information. More specifically, a travel mode to be tested can be designed by editing age and gender to generate the travel mode information to be tested; of course, in some embodiments, the travel mode information to be tested of the individual to be constructed and trained by the ABM agent model can also be selected based on the reference travel object, and then the travel time and space information corresponding to the travel mode information to be tested can be predicted by the ABM agent model, and based on the travel time and space information, it can be determined how to perform further ABM agent model construction and training work.

[0075] Load factor (for public transport, ride-sharing, etc.):

[0076] For multi-person shared travel modes such as public transportation and carpooling, passenger load factor is an important factor affecting carbon emissions.

[0077] Public transport: Higher occupancy means lower carbon emissions per passenger. For example, a bus with a full load of passengers will have lower carbon emissions per passenger than a bus with only a few passengers, even though they are travelling the same route, because the total carbon emissions of the vehicle are shared by more passengers.

[0078] Carpooling (including taxi carpooling, online car-hailing carpooling, etc.): Carpooling can improve vehicle utilization, reduce the number of vehicles on the road, and thus reduce overall carbon emissions. If a taxi originally only carried one passenger, but now carries three passengers through carpooling, then the carbon emissions per passenger will be reduced to one-third of the original when traveling the same distance.

[0079] S104. Use the test set to evaluate the trained ABM intelligent agent model, calculate the evaluation index, and apply the trained ABM intelligent agent model to the actual personal travel carbon emission calculation and analysis.

[0080] In summary, based on the ABM agent model construction and training method provided by the embodiment of the present invention, the multi-dimensional travel information of the reference travel object and the ABM agent model can be combined to obtain an ABM agent model with a more accurate proportion of the ABM agent model construction and training traits. Therefore, based on the ABM agent model, the travel time and space of the individual to be constructed and trained by the ABM agent model can be more accurately predicted for the travel mode to be tested of the reference travel object, thereby making the ABM agent model construction and training work of the travel object more efficient, accurate and convenient.

[0081] In some embodiments, the relevant characteristic data also includes: at least one of traveler attribute characteristic data, road network characteristic data, and meteorological characteristic data collected based on the multi-dimensional travel information.

[0082] In at least one embodiment, the traveler attribute feature data includes at least one of traveler feature data of a non-single-person feature sequence, ratio feature data of single-person feature age and gender, and variance feature data of single-person feature age and gender.

[0083] Personal behavior habits include:

[0084] (1) Driving habits: For motor vehicle drivers, bad driving habits such as sudden acceleration, sudden braking, and speeding will increase fuel consumption and thus carbon emissions. For example, frequent sudden braking and acceleration will cause incomplete combustion of fuel in the car engine, wasting fuel and increasing carbon emissions.

[0085] (2) Travel preference: Individuals’ preferences for different modes of transportation will also affect carbon emissions. Some people prefer to travel by private car, while others prefer public transportation or non-motorized vehicles. Different choices will lead to different carbon emissions.

[0086] The types and collection of multi-source travel data include:

[0087] (1) Public transportation data

[0088] Public transportation IC card data: This is a common data type in public transportation. It records the passenger's riding information, such as card number, boarding time, boarding and alighting stations, etc. By analyzing the public transportation IC card data, we can understand the passenger's travel time pattern, common routes and stations, etc. For example, by analyzing the IC card data of a large number of passengers, we can find out which stations on a certain bus line have the largest passenger flow during the morning rush hour on weekdays, thus providing a basis for optimizing bus routes and operational scheduling.

[0089] Vehicle GPS data: mainly comes from GPS devices installed on public transportation vehicles. It can provide information such as the real-time location, driving speed, and driving direction of the vehicle. Using vehicle GPS data, the operating status of public transportation vehicles can be monitored, such as whether they are running on time, whether they deviate from the normal driving route, etc. At the same time, combined with bus IC card data, the passengers' boarding and alighting stations can be more accurately calculated.

[0090] Operation management data: including bus route planning, vehicle scheduling, operation schedule and other data. These data reflect the operation organization of public transportation and are of great significance for analyzing the service level and efficiency of public transportation. For example, by comparing the actual operation data with the operation schedule, the punctuality rate of public transportation can be evaluated.

[0091] (2) Private transportation data

[0092] Data from private car equipment: such as data recorded by the car's onboard computer, including mileage, speed, fuel consumption and other information. Some high-end cars may also be equipped with more advanced sensors that can record the vehicle's acceleration, braking conditions, etc. These data are very useful for analyzing the driving patterns and energy consumption of private cars. For example, based on mileage and fuel consumption data, the fuel efficiency of a private car can be calculated, and then its carbon emissions can be estimated.

[0093] Taxi / online ride-hailing platform data: Contains order information, such as passenger boarding and alighting locations, time, travel distance, vehicle type, etc. These data not only reflect passengers' travel needs, but can also be used to analyze the operating efficiency and energy consumption of taxis / online ride-hailing vehicles. Since the operating model of taxis / online ride-hailing vehicles is different from that of private cars, the calculation of their carbon emissions also needs to consider factors such as vehicle utilization and empty driving rate. Geographic Information System (GIS) data: GIS data can provide geographic spatial information, such as terrain, road network, road slope, etc., which is very important for accurately calculating the carbon emissions of transportation modes.

[0094] ABM (Agent-Based Model) is a class of computational models used to simulate the behavior and interactions of autonomous agents (individuals or collective entities such as organizations or groups) to evaluate their impact on the entire system.

[0095] Deep learning models, such as convolutional neural networks (CNN) and recurrent neural networks (RNN), have powerful feature learning and data fitting capabilities. When combined with ABM, ABM can provide structured simulation data for deep learning, which reflects the behavior and interaction of individual agents. Deep learning models can deeply mine and analyze the data generated by ABM, and learn the complex nonlinear relationship between agent behavior patterns and carbon emissions. For example, in a transportation scenario, ABM can simulate the travel choice behaviors of different types of travelers (such as office workers, students, etc.) under different transportation networks and travel policies, and deep learning models can analyze these behavioral data to find out the key factors and patterns that affect carbon emissions.

[0096] Based on the above, continue as Figure 2 As shown, in some embodiments, when the relevant feature data also includes the traveler attribute feature data, step S101, collecting relevant feature data of the reference travel object, includes:

[0097] S1011, collecting the multi-dimensional travel information based on a sample group of reference travel objects;

[0098] S1012. Collect the traveler attribute feature data based on the traveler information, the traffic facility information, and the environmental information.

[0099] Furthermore, when the relevant characteristic data further includes the road network characteristic data, step S101, collecting relevant characteristic data of reference travel objects, further includes:

[0100] S1013. Collect the road network characteristic data based on the traveler attribute characteristic data.

[0101] Furthermore, when the relevant characteristic data also includes the meteorological characteristic data, step S101, collecting relevant characteristic data of reference travel objects, further includes:

[0102] S1014. Collect the meteorological characteristic data based on the road network characteristic data.

[0103] In some embodiments, Figure 3 As shown, step S102, collecting an ABM agent model based on deep learning based on the relevant feature data, includes:

[0104] S1021. Collecting an ABM intelligent agent model, the ABM intelligent agent model includes an input layer, an output layer, and at least one hidden layer and at least one convolutional layer located between the input layer and the output layer; wherein the collected information of each hidden layer includes one of road network feature data, traveler attribute feature data, and meteorological feature data.

[0105] S1022. Optimize the hyperparameters of each layer of the ABM intelligent agent model, determine the optimal hyperparameters, and obtain the optimized ABM intelligent agent model.

[0106] S1023. Training the optimized ABM intelligent agent model based on the relevant feature data to obtain the ABM intelligent agent model.

[0107] In some embodiments, the input layer, the at least one hidden layer, the at least one hidden layer, and the output layer are connected in sequence. For example, the at least one hidden layer includes a first hidden layer, a second hidden layer, and a third hidden layer, and the at least one hidden layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer; then the input layer, the first hidden layer, the second hidden layer, the third hidden layer, the first convolutional layer, the second convolutional layer, the third convolutional layer, and the output layer are connected in sequence.

[0108] It is easy to understand that the collected information of each hidden layer includes one of the road network feature data, the traveler attribute feature data, and the meteorological feature data, and may also include the output data of the previous layer connected thereto. For example, the collected information of the first hidden layer may also include the output data of the input layer, and the collected information of the second hidden layer may also include the output data of the first hidden layer. Furthermore, in some embodiments, the collected information of the first hidden layer includes the road network feature data and the output data of the input layer; the collected information of the second hidden layer includes the traveler attribute feature data and the output data of the first hidden layer; and the collected information of the third hidden layer includes the traveler network information and the output data of the second hidden layer.

[0109] In this embodiment, data collection and preprocessing are first performed:

[0110] Collect personal travel-related data, including basic information of travelers (such as age, occupation, etc.), travel mode selection, travel time, travel distance, etc. These data can be

[0111] The data is obtained through traffic surveillance camera data, traffic big data from Amap, vehicle GPS, mobile communication data, social media data questionnaires, and transportation card data.

[0112] Collect data related to transportation facilities, such as road network information (including road length, width, grade, etc.), bus and subway operation information (such as route distribution, number of vehicles, departure intervals, etc.), distribution of gas stations and charging stations, etc.

[0113] Collect environmental data, such as meteorological data (temperature, humidity, wind speed, etc.), because meteorological conditions will affect traffic operation efficiency and energy consumption, and thus affect carbon emissions.

[0114] Clean the collected data to remove erroneous data and outliers. For example, in the transportation card data, there may be unreasonable travel records caused by card swiping errors, which need to be identified and removed.

[0115] Standardize the data to make different types of data comparable. For example, normalize the travel distance, time and other data to facilitate subsequent model calculations.

[0116] In some embodiments, S1023, training the optimized ABM agent model based on the relevant feature data to obtain the ABM agent model includes:

[0117] In the iterative update of the ABM agent model, the carbon footprint mark of each hidden layer of the ABM agent model is collected based on the relevant feature data; in some embodiments, the carbon footprint mark carries diffusible and detectable carbon footprint sequence information, carbon footprint information or age and gender variance feature data;

[0118] Collecting the false discovery rate of the weights of the carbon footprint markers of each hidden layer;

[0119] Calculate the significance information of the carbon footprint mark of each hidden layer and the travel time and space association based on the false discovery rate;

[0120] Based on the significant information, collecting carbon footprint marks significantly associated with travel time and space from the carbon footprint marks of each hidden layer; and

[0121] Based on the significantly associated carbon footprint markers, a correspondence between each hidden layer and the relevant feature data is established. Specifically, in some embodiments, a correspondence between the meteorological feature data and the third hidden layer, a correspondence between the traveler attribute feature data and the second hidden layer, and a correspondence between the road network feature data and the first hidden layer is established.

[0122] The purpose of the above embodiment is to iteratively update the ABM agent model during the training process to acquire a more accurate ABM agent model.

[0123] Specifically, in the iterative update, the weights of the carbon footprint labels of each hidden layer of the ABM intelligent agent model are obtained by reading and retrieving the variables and variable value files saved in the ABM intelligent agent model, the weights of the carbon footprint labels under random circumstances are collected through the rank sum test, and the false discovery rate of the carbon footprint label weights is calculated to calculate the significance information of the carbon footprint labels of each hidden layer and the travel time and space, so that the carbon footprint labels of each hidden layer that are significantly associated with the travel time and space can be collected.

[0124] Define different agents according to the types of travelers in the data, such as the office workers, students, etc. Assign corresponding attributes to each agent, such as travel preferences, economic conditions, etc.

[0125] Define transportation facility entities, such as roads and stations, and set their properties, such as the road capacity and station service capacity.

[0126] Set travel selection rules for traveler agents, which can be based on actual travel surveys and behavior analysis. For example, determine the probability of travelers choosing different modes of transportation based on factors such as travel distance, travel time, and traffic congestion.

[0127] Set operating rules for intelligent entities of transportation facilities, such as traffic flow models for roads, operation and scheduling rules for buses and subways, etc.

[0128] After building the intelligent agent and setting the rules, the intelligent agent is placed in a simulated traffic environment and the state of the model is initialized, such as setting the initial traffic flow, traveler distribution, etc.

[0129] In some embodiments, the at least one convolutional layer is equal in number to the at least one hidden layer.

[0130] In some embodiments, Figure 4 As shown, step S1011, collecting the multi-dimensional travel information based on a multi-person travel set of a reference travel object includes:

[0131] S10111. Collect a multi-person travel set of reference travel objects. Specifically, a sample group of reference travel objects may be selected first, and then the multi-person travel set may be collected and saved for later use. For example, relevant feature data of each individual in the sample group may be collected, recorded and saved to obtain the multi-person travel set.

[0132] S10112. Collect the travel mode information and the travel time and space information based on the characteristics of a single person in the multi-person travel set; it is easy to understand that each individual has a set of travel mode information and travel time and space information, so by selecting or determining an individual from the sample group, the travel mode information and travel time and space information of the reference travel object can be collected based on the characteristics of the single person.

[0133] S10113. Collect the traveler information, the traffic facility information, and the environmental information based on the multi-person travel set.

[0134] It should be noted that, in certain embodiments, the traveler information and traffic facility information here can be collected based on different tissues of the sample population at different developmental stages; it should be noted that the tissues here refer to biological organ tissues.

[0135] The simulation runs according to the set time step. In each time step, the traveler agent chooses the travel mode and route according to its own rules and interacts with the transportation facility agent. For example, if a traveler chooses to take the bus, he will interact with the bus station and bus agents, affecting the bus's passenger capacity and operating status.

[0136] During the simulation, the travel behavior data of each traveler agent is recorded, including the selected mode of transportation, travel distance, travel time, etc. These data will be used as input data for the subsequent deep learning model. The deep learning model can process a large amount of traffic travel data, including historical travel records, traffic flow data, meteorological data, etc. Choose the appropriate deep learning model structure according to the problem to be solved. If you are analyzing time series data, such as travel time series data, you can choose a recurrent neural network (RNN) or its variants (such as LSTM, GRU) to mine the laws of travel demand in different time periods. If you are analyzing image data (such as traffic flow images), you can choose a convolutional neural network (CNN).

[0137] Determine the structure of the model’s input layer, hidden layer, and output layer. The input layer receives the travel behavior data generated by the ABM simulation, and the output layer can output the carbon emissions of individual travel or indicators related to carbon emissions.

[0138] The data obtained by ABM simulation is divided into training set, validation set and test set. The deep learning model is trained with the training set, and the parameters of the model are adjusted so that the model can accurately fit the data.

[0139] During the training process, the validation set is used to evaluate the performance of the model to prevent the model from overfitting. The hyperparameters of the model, such as the learning rate and the number of hidden layer nodes, are adjusted based on the evaluation results of the validation set.

[0140] In some embodiments, step S1012, collecting the traveler attribute characteristic data based on the traveler information, the traffic facility information, and the environmental information, includes:

[0141] S10121. Based on the location information of the peak area on the heat map corresponding to the traveler information collection, a labeling matrix of the traveler information is constructed, and the traveler feature data of the non-single-person feature sequence is collected based on the labeling matrix.

[0142] Furthermore, the label matrix and peak area sequence information can be modeled through the ABM intelligent model to obtain the traveler characteristic data of the non-single-person characteristic sequence.

[0143] The trained deep learning model is evaluated using the test set, and evaluation metrics such as mean square error (MSE), mean absolute error (MAE), etc. are calculated to measure the accuracy of the model.

[0144] Compare the evaluation indicators under different model structures or parameter settings and select the model with the best performance.

[0145] Apply the trained ABM+ deep learning model to the actual calculation and analysis of personal travel carbon emissions. For example, provide traffic management departments with prediction results of personal travel carbon emissions to help formulate emission reduction policies; provide travelers with personalized low-carbon travel suggestions to guide them to choose more environmentally friendly travel methods.

[0146] S10122. Based on the traffic facility information, age-gender ratio level information is collected to construct an age-gender ratio matrix. Based on the age-gender ratio matrix and the environmental information, ratio characteristic data of the characteristic age and gender of a single person and variance characteristic data of the characteristic age and gender of a single person are collected.

[0147] Furthermore, the age-sex ratio matrix can be decomposed by principal component analysis (PCA) method, and the ratio characteristic data of the age and sex of a single person and the variance characteristic data of the age and sex of a single person can be obtained by combining the environmental information.

[0148] In some embodiments, step S1013, collecting road network characteristic data based on the traveler attribute characteristic data, includes:

[0149] S10131. Based on the traveler characteristic data of the non-single-person feature sequence, the variance characteristic data of the road attribute information in the non-single-person feature is collected; illustratively, the DeepSEA model can be used to learn the apparent age and gender group signal characteristics of the non-single-person feature within the range of all age and gender groups, so as to collect the variance characteristic data of the road attribute information in the non-single-person feature.

[0150] S10132. Based on the proportion feature data of the individual's age and gender and the variance feature data of the individual's age and gender, collect the location information of the road attribute information and the individual's age and gender; based on the information on the degree of influence of the location information and the road attribute information on the carbon footprint structure and curve, collect the variance feature data of the road attribute information in the individual's characteristics.

[0151] In some embodiments, step S1014, collecting meteorological characteristic data based on the road network characteristic data, includes:

[0152] The traveler information is collected based on the variance feature data of the road attribute information in single-person characteristics and the variance feature data of the road attribute information in non-single-person characteristics.

[0153] In some embodiments, the ABM agent model construction and training method further includes:

[0154] The prediction effect of the ABM agent model was evaluated based on the Pearson correlation coefficient (PCC).

[0155] More specifically, the ABM agent model was compared and analyzed with classic ABM agent models such as rrBLUP in combination with the Pearson correlation coefficient, and then the effect of the ABM agent was comprehensively evaluated through the established independent validation set.

[0156] Based on the same concept as the ABM agent model construction and training method provided in the first aspect embodiment, refer to Figure 5 The second aspect of the embodiment of the present invention provides a personal travel carbon emission calculation device based on ABM and deep learning, including:

[0157] The relevant characteristic data collection unit 100 is used to collect relevant characteristic data of the reference travel object; wherein the relevant characteristic data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of the traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information;

[0158] An ABM agent model acquisition unit 200 is used to acquire an ABM agent model based on deep learning based on the relevant feature data;

[0159] A travel time and space collection unit 300 is used to collect the travel time and space corresponding to the travel mode to be tested of the selected individuals of the reference travel object based on the ABM intelligent agent model, so as to be used for ABM intelligent agent model construction and training;

[0160] The carbon emission analysis unit 400 uses the test set to evaluate the trained ABM intelligent agent model, calculates the evaluation index, and applies the trained ABM intelligent agent model to actual personal travel carbon emission calculation and analysis.

[0161] ABM is a computational model that simulates individual behaviors and their interactions in complex systems. In the calculation of personal travel carbon emissions, ABM can simulate factors such as different individuals' travel habits, chosen means of transportation, travel distance, etc., so as to calculate each person's carbon emissions.

[0162] Personal travel carbon emissions are affected by many factors, such as the traveler's personal attributes (age, gender, income, etc.), travel time (morning rush hour, evening rush hour, etc.), travel purpose (work, shopping, leisure, etc.), traffic environment (road congestion, convenience of public transportation, etc.). Deep learning models can handle the nonlinear relationship between these complex factors, while traditional calculation methods often find it difficult to simultaneously consider so many complex factors and their interrelationships.

[0163] Deep learning models can integrate data from different sources, such as operational data from the transportation department, personal travel record data, geographic information data, etc. For example, by combining the transportation department's bus and subway passenger flow data with personal mobile phone positioning travel trajectory data, deep learning models can mine more comprehensive travel patterns and more accurately calculate personal travel carbon emissions. Traditional calculation methods often rely on fixed carbon emission factors and travel data, which may not accurately reflect the actual situation.

[0164] Personal travel behavior and traffic environment are changing dynamically. For example, as cities develop and new transportation facilities are built, people's travel habits will change accordingly. Deep learning models can adapt to this dynamic change by constantly updating data. The parameters of the model can be readjusted according to the new data to maintain the accuracy of the calculation of personal travel carbon emissions. Traditional calculation methods may require rebuilding the model or a lot of manual adjustments to adapt to such changes.

[0165] The ABM+deep learning model personal travel carbon emissions calculation method can fully utilize the advantages of both. First, ABM can simulate individual travel behavior and generate a large amount of travel data. Then, this data can be used as input to the deep learning model to train the model to predict carbon emissions. Finally, through continuous iteration and optimization, the model can predict the carbon emissions of personal travel more and more accurately.

[0166] The personal travel carbon emissions calculation method based on ABM+ deep learning model is an innovative method that can more accurately calculate the carbon emissions of personal travel. This method not only takes into account individual behavioral differences, but also processes and analyzes a large amount of travel data, thereby improving the accuracy and practicality of the calculation.

[0167] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0168] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units. The above is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

[0169] The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims

1. A method for calculating carbon emissions from personal travel based on ABM and deep learning, characterized in that: include: Collect relevant characteristic data of the reference travel object; wherein the relevant characteristic data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of the traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information; Based on the relevant feature data, an ABM agent model based on deep learning is collected; Based on the ABM intelligent agent model, travel time and space information corresponding to the travel mode information to be tested of the reference travel object is collected for use in ABM intelligent agent model construction and training; Use the test set to evaluate the trained ABM agent model, calculate the evaluation indicators, and apply the trained ABM agent model to the actual calculation and analysis of personal travel carbon emissions.

2. The personal travel carbon emission calculation method based on ABM and deep learning according to claim 1 is characterized in that: The relevant characteristic data also includes: at least one of the traveler attribute characteristic data, road network characteristic data, and meteorological characteristic data collected based on the multi-dimensional travel information; wherein, The traveler attribute feature data includes at least one of traveler feature data of non-single-person feature sequence, single-person feature age and gender ratio feature data, and single-person feature age and gender variance feature data; The road network feature data includes variance feature data of road attribute information in non-single-person features and variance feature data of road attribute information in single-person features; The traveler network information includes traveler information formed based on road attribute information, age and gender traveler elements, age and gender, and age and gender curves.

3. The personal travel carbon emission calculation method based on ABM and deep learning according to claim 1 is characterized in that: The relevant characteristic data also includes traveler attribute characteristic data, road network characteristic data, and meteorological characteristic data; The collection of relevant characteristic data of the reference travel object includes: Collecting the multi-dimensional travel information based on a multi-person travel set of a reference travel object; Collecting the traveler attribute characteristic data based on the traveler information, the traffic facility information, and the environmental information; Collecting the road network characteristic data based on the traveler attribute characteristic data; The meteorological characteristic data is collected based on the road network characteristic data.

4. The method for calculating carbon emissions from personal travel based on ABM and deep learning according to claim 1, characterized in that: Training the selected ABM agent model based on the relevant feature data to obtain the ABM agent model includes: Acquire an ABM agent model, wherein the ABM agent model includes an input layer, an output layer, and at least one hidden layer and at least one convolutional layer located between the input layer and the output layer; The optimized ABM intelligent agent model is trained based on the relevant feature data to obtain the ABM intelligent agent model.

5. The method for calculating carbon emissions from personal travel based on ABM and deep learning according to claim 4 is characterized in that: The at least one hidden layer includes a first hidden layer, a second hidden layer, and a third hidden layer, and the at least one hidden layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer; The input layer, the first hidden layer, the second hidden layer, the third hidden layer, the first convolutional layer, the second convolutional layer, the third convolutional layer, and the output layer are connected in sequence; The collected information of the input layer includes the travel mode information; The collected information of the first hidden layer includes the road network feature data; The collected information of the second hidden layer includes the attribute feature data of the travelers; The collected information of the third hidden layer includes the traveler network information; The collected information of the output layer includes the travel time and space information.

6. The method for calculating carbon emissions from personal travel based on ABM and deep learning according to claim 1, characterized in that: The collection of relevant characteristic data of the reference travel object includes: Collect a multi-person trip set of reference trip objects; Collecting the travel mode information and the travel time and space information based on the individual characteristics of the multi-person travel set; Collecting the traveler information, the traffic facility information, and the environmental information based on the multi-person travel set; Wherein, the traveler information and the traffic facility information are collected based on different tissue information of the sample population at different developmental stages; The traveler information includes at least chromatin open region data.

7. The method for calculating carbon emissions from personal travel based on ABM and deep learning according to claim 3 is characterized in that: The multi-traveler attribute feature data includes at least one of the traveler feature data of non-single-person feature sequence, the ratio feature data of single-person feature age and gender, and the variance feature data of single-person feature age and gender; the road network feature data includes the variance feature data of road attribute information in non-single-person feature and the variance feature data of road attribute information in single-person feature; the traveler network information includes the traveler information formed based on road attribute information, age and gender traveler elements, age and gender, and age and gender curves; Collecting the traveler attribute feature data based on the traveler information, the traffic facility information, and the environmental information, including: constructing a labeling matrix of the traveler information based on the location information of the peak area on the heat map corresponding to the traveler information, and collecting the traveler feature data of the non-single-person feature sequence based on the labeling matrix; collecting age-sex ratio level information based on the traffic facility information, constructing an age-sex ratio matrix, and collecting the single-person feature age-sex ratio feature data and the single-person feature age-sex variance feature data based on the age-sex ratio matrix and the environmental information; The collecting of road network characteristic data based on the traveler attribute characteristic data includes: collecting the variance characteristic data of the road attribute information in the non-single-person characteristic based on the traveler characteristic data of the non-single-person characteristic sequence; collecting the location information of the road attribute information and the single-person characteristic age and gender based on the proportion characteristic data of the single-person characteristic age and gender and the variance characteristic data of the single-person characteristic age and gender; collecting the variance characteristic data of the road attribute information in the single-person characteristic based on the influence degree information of the location information and the road attribute information on the carbon footprint structure and curve; The collecting of meteorological characteristic data based on the road network characteristic data includes: collecting the traveler information based on the variance characteristic data of the road attribute information in single-person characteristics and the variance characteristic data of the road attribute information in non-single-person characteristics.

8. A personal travel carbon emission calculation device based on ABM and deep learning, characterized in that: include: A relevant feature data collection unit, used to collect relevant feature data of a reference travel object; wherein the relevant feature data includes multi-dimensional travel information, and the multi-dimensional travel information includes at least one of traveler information, traffic facility information, and environmental information of the reference travel object, travel mode information, and travel time and space information; An ABM agent model acquisition unit, used for acquiring an ABM agent model based on deep learning based on the relevant feature data; A travel time-space collection unit, used for collecting travel time-space information corresponding to the travel mode information to be tested of the reference travel object based on the ABM intelligent agent model, so as to be used for ABM intelligent agent model construction and training; The personal travel carbon emission calculation device based on ABM and deep learning is used to execute the personal travel carbon emission calculation method based on ABM and deep learning as described in any one of claims 1 to 7.

Images