Traffic multi-mode connection calculation method based on travel demand prediction

Abstract

The invention discloses a traffic multi-mode connection calculation method based on travel demand prediction, and belongs to the technical field of computers, and the method comprises the steps: S1, obtaining multi-mode traffic data, and carrying out the preprocessing of the multi-mode traffic data; s2, according to historical data and real-time monitoring data, combined with a multivariable time sequence analysis and machine learning model, carrying out subway station passenger flow demand prediction; s3, constructing a multi-objective optimization hierarchical model; s4, solving the multi-objective optimization hierarchical model; according to the traffic multi-mode connection calculation method based on travel demand prediction provided by the invention, the established optimization model not only provides a theoretical optimal solution, but also generates a traffic connection scheme with actual operability, and the scheme can effectively reduce the total travel time, cost and carbon emission, so that the method is suitable for large-scale popularization and application. The operation efficiency and the management level of an urban multi-mode traffic system can be remarkably improved, and powerful support is provided for smart city construction.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a traffic multi-mode connection calculation method based on travel demand prediction. Background Art

[0002] The current research and application of multimodal transportation connection optimization mainly focuses on the following aspects:

[0003] Single-mode traffic optimization: Traditional traffic optimization models usually focus on the optimization of a single mode of transportation. For example, bus scheduling optimization, subway operation optimization, etc. These models usually only consider factors such as time, cost or service level of a single travel mode, ignoring the complexity and coordination of multi-mode transportation connections, and it is difficult to meet the diverse travel needs of passengers.

[0004] Preliminary research on multimodal transportation connection: In recent years, with the complexity of urban transportation demand, research on multimodal transportation connection has gradually attracted attention. Researchers have explored how to effectively connect and optimize multiple transportation modes by establishing a multimodal transportation network model that integrates different transportation modes. However, these studies usually only focus on minimizing time or cost, and lack consideration of factors such as environmental impact (such as carbon emissions).

[0005] Application of intelligent transportation system: With the development of intelligent transportation technology, some studies have begun to try to apply technologies such as big data and artificial intelligence to the optimization of multi-modal transportation connections. The introduction of these technologies has improved the real-time and intelligent level of the model, but it has also brought about an increase in computational complexity, and put forward higher requirements for the real-time solution capability of the model.

[0006] The existing research still has the following deficiencies:

[0007] 1. Lack of global research: Currently, most studies are still limited to the local optimization of a certain mode of transportation, and lack global and coordinated analysis of the entire transportation system.

[0008] 2. Insufficient consideration of environmental factors: Existing multimodal connection models usually focus on the optimization of time and cost, and rarely consider the impact of environmental factors such as carbon emissions.

[0009] 3. Insufficient dynamic demand prediction: Existing models often assume that travel demand is static and lack the ability to predict and respond to dynamic changes in travel demand. This may limit the applicability of the model in practical applications.

[0010] In summary, although the research on multimodal transportation connection has made some progress, there are still many problems to be solved, especially in how to optimize travel time, cost and carbon emissions at the same time, how to deal with dynamic demand changes, etc. There is still a lot of room for improvement. Therefore, on this basis, a multimodal transportation connection optimization model integrating multi-objective optimization and dynamic demand prediction is proposed, which can better meet the actual transportation needs. Summary of the invention

[0011] The purpose of the present invention is to provide a traffic multi-mode connection calculation method based on travel demand prediction to solve the problems existing in the above-mentioned background technology.

[0012] To achieve the above object, the present invention provides a method for calculating multi-modal traffic connection based on travel demand prediction, comprising the following steps:

[0013] S1. Acquire and pre-process multi-modal traffic data, which includes subway passenger flow data, shared bicycle data, bus data, community data, and road and parking lot data;

[0014] S2. Based on historical data and real-time monitoring data, combined with multivariate time series analysis and machine learning models, subway station passenger flow demand forecasting is performed;

[0015] S3. Construct a multi-objective optimization hierarchical model, including:

[0016] S31. define a multi-objective optimization hierarchical model;

[0017] S32, construct constraint conditions;

[0018] S33, introduce dynamic demand;

[0019] S4. Solve the multi-objective optimization hierarchical model.

[0020] Preferably, preprocessing the multi-modal traffic data in S1 includes:

[0021] S11. Perform data cleaning on multimodal traffic data, remove outliers (such as missing latitude and longitude information or unreasonable timestamps), and unify the time formats and coordinate systems of different data sources;

[0022] S12, uniformly converting the geographic coordinate information in the data into a plane coordinate system for subsequent path calculation and optimization analysis;

[0023] S13, according to the actual road network structure calculation, determine the shortest path distance of a single vehicle by using the Dijkstra algorithm;

[0024] S14, dividing the time-related data into fixed time windows to ensure data synchronization, as follows:

[0025] T segments ={T k |T k =[t start +k·Δt,t start +(k+1)·Δt],k=0,1,...,n-1};

[0026] Among them, T segments is the set of divided time periods, T k is the kth time period, and Δt is the length of the time period.

[0027] Preferably, S2 specifically includes:

[0028] S21. Carry out traffic modeling:

[0029]

[0030] Among them, N t is the passenger flow at time t, α is a constant term, β i is the autoregressive term coefficient, θ j is the moving average coefficient, ∈ t is the error term;

[0031] S22. Considering the impact of exogenous factors on passenger flow, such as weather, holidays, etc., exogenous variables are introduced to improve the prediction accuracy, which is:

[0032]

[0033] Among them, X k,t is the value of the kth exogenous variable at time t, γ k is the influence coefficient of the exogenous variable;

[0034] S23, comparing the prediction result with the real-time monitoring data, dynamically correcting the prediction result using the real-time monitoring data, and realizing online correction of the prediction result through the Kalman filter, which is:

[0035]

[0036] in, is the revised predicted passenger flow, To monitor the passenger flow in real time, K t is the Kalman gain.

[0037] Preferably, S31 specifically includes:

[0038] S311. Definition level 1: The goal of minimizing total travel time and total carbon emissions is:

[0039]

[0040] Where T is the total travel time, x i,j,k Is the kth mode of transportation selected on path i, j? If selected, it is 1, otherwise it is 0. is the bus route set, t i,j,k is the travel time of the kth mode of transportation on path i, j, w i,j,k is the waiting time of the kth mode of transportation on path i, j, E is the total carbon emission, e i,j,k is the carbon emission of the kth mode of transportation on path i, j;

[0041] S312. Define the objective of minimizing the total cost at level 2 as:

[0042]

[0043] Among them, C is the total cost, c i,j,k is the operating cost of the kth mode of transportation on path i, j;

[0044] S313. Define a two-stage objective function:

[0045] MinZ = α·T + β·E + C;

[0046] Among them, α and β are the weight factors of time and carbon emissions.

[0047] Preferably, S32 specifically includes:

[0048] S321. Constructing flow conservation constraints:

[0049]

[0050] Among them, p i,j,k is the number of passengers on path i, j using the kth mode of transportation; P is the total flow in the system;

[0051] S322. Build capacity limit constraints:

[0052]

[0053] Among them, C k is the capacity of the k-th transportation mode;

[0054] S323. Build time window constraints:

[0055]

[0056] S324. Construct carbon emission constraints and set threshold E max :

[0057]

[0058] S325, construct path selection constraints and set the time threshold to T th :

[0059]

[0060] S326, build requirements to meet constraints:

[0061]

[0062] Among them, d i,j is the passenger demand on path i, j;

[0063] S327. Construct non-negative constraints and integer constraints:

[0064] x i,j,k ∈{0,1},p i,j,k ≥0.

[0065] Preferably, in S33, the passenger flow demand d in a future time window is ij :

[0066] d ij (t) = f predict (d ij (t-1),...,d ij (tn)).

[0067] Preferably, S4 specifically includes:

[0068] S41, initialization confirmation;

[0069] S42. Calculate the fitness function. The fitness function is used to evaluate the quality of each individual solution. The formula is:

[0070] F i =w1T i +w2C i +w3E i ;

[0071] Among them, F i is the fitness of the ith individual, T i , C i 、E i are the travel time, cost and carbon emission value of the individual, respectively, w1, w2 and w3 are the weights of different objectives;

[0072] S43, fitness selection, using roulette selection method, selection probability P i The calculation formula is:

[0073]

[0074] Among them, f i is the fitness value of the ith individual, f j is the fitness value of the jth individual, and N is the number of individuals in the population;

[0075] S44, use single-point crossover for crossover selection, the formula is as follows:

[0076] X'=X1·(1-λ)+X2·λ;

[0077] Among them, X1 and X2 are two parent individuals, and λ is a randomly selected crossover point;

[0078] S45. Perform mutation operation on the individuals after crossover. The formula is:

[0079] x'=x+rand(-Δx,Δx);

[0080] Among them, Δx is the variation amplitude, and rand is the random function;

[0081] S46. Update particle velocity and position to:

[0082] v i =ωv i +c1r1(p i -x i )+c2r2(gx i );

[0083] x i =x i +v i ;

[0084] Among them, v i is the velocity of the particle, x i is the position of the particle, ω is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers, and p i is the individual optimal position of the particle, and g is the global optimal position;

[0085] S47, iterative update until the maximum number of iterations is reached or the convergence condition is met;

[0086] S48, output the optimal solution;

[0087] S49, adjust the time weight w1 and carbon emission weight w3, observe their impact on the optimization results and conduct sensitivity analysis, which is:

[0088] F'(w1,w3)=w1'T+w3'E;

[0089] Among them, w'1 and w'3 are the adjusted weight values;

[0090] S410, dynamic adjustment and global output.

[0091] Preferably, S41 includes:

[0092] S411, optimize total travel time and carbon emissions;

[0093] S412. Optimize total travel cost.

[0094] Preferably, the optimal solution output by S48 includes the optimal travel time, travel cost, carbon emission data and the corresponding proportion of transportation mode selection.

[0095] Preferably, the global output of S410 is a transportation mode ratio table, which includes the selection probability of each transportation mode at each time period and each node, the actually adopted connection plan, and the corresponding travel time, travel cost and carbon emission data.

[0096] Therefore, the present invention adopts the above-mentioned traffic multi-mode connection calculation method based on travel demand prediction, which has the following beneficial effects:

[0097] (1) By introducing a hierarchical setting of multi-objective optimization, the connection between multiple modes of transportation such as subways, bicycles and buses is effectively coordinated. Its goal is to minimize travel time, cost and carbon emissions, achieve the optimization of the multi-modal transportation system, and significantly improve the overall efficiency and sustainability of urban transportation;

[0098] (2) Using dynamic demand forecasting operations, the system can accurately predict future changes in passenger demand based on real-time passenger flow data and historical travel patterns. This forecasting capability enables transportation management departments to more effectively adjust resource allocation and scheduling strategies, avoid resource waste and over-scheduling, and improve the system's response speed and flexibility.

[0099] (3) The hierarchical multi-objective optimization algorithm is combined with the genetic algorithm and the particle swarm optimization algorithm to solve the global optimal solution under complex constraints. Compared with the traditional single-objective optimization algorithm, this method can find a balance between multiple objectives and ensure that the optimal state is achieved in multiple dimensions such as time, cost and carbon emissions.

[0100] (4) By integrating real-time data analysis, machine learning algorithms and optimization models, the system can intelligently select transportation modes and dispatch resources. This improves the automation and intelligence level of the entire transportation system and contributes to the modern management and sustainable development of urban transportation.

[0101] (5) While optimizing transportation connections, the minimization of carbon emissions is taken into consideration, and the promotion of green travel modes (such as subway + shared bicycles) is encouraged, which helps to reduce the carbon footprint of urban transportation and promote the realization of environmental protection goals.

[0102] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 The present invention is a flow chart of a method for calculating multi-modal transportation connection based on travel demand prediction. DETAILED DESCRIPTION

[0104] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0105] See also Figure 1 , a traffic multi-mode connection calculation method based on travel demand prediction includes the following steps:

[0106] S1. Data collection and preprocessing

[0107] Multimodal traffic data collection:

[0108] The multi-modal transportation connection optimization design method of this embodiment first relies on comprehensive and accurate data collection and preprocessing. The integrity and accuracy of the data are the basis for the subsequent optimization model construction and solution.

[0109] The input data of this embodiment is mainly divided into two categories: static and dynamic. Static input includes the layout information of subway stations and bus routes, and the spatial distribution data of shared bicycles and parking lots. Dynamic input includes real-time subway passenger flow data, bus and shared bicycle usage, historical traffic flow data, and data related to external influencing factors such as weather and holidays. These input data are collected through the Amap API, subway operator interface and other external data sources, and are cleaned, formatted and uniformly mapped to a two-dimensional coordinate system before entering the prediction model to ensure data synchronization and consistency, including:

[0110] Subway passenger flow data: Obtain the passenger flow of the target subway station in different time windows through the historical passenger flow data provided by the subway operator. Assume that the number of people entering and leaving the target subway station is N t , the data records the distribution of the number of people entering and leaving in different time windows t

[0111]

[0112] Among them, N t represents the total passenger flow in time window t, P i,tIt refers to the passenger flow of the i-th entrance and exit within the time window t, and m refers to the number of entrances and exits of the target subway.

[0113] Shared bicycle data: Use the Amap API to obtain shared bicycle travel data in the target area. The data includes the starting and ending locations of the bicycle, the usage period, etc. Assume that the bicycle's travel path is P s , its travel distance d s Calculated based on the shortest path in the road network.

[0114]

[0115] Among them, d s is the shortest path distance for a single vehicle, P refers to the set of all feasible paths, d ij Refers to the distance between adjacent nodes i and j on path p.

[0116] Bus data: Use the Amap API to obtain the location of bus stops and their route information, and determine the coverage and stops of each bus route. Assume that the location of the bus stop is (x b ,y b ), its line coverage area R b It is expressed as:

[0117]

[0118] In the formula, (x b ,y b ) is the location coordinate of the bus stop, R b is the coverage of the bus line, and r is the coverage radius of the bus stop.

[0119] Cell data: Get the cell distribution information in the target area, including the cell location, size, population density, etc. Assume the cell location is (x c ,y c ), its service population P c According to the population density ρ c and the area of ​​the plot A c calculate:

[0120] P c =ρ c ·A c ;

[0121] Where P c is the service population of the community, ρ c is the population density of the community, A c is the area of ​​the community.

[0122] Road and parking lot data: Obtain the road network structure and parking lot distribution in the target area through the Amap API. Assume that the node of the road is (xr ,y r ), the parking lot capacity is C p :

[0123]

[0124] In the formula, C p Refers to the total capacity of the parking lot, c p,k is the capacity of the kth parking space, and n is the number of parking spaces.

[0125] Multimodal traffic data preprocessing

[0126] S11. Data cleaning and format unification

[0127] For the collected raw data, we first clean the data to remove outliers (such as missing latitude and longitude information or unreasonable timestamps), and process the time formats and coordinate systems of different data sources:

[0128]

[0129] Where D cleaned Cleaned dataset, Clean(d i )Data cleaning function, processing single data d i ;

[0130] S12. Geographic coordinate conversion and mapping

[0131] The geographic coordinate information (latitude and longitude) in the shared bicycle, bus and community data is uniformly converted to the plane coordinate system to facilitate subsequent path calculation and optimization analysis. The formula is:

[0132] x=R·lon·cos(lat), y=R·lat;

[0133] Where x is the transformed plane coordinate x, y is the transformed plane coordinate y, and R is the radius of the earth.

[0134] S13. Calculation of shared bicycle path distance:

[0135] According to the actual road network structure, the shortest path distance of a single vehicle is determined by the Dijkstra algorithm:

[0136]

[0137] Among them, d st is the shortest path distance from the starting point to the end point of a bicycle, p is all feasible paths from the starting point to the end point, (x i ,y i ), (x j ,y j) are the coordinates of the adjacent nodes on the path;

[0138] S14. Time window division and synchronization

[0139] Divide all time-related data into fixed time windows, for example, divide the morning peak (7:00-9:00) into multiple 15-minute sub-time periods to ensure data synchronization:

[0140] T segments ={T k |T k =[t start +k·Δt,t start +(k+1)·Δt],k=0,1,...,n-1};

[0141] Where, T segments is the set of divided time periods, T k is the kth time period, and Δt is the length of the time period.

[0142] S2. Travel Demand Forecast

[0143] Demand forecasting is a key part in the optimization of multi-modal transportation connections. This embodiment uses historical passenger flow data and real-time monitoring data, combined with multivariate time series analysis and machine learning models, to forecast passenger flow demand at subway stations. The specific steps are as follows:

[0144] S21. Traffic modeling

[0145]

[0146] Where N t is the passenger flow at time t, α is a constant term, β i is the autoregressive term coefficient, θ j is the moving average coefficient, ∈ t is the error term;

[0147] S22. Introduction of exogenous variables

[0148] Consider the impact of exogenous factors on passenger flow, such as weather, holidays, etc. This embodiment introduces these exogenous variables into the time series model to improve the prediction accuracy.

[0149]

[0150] Among them, X k,t is the value of the kth exogenous variable at time t, γ k is the influence coefficient of the exogenous variable;

[0151] S23. Adjustment and correction of forecast results

[0152] In order to improve the prediction accuracy, this embodiment compares the model prediction results with the real-time monitoring data, and uses the real-time monitoring data to dynamically correct the prediction results. The online correction of the prediction results is achieved through the Kalman Filter.

[0153]

[0154] in, is the revised predicted passenger flow, To monitor the passenger flow in real time, K t is the Kalman gain;

[0155] S3, model construction

[0156] In the model building part, in order to ensure the integrity of the model, it is necessary to define the basic building elements in the transportation network. These elements include bus station sets, subway station sets, bus line sets, urban roads and shortest paths sets, and arrival community sets. These basic building elements provide data support for subsequent model solving and optimization.

[0157] Construction of basic transportation network

[0158] Bus stop set B: Define the locations of all bus stops in the city, using set B = {b1, b2, …, b n} means that b i represents the location of the ith bus stop. The bus stop set is used to describe the spatial distribution of the bus system and is an important node for passengers to transfer and choose a route.

[0159] Subway station set S: all subway stations in the city subway system, represented by the set S = {s1, s2, …, s m} means, where s n represents the location of the i-th subway station. The subway station set is used to describe the layout of the subway system and the distribution of passenger flow.

[0160] Bus route set R: contains the information of all bus routes in the city, represented by the set R = {r1, r2, …, r p} means, where r i represents the ith bus line. The bus line set is used to analyze the operation routes and coverage of buses.

[0161] Urban roads and shortest path set T: used to describe the urban road network structure and the shortest distance between each path, using the set T = {t1, t2, …, t q} means, where t i represents the i-th urban road. This set is particularly suitable for path planning for multiple travel modes such as online car-hailing and walking.

[0162] Arrival cell set D: contains all cell information in the target area, with set D = {d1, d2, ..., d l} means, where d i Represents the i-th community. This set is used to describe the starting or ending point requirements of passengers

[0163] S31. Define a multi-objective optimization hierarchical model

[0164] S311. Definition level 1: Goals for minimizing total travel time and total carbon emissions:

[0165]

[0166] Where T is the total travel time, x i,j,k Is the kth mode of transportation selected on path i, j? If selected, it is 1, otherwise it is 0. is the bus route set, t i,j,k is the travel time of the kth mode of transportation on path i, j, w i,j,k is the waiting time of the kth mode of transportation on path i, j, E is the total carbon emission, e i,j,k is the carbon emission of the kth mode of transportation on path i, j;

[0167] S312. Define the level 2 total cost minimization objective:

[0168]

[0169] Among them, C is the total cost, c i,j,k is the operating cost of the kth mode of transportation on path i, j;

[0170] S313. Define the two-stage objective function:

[0171] MinZ = α·T + β·E + C;

[0172] Among them, α and β are the weight factors of time and carbon emissions.

[0173] S32. Construction of constraints:

[0174] S321. Constructing flow conservation constraints:

[0175]

[0176] Among them, p i,j,k is the number of passengers on path i, j using the kth mode of transportation; P is the total flow in the system;

[0177] S322. Build capacity limit constraints:

[0178]

[0179] Among them, C k is the capacity of the k-th transportation mode;

[0180] S323. Build time window constraints:

[0181]

[0182] S324. Construct carbon emission constraints and set threshold E max :

[0183]

[0184] S325, construct path selection constraints and set the time threshold to T th :

[0185]

[0186] S326, build requirements to meet constraints:

[0187]

[0188] Among them, d i,j is the passenger demand on path i, j;

[0189] S327. Construct non-negative constraints and integer constraints:

[0190] x i,j,k ∈{0,1},p i,j,k ≥0;

[0191] S33, Dynamic Demand Introduction

[0192] Refer to the travel demand forecast section, for the passenger flow demand d in a certain time window in the future ij

[0193] d ij (t) = f predict (d ij (t-1),...,d ij (tn)).

[0194] S4. Model solution

[0195] S41, initialization determination:

[0196] This embodiment defines three main optimization objectives: minimizing the total travel time T, minimizing the total cost C, and minimizing the total carbon emissions E. When solving multi-objective optimization problems, the model adopts hierarchical goal setting. The first thing to optimize is the total travel time and total carbon emissions, ensuring that the choice of transportation mode can achieve the optimal solution while meeting the time and environmental friendliness requirements. After completing the first-level goal optimization, the total travel cost is optimized on this basis. Therefore, the entire solution process is divided into two stages:

[0197] S411, optimize total travel time and carbon emissions;

[0198] S412, optimize total travel cost;

[0199] In this process, the population needs to be initialized. In genetic algorithms and particle swarm optimization algorithms, the population is represented as a set of possible solutions. The initial population is determined by randomly generating possible solutions. The population size is N, and the algorithm parameters are initialized, including the crossover rate and mutation rate in genetic algorithms, and the inertia weight and acceleration constant in particle swarm optimization algorithms.

[0200] S42, fitness function calculation

[0201] The fitness function is used to evaluate the quality of each individual solution. Its core is to measure the performance of the individual solution on the optimization target. The specific formula is as follows:

[0202] F i =w1T i +w2C i +w3E i ;

[0203] Among them, F i is the fitness of the ith individual, T i , C i 、E i are the individual's travel time, cost and carbon emission values, respectively, and w1, w2 and w3 are the weights of different objectives. In the solution process, according to the needs of hierarchical objectives, the first stage will focus on adjusting w1 and w3 to optimize time and carbon emissions. In the second stage, w2 will be adjusted to further optimize travel costs.

[0204] S43, fitness selection

[0205] The fitness selection operation determines which individual solutions can enter the next generation. In order to ensure the priority optimization of travel time and carbon emissions, the probability of fitness selection is allocated according to the fitness value of the first-level target. The roulette selection method is used to select the probability P i The calculation formula is:

[0206]

[0207] Among them, f i is the fitness value of the ith individual, f j is the fitness value of the jth individual, N is the number of individuals in the population, the roulette wheel selection method, the higher the fitness, the greater the probability, i is the individual for calculating the probability (the individual currently being calculated), and j is the index of the traversal population (all individuals in the population).

[0208] S44, cross selection

[0209] In the crossover operation, the selected individuals generate new individuals through crossover. The crossover rate determines the probability of the crossover operation. Commonly used crossover methods include single-point crossover, multi-point crossover, and uniform crossover. The formula for single-point crossover is as follows:

[0210] X'=X1·(1-λ)+X2·λ;

[0211] Among them, X1 and X2 are two parent individuals, and λ is a randomly selected crossover point;

[0212] S45. Mutation operation

[0213] Perform mutation operations on individuals after crossover to avoid local optimal solutions. Mutation rate p m The probability of each gene locus mutating is determined. The formula for mutation is:

[0214] x'=x+rand(-Δx,Δx);

[0215] Among them, Δx is the variation amplitude, and rand is the random function;

[0216] S46. Update particle speed and position

[0217] For the particle swarm optimization algorithm, the update formulas for particle velocity and position are as follows:

[0218] v i =ωv i +c1r1(p i -x i )+c2r2(gx i );

[0219] x i =x i +v i ;

[0220] Among them, v i is the velocity of the particle, x i is the position of the particle, ω is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers, and p i is the individual optimal position of the particle, and g is the global optimal position;

[0221] S47, delivery update

[0222] The model is updated through generations until the maximum number of iterations is reached or the convergence condition is met. During the entire iteration process, the individuals with the best fitness are retained and the global optimal solution is updated according to their fitness values.

[0223] S48, output optimal solution

[0224] After all iterations are completed, the optimal individual and its corresponding solution are output, including the optimal time, cost and carbon emission data, as well as the corresponding proportion of transportation mode selection.

[0225] S49, Sensitivity analysis

[0226] In order to verify the robustness of the optimization results, the model performs sensitivity analysis on the changes in different target weights. The time weight w1 and carbon emission weight w3 are adjusted to observe their impact on the optimization results. The formula for sensitivity analysis is as follows:

[0227] F'(w1,w3)=w1'T+w3'E;

[0228] Among them, w'1 and w'3 are the adjusted weight values;

[0229] S410, dynamic adjustment and global output

[0230] The traffic system is adjusted according to the generated scheme, and a dynamic adjustment module is set up through real-time monitoring and threshold feedback mechanism. If the predicted demand exceeds the threshold of RMSE = 0.05 compared with the actual demand, the connection scheme is dynamically optimized according to the actual demand. The global output after dynamic adjustment not only includes the optimal values ​​of the connection ratio, travel time, cost and carbon emissions of a single mode of transportation, but also includes the synergy of the entire multimodal transportation system. Through the coordinated optimization of various modes of transportation, the system can output the transportation mode selection weight of each node, as well as the comprehensive optimal solution of the total travel time, total cost and carbon emissions of the entire transportation network. Finally, the optimized global output is a transportation mode ratio table, which contains the selection probability of each mode of transportation at each time period and each node, the actual connection scheme adopted, and the corresponding time, cost and carbon emission data. This output result will be used for dynamic adjustment and decision support.

[0231] Therefore, the present invention adopts the above-mentioned traffic multi-mode connection calculation method based on travel demand prediction. In the data collection and preprocessing stage, it provides basic data support for the subsequent demand prediction and optimization model construction; in the demand prediction stage, it uses time series analysis and Kalman filter model to dynamically predict the travel demand within a certain time window in the future; by analyzing historical data and real-time data, the travel demand forecast value of each traffic node in different time periods is generated; the prediction results will be used as the input of the subsequent optimization model to provide a basis for the formulation of the connection plan; the mixed integer linear programming (MILP) method is used to construct the optimization model. The model comprehensively considers the three major goals of passengers' travel time, travel cost and carbon emissions. Through the hierarchical setting of multi-objective optimization, the total travel time and carbon emissions are optimized first, and then the total travel cost is further optimized. The model also introduces dynamic demand forecast results to comprehensively optimize the traffic connection plan. In order to improve the efficiency of solution, the model genetic algorithm and particle swarm optimization algorithm are used to solve the problem to ensure the global optimality of the solution results; in the result analysis and solution output stage, the connection scheme of each mode of transportation is generated according to the results of the model solution, including the selection ratio, total travel time, total cost and carbon emissions. Through the sensitivity analysis of the results, the optimization effect under different weight settings is explored, and the final connection scheme is output. This scheme can provide decision support for urban transportation planning, especially for the optimization design of large-scale urban transportation systems.

[0232] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for calculating multi-modal transportation connection based on travel demand prediction, characterized in that: The following steps are involved: S1. Acquire and pre-process multi-modal traffic data, which includes subway passenger flow data, shared bicycle data, bus data, community data, and road and parking lot data; S2. Based on historical data and real-time monitoring data, combined with multivariate time series analysis and machine learning models, subway station passenger flow demand forecasting is performed; S3. Construct a multi-objective optimization hierarchical model, including: S31. define a multi-objective optimization hierarchical model; S32, construct constraint conditions; S33, introduce dynamic demand; S4. Solve the multi-objective optimization hierarchical model.

2. The method for calculating multi-modal connection of traffic based on travel demand prediction according to claim 1, characterized in that: In S1, multimodal traffic data is preprocessed, including: S11, cleaning the multi-modal traffic data, removing outliers, and processing the time formats and coordinate systems of different data sources; S12, converting the geographic coordinate information in the data into a plane coordinate system; S13, determining the shortest path distance of a bicycle by using Dijkstra algorithm; S14, dividing the time-related data according to fixed time windows into: T segments ={T k |T k =[t start +k·Δt,t start +(k+1)·Δt],k=0,1,...,n-1}; Among them, T segments is the set of divided time periods, T k is the kth time period, and Δt is the length of the time period.

3. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 1, characterized in that: S2 specifically includes: S21. Carry out traffic modeling: Among them, N t is the passenger flow at time t, α is a constant term, β i is the autoregressive term coefficient, θ j is the moving average coefficient, ∈ t is the error term; S22, introduce exogenous variables, which are: Among them, X k,t is the value of the kth exogenous variable at time t, γ k is the influence coefficient of the exogenous variable; S23, comparing the prediction result with the real-time monitoring data, and implementing online correction of the prediction result through the Kalman filter, which is: in, is the revised predicted passenger flow, To monitor the passenger flow in real time, K t is the Kalman gain.

4. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 1, characterized in that: S31 specifically includes: S311. Definition level 1: The goal of minimizing total travel time and total carbon emissions is: Where T is the total travel time, x i,j,k Is the kth mode of transportation selected on path i, j? If selected, it is 1, otherwise it is 0. is the bus route set, t i,j,k is the travel time of the kth mode of transportation on path i, j, w i,j,k is the waiting time of the kth mode of transportation on path i, j, E is the total carbon emission, e i,j,k is the carbon emission of the kth mode of transportation on path i, j; S312. Define the objective of minimizing the total cost at level 2 as: Among them, C is the total cost, c i,j,k is the operating cost of the kth mode of transportation on path i, j; S313, define a two-stage objective function: MinZ = α·T + β·E + C; Among them, α and β are the weight factors of time and carbon emissions.

5. The method for calculating multi-modal connection of traffic based on travel demand prediction according to claim 1, characterized in that: S32 specifically includes: S321. Constructing flow conservation constraints: Among them, p i,j,k is the number of passengers on path i, j using the kth mode of transportation; P is the total flow in the system; S322. Build capacity limit constraints: Among them, C k is the capacity of the k-th transportation mode; S323. Build time window constraints: S324. Construct carbon emission constraints and set threshold E max : S325, construct path selection constraints and set the time threshold to T th : S326, build requirements to meet constraints: Among them, d i,j is the passenger demand on path i, j; S327. Construct non-negative constraints and integer constraints: x i,j,k ∈{0,1},p i,j,k ≥0。 6. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 1, characterized in that: The passenger flow demand d in a future time window in S33 ij : d ij (t)=f predict (d ij (t-1),...,d ij (t-n))。 7. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 1, characterized in that: S4 specifically includes: S41, initialization confirmation; S42, calculate the fitness function, the formula is: F i =w1T i +w2C i +w3E i ; Among them, F i is the fitness of the ith individual, T i , C i 、E i are the travel time, cost and carbon emission value of the individual, respectively, w1, w2 and w3 are the weights of different objectives; S43, fitness selection, using roulette selection method, selection probability P i The calculation formula is: where f i is the fitness value of the ith individual, f j is the fitness value of the jth individual, and N is the number of individuals in the population; S44, use single-point crossover for crossover selection, the formula is as follows: X'=X1·(1-λ)+X2·λ; Among them, X1 and X2 are two parent individuals, and λ is a randomly selected crossover point; S45. Perform mutation operation on the individuals after crossover. The formula is: x'=x+rand(-Δx,Δx); Among them, Δx is the variation amplitude, and rand is the random function; S46. Update particle velocity and position to: v i =ωv i +c1r1(p i -x i )+c2r2(g-x i ); x i =x i +v i ; Among them, v i is the velocity of the particle, x i is the position of the particle, ω is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers, and p i is the individual optimal position of the particle, and g is the global optimal position; S47, iterative update until the maximum number of iterations is reached or the convergence condition is met; S48, output the optimal solution; S49, adjust the time weight w1 and carbon emission weight w3, observe their impact on the optimization results and conduct sensitivity analysis, which is: F'(w1,w3)=w1'T+w3'E; Among them, w'1 and w'3 are the adjusted weight values; S410, dynamic adjustment and global output.

8. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 7, characterized in that: S41 includes: S411, optimize total travel time and carbon emissions; S412. Optimize total travel cost.

9. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 7, characterized in that: The optimal solution output by S48 includes the optimal travel time, travel cost, carbon emission data and the corresponding proportion of transportation mode selection.

10. The method for calculating multi-modal traffic connection based on travel demand prediction according to claim 7, characterized in that: The global output of S410 is a transportation mode ratio table, which includes the selection probability of each transportation mode at each time period and each node, the actual connection plan adopted, and the corresponding travel time, travel cost and carbon emission data.

Images