CN117332909A - Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent - Google Patents
Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent Download PDFInfo
- Publication number
- CN117332909A CN117332909A CN202311637750.5A CN202311637750A CN117332909A CN 117332909 A CN117332909 A CN 117332909A CN 202311637750 A CN202311637750 A CN 202311637750A CN 117332909 A CN117332909 A CN 117332909A
- Authority
- CN
- China
- Prior art keywords
- road
- exposure
- scale
- urban
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 94
- 238000004364 calculation method Methods 0.000 claims abstract description 25
- 238000004458 analytical method Methods 0.000 claims abstract description 22
- 238000013461 design Methods 0.000 claims abstract description 16
- 239000003795 chemical substances by application Substances 0.000 claims description 53
- 238000001556 precipitation Methods 0.000 claims description 42
- 230000008859 change Effects 0.000 claims description 39
- 230000006399 behavior Effects 0.000 claims description 35
- 238000012706 support-vector machine Methods 0.000 claims description 33
- 238000012549 training Methods 0.000 claims description 31
- 230000000694 effects Effects 0.000 claims description 27
- 241000282461 Canis lupus Species 0.000 claims description 26
- 241000976924 Inca Species 0.000 claims description 24
- 230000002159 abnormal effect Effects 0.000 claims description 21
- 239000000463 material Substances 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 19
- 241000282421 Canidae Species 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 16
- 238000004088 simulation Methods 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000003066 decision tree Methods 0.000 claims description 12
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 11
- 238000005094 computer simulation Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000005516 engineering process Methods 0.000 claims description 9
- 238000012502 risk assessment Methods 0.000 claims description 9
- 238000010801 machine learning Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000011160 research Methods 0.000 claims description 8
- 230000008878 coupling Effects 0.000 claims description 7
- 238000010168 coupling process Methods 0.000 claims description 7
- 238000005859 coupling reaction Methods 0.000 claims description 7
- 230000032683 aging Effects 0.000 claims description 6
- 238000013145 classification model Methods 0.000 claims description 6
- 238000006073 displacement reaction Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000033001 locomotion Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000010792 warming Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000000638 solvent extraction Methods 0.000 claims description 5
- 230000007704 transition Effects 0.000 claims description 5
- 230000003139 buffering effect Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 230000035772 mutation Effects 0.000 claims description 4
- 206010034701 Peroneal nerve palsy Diseases 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 3
- 230000000903 blocking effect Effects 0.000 claims description 3
- 238000012512 characterization method Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000009792 diffusion process Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000001303 quality assessment method Methods 0.000 claims description 3
- 238000013139 quantization Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 230000011273 social behavior Effects 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000011002 quantification Methods 0.000 claims description 2
- 230000005856 abnormality Effects 0.000 claims 2
- 238000009395 breeding Methods 0.000 claims 1
- 230000001488 breeding effect Effects 0.000 claims 1
- 238000000556 factor analysis Methods 0.000 claims 1
- 238000010561 standard procedure Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000010485 coping Effects 0.000 abstract description 2
- 238000009825 accumulation Methods 0.000 description 7
- 230000003542 behavioural effect Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000012876 topography Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000001376 precipitating effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 230000000630 rising effect Effects 0.000 description 2
- 230000007480 spreading Effects 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 239000010426 asphalt Substances 0.000 description 1
- 239000004568 cement Substances 0.000 description 1
- 239000004035 construction material Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000001970 hydrokinetic effect Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Primary Health Care (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a multi-scale urban waterlogging road traffic exposure prediction method based on an intelligent agent, which mainly comprises the following steps: s1: road exposure influence characteristic analysis under waterlogging situations; s2: dividing a road traffic space multi-scale exposure unit; s3: a multi-scale urban road traffic exposure prediction coupled with future climate data; s4: carrying out quantitative calculation on the road traffic exposure of rainstorm and waterlogging; s5: urban road traffic exposure in future climate is predicted based on the agent. The invention has the beneficial effects that the invention designs a theoretical framework of road traffic exposure, and utilizes an intelligent body model to couple a future climate mode to realize multi-scale prediction of road traffic exposure, thereby assisting government decision-making departments to formulate effective early warning systems and coping strategies.
Description
Technical Field
The invention relates to the technical field of disaster assessment prediction, in particular to a multi-scale urban waterlogging road traffic exposure prediction method based on an intelligent agent.
Background
Urban and climate change causes frequent occurrence of rainstorm and waterlogging, which threatens urban sustainable development and mass life. Traffic networks are critical to daily travel and social development. Once being affected by storm and waterlogging, abnormal road traffic can affect urban road operation. Therefore, how to simulate and predict the evolution rule of road traffic in a rainstorm waterlogging event is helpful for government decision-making departments to formulate effective early warning systems and coping strategies so as to ensure the reliability and safety of urban traffic and promote the sustainable development of cities.
The traditional method extracts and analyzes the data of the disaster-stricken road according to a specific rule from the angle of traffic simulation; or from the perspective of disaster simulation, a hydrokinetic model is used to simulate and calculate the submerged road conditions. Urban inland inundation is a natural-social composite disaster, and road traffic exposure is influenced from a multidimensional perspective, and the influence of hydrologic process and human activities is comprehensively considered.
The traditional method lacks of dynamic simulation of traffic exposure under geographic environment constraint, the simulation method focuses on commute rule design, and space-time constraint on urban geographic environment factors is not considered enough in disaster response.
In traditional flood forecasting, the input of a waterlogging model mainly depends on ground observation station data. The observation sites are sparse, the precipitation distribution difference is large, the precipitation characteristics are difficult to accurately describe by means of rainfall, and the requirement of multi-scale urban road traffic waterlogging prediction cannot be met.
Disclosure of Invention
Aiming at the problems of the technology, the invention provides a multi-scale urban waterlogged road traffic exposure prediction method based on an intelligent agent, which aims at designing a road traffic exposure theoretical frame and realizing multi-scale road traffic exposure prediction based on a future climate mode by utilizing an intelligent agent model. The invention is realized by the following technical scheme:
The invention discloses a multi-scale urban waterlogging road traffic exposure prediction method based on an intelligent agent, which mainly comprises the following steps:
s1: road exposure influence characteristic analysis under waterlogging situations;
s2: dividing a road traffic space multi-scale exposure unit;
s3: a multi-scale urban road traffic exposure prediction coupled with future climate data;
s4: carrying out quantitative calculation on the road traffic exposure of rainstorm and waterlogging;
s5: urban road traffic exposure in future climate is predicted based on the agent.
Further, the step S1 includes the following substeps:
s11: analyzing road traffic exposure influence factors;
s12: analyzing the road traffic exposure forming process;
s13: dominant and recessive impact feature recognition.
Further, the S11 urban waterlogging disaster is mainly generated by the comprehensive action of three parts, and is specifically as follows:
s111: and analyzing disaster causing factors. Disaster causing factors are direct causes of disasters, and refer to dynamic factors possibly causing disasters and subsequent socioeconomic losses;
s112: and analyzing the pregnant disaster environment. The regional differences determine the spatial-temporal distribution characteristics of the disaster-causing factors. The pregnant disaster environment refers to natural and artificial environments for the pregnant urban waterlogging disasters. Natural environments include earth surface topography and the like, while artificial environments include urban drainage capacity, urban land function area distribution and the like;
S113: and analyzing the disaster bearing body. The disaster-bearing body of the invention is an urban road, comprising:
road speed. Since the road speed is mainly detected by a sensor such as a loop detector, it is difficult to acquire the value. In order to obtain road speed data at different time points, the invention utilizes vehicle GPS data to match vehicle track points with corresponding roads by using a map matching method. The average speed of the road is the average speed of all track points passing through the road section in the time unit, and the calculation formula is as follows:
，
wherein,the number of target road track points per unit time, +.>For the track point on the road section +.>Instantaneous speed;
road elevation. In order to correlate raw elevation (Digital Elevation Model, DEM) data with urban road data, the invention assigns grid-type elevation data to the corresponding vector roads;
road intermediation centrality. The mediating centrality is used to quantify the importance of a node in a network graph, reflecting the number of shortest paths through the node in the graph;
land use type. Land use types are divided into park, residential and commercial uses;
road surface material. There is a certain relationship between the road surface material and the road traffic exposure. The modes of different pavement materials influencing traffic exposure are different, and the roughness of the pavement materials can influence the running safety and the traffic exposure of the vehicle;
Road grade. The roads are divided into expressways, urban main roads, secondary main roads and branch roads according to the grades;
number of lanes. The number of lanes refers to the number of lanes on a road that a vehicle can travel. Since the road grade is not directly related to the number of lanes, the road traffic exposure needs to be studied by means of the feature of the number of lanes;
road length. The length of a link refers to the distance between two end points and is generally used to measure the buffering capacity provided by the link when it is congested.
Further, the step S12 includes the following substeps:
s121: and detecting rainfall abnormal roads. The rainfall abnormal road means that after rainy days, the speed change of part of vehicles on the road causes the creep or congestion of the surrounding road. In order to detect rainfall abnormal roads, the invention carries out statistical comparison on two types of road speed data under the normal situation of a target area and under the rainstorm waterlogging situation;
s122: road extraction affected by rainfall
According to the statistical result, the road types are divided into three types: a road with a significantly rising speed, a road with no significant change in speed, and a road with a significantly falling speed. In order to analyze more accurately, the invention combines the rainfall monitoring point data of the target area to carry out statistical analysis and calculate the average change rate of the road speed;
S123: and (5) extracting the rainfall abnormal road. Road speed time series data during rainfall is analyzed using a Mann-Kendall abrupt change detection method to detect a speed change situation. The method analyzes all roads in a target area, counts the roads with speed mutation and represents the position through a map;
s124: road traffic congestion spread analysis. In urban traffic, sporadic road congestion usually initially occurs at a specific point, such as an intersection, and only directly affects roads associated with the space. In order to analyze the influence of an abnormal road caused by initial heavy rain on the surrounding road conditions, the invention adopts a local Morgan index to analyze the spatial autocorrelation of the abnormal road caused by the heavy rain. The spatial autocorrelation analysis method can be used for researching the spatial distribution characteristics of the geographic elements, namely analyzing the potential correlation existing between the geographic elements in the same research area, and further analyzing the road traffic jam diffusion phenomenon caused by rainfall and waterlogging. The local molan index is calculated as follows:
，
in the method, in the process of the invention,is a variable->Is the variable mean value->Is the number of geographic elements->Is a weight matrix.
Further, the step S13 includes the following substeps:
S131: dominant impact feature recognition. The dominant impact features refer to the dominant impact features of the occurrence of puddles and the like on the road when the exposure is calculated. Puddles may cause traffic anomalies, resulting in traffic jams. Sometimes, the blocking situation is propagated through the adjacent roads, so that the exposure degree of the adjacent roads is increased;
s132: implicit impact feature recognition. Implicit impact characteristics refer to the potential impact on the exposure of a road when calculating the exposure. The invention calculates the exposition by quantifying the indirect exposition feature;
further, the spatial multi-scale effect of the S2 urban waterlogging disaster generates obvious multi-scale characteristics on the exposure distribution pattern and change mechanism of road traffic. The invention designs four road traffic space scale exposure units, which specifically comprise: primary large scale exposition units, i.e. only including urban expressways; a second-level larger exposure unit, namely only comprising urban expressways and main roads; three-level mesoscale exposure units, namely, urban expressways, arterial roads and secondary arterial roads; four-level small-scale exposition units, namely, urban expressways, main roads, secondary main roads and branches. Specifically, the road traffic space multi-scale exposure unit dividing steps are as follows:
S21: by analyzing the space scale effects of the four scales, researching geometric form characteristics of the space scale effects and summarizing the difference and similarity rules of the space scale effects, the space association and constraint relation of exposure units among different scales are obtained;
s22: carrying out attribute feature analysis by adopting a correlation analysis method, exploring the rule of the road traffic exposure characterization along with the change of the scale, researching the conversion and change mechanism of the exposure elements, and obtaining attribute association and constraint relation of the exposure units among different scales;
s23: the geometry of the multiscale exposed cell is constructed. The geometry of the primary large scale exposed cells is obtained first. And then, based on the spatial association and constraint relation between the S21 primary exposure unit and the secondary scale, combining higher-precision data and water flow motion characteristics, and obtaining the geometric form of the secondary larger-scale exposure unit under a specific space-time range and organization mode. By a similar method, spatial information of other exposing units is acquired, and geometric construction of the exposing units from large scale to small scale is realized;
s24: and designing attribute characteristics of the multi-scale exposition unit according to the exposition simulation prediction requirements of different scales. And designing proper attribute characteristics for the exposure units of different scales by taking into consideration attribute association and constraint relation among the scales S22.
Further, the S3 utilizes a weather hydrographic coupling technology to respectively take weather forecast rainfall forecast data (Coupled Model Intercomparison Project Phase 6, CMIP6) and high-resolution short-time adjacent rainfall forecast data (The Integrated Nowcasting throughComprehensive Analysis, INCA) as rainfall input of a waterlogging model, and simulate and forecast urban road exposition under different scales. Specifically, the method comprises the following steps:
s31: obtaining and preprocessing CMIP6 climate estimated data;
s32: acquiring and processing INCA high-resolution short-time adjacent precipitation forecast data;
s33: and (5) predicting the multi-scale road traffic exposure based on the meteorological waterlogging coupling technology.
Further, the step S31 includes the following substeps:
s311: CMIP6 predicts data acquisition. According to the invention, the urban waterlogging model is driven by the recently released rainfall estimated data of the future situation in the CMIP6, so that the waterlogging condition of the target area of the 5 years in the future is estimated. The climate pattern used was a CNRM climate pattern developed by the french national weather research center, which provided lattice data as simulation data. By the method, the situation of future urban inland inundation can be simulated, and references are provided for related decisions and planning. The predicted time period is 2021-2026, and the meteorological elements include: daily precipitation, daily gas temperature maximum and daily gas Wen Junzhi;
S312: CMIP6 predicts the downscaling of the data. The invention downscales CMIP6 mode data to the resolution of observed data as follows:
(1) Interpolating the average value of the observation data in each month of years to the mode resolution to obtain the deviation of the observation field and the simulation field;
(2) The resolution ratio of the deviation field data is kept consistent with that of the original observed data through interpolation, and the deviation field data and the original observed data are summed;
(3) When the resolution ratio of the observed data is higher than 0.25 DEG, the CMIP6 data can not accurately reflect the spatial heterogeneity of urban rainfall forecast, and the climate estimated grid point data is interpolated into the values；
S313: and (5) extremely precipitating water. The invention researches the extreme of strong rainfall which causes waterlogging, and selects 80 percentiles as the threshold value of the extreme rainfall. By analyzing the total extreme precipitation of 2021-2026, samples with estimated daily precipitation of more than or equal to 0.2mm are extracted, and the samples are arranged according to the precipitation amount from large to small. The extreme precipitation threshold is the daily precipitation at the 80 th percentile, and precipitation exceeding the threshold is identified as an extreme precipitation event.
Further, the step S32 includes the following substeps:
s321: INCA data acquisition. INCA is developed by the Austrian weather bureau, and is a short-time proximity prediction system with the time resolution of 1 hour and the spatial resolution of 0.01 degrees;
S322: and forecasting quality assessment by INCA data. In order to evaluate the quality of INCA rainfall forecast products, the invention utilizes the rainfall observation data of 2016-2018 of Meteorological departments in Nanjing city to evaluate the forecast accuracy of the short-time proximity forecast system in the embodiment region.
Further, the step S33 adopts the urban waterlogging disaster exposure multiscale expression partitioning criteria in step S2, including the following four spatio-temporal scales:
s331: under the future climate warming background, predicting road traffic exposure of only urban expressways after 5 years of extreme rainfall by adopting CMIP6 climate estimated data in S31, wherein the time scale is year, and the space scale is a primary large-scale exposure unit;
s332: under the future climate warming background, predicting road traffic exposition which only comprises urban expressways and trunk roads in the future under the extreme rainfall of 1 year by adopting the CMIP6 climate estimated data in S31, wherein the time scale is a month, and the space scale is a secondary larger scale exposition unit;
s333: within the near aging of 0-24 hours in the future, predicting the road traffic exposition including urban expressways, arterial roads and secondary arterial roads by adopting short-time near precipitation forecast (INCA) data with high resolution in S32, wherein the time scale is hour, and the space scale is a three-level mesoscale exposition unit;
S334: and in the short time adjacent aging of 0-3h in the future, predicting the road traffic exposure comprising the urban expressway, the arterial road, the secondary arterial road and the branches by adopting short time adjacent precipitation forecast (INCA) data with high resolution in S32, wherein the time scale is minute, and the space scale is a four-level small-scale exposure unit.
Further, the step S4 includes the following substeps:
s41: preprocessing track data;
s42: selecting an exposure computing index;
s43: characteristic index quantization treatment;
s44: identifying the direct exposure characteristic of road traffic based on a gray wolf optimization-support vector machine;
s45: and calculating the indirect exposure characteristic weight of the road traffic by using the gradient lifting tree.
Further, the step S41 includes the following substeps:
s411: and (5) converting a coordinate system. The invention relates to a coordinate system conversion method, which is used for converting longitude and latitude coordinates of a track point from a hundred-degree coordinate system to a Mars coordinate system, and converting the coordinate system into a world geodetic coordinate system 1984 unified with road network data through coordinate inverse solution, so that the unification and the accuracy of the data are realized;
s412: and (5) spatial filtering. The invention needs to delete the track point data outside the target area, and in order to remove redundant data, the invention needs to delete the track point data outside the target area;
S413: and removing the trace abnormal points. In the process of sampling track data, partial abnormal points may appear, and need to be removed in data preprocessing. These outliers can be categorized into two categories:
multi-point problem for bicycle. It means that at the same point in time, a vehicle leaves a repeated track point. This may be because the sensor sends multiple signals in a short time, resulting in a net jockey track point number that far exceeds 1 at many points in time;
vehicle parking problems. It is characterized by that the speed and displacement of the vehicle are not zero, but the displacement value is very small, so that it can be considered that the vehicle has not undergone actual movement.
S414: map matching. In order to study the space-time law of road traffic under the condition of urban inland inundation and further analyze the exposition of urban roads, the invention carries out data matching on the GPS track data of vehicles and an electronic map. In order to associate the track points with the target area road network, a corresponding map matching process is required.
Further, the step S414 specifically includes the following substeps:
step 1: setting the threshold value of the azimuth angle difference value to be 25 degrees;
step 2: searching all roads by taking each track point as a center and taking 30m as a searching radius;
step 3: making a vertical line from the track point to the matched road, and taking the foot drop as the position of the matched track point;
Step 4: and (3) repeating the steps 2-3 until all the track point data are matched with the map.
Further, the step S42 selects the road as the disaster-bearing body, and performs the exposure research by analyzing the characteristics related to the exposure, and the specific characteristic indexes according to the step S113 include: road intermediate center line, road elevation, land use type, road class, number of lanes, road surface material, road length, and road speed.
Further, the step S43 includes the following substeps:
s431: and (5) value range scaling processing. In order to unify standards, the invention performs value domain scaling on the original data of various numerical type indexes. A common value range scaling method is to scale with the highest value to ensure that each index value is between 0 and 1. The invention adopts a range normalization method to carry out value range scaling treatment on the running speed, the intermediation centrality, the DEM, the road length and the lane number of the road. After the processing, the value ranges of all the characteristics are between 0 and 1;
s432: one-hot Encoding (One Encoding) is a commonly used data Encoding technique for converting classification variables into a format that can be processed by machine learning algorithms. Because the attribute fields such as pavement materials, road types and the like are not numerical values, the invention converts the category characteristic values of the areas based on the independent-heat coding method so as to avoid the problem of directly using integer digital conversion. The idea of one-hot encoding is to encode M states using M state vectors, where only one position is 1 at each point in time and the other positions are 0.
Further, the step S44 of identifying the water accumulation road section in the area through the machine learning model specifically includes the following substeps:
s441: and (3) constructing a ponding road classification recognition model based on a gray wolf optimization-support vector machine (Grey Wolf Optimizer-Support Vector Machine, GWO-SVM). Since road water accumulation is closely related to the existence of direct exposure, and the degree of water accumulation is mainly reflected by road speed. In order to accurately and efficiently evaluate the direct exposure of the road, the invention trains a model for specially classifying and identifying the ponding road. Therefore, a support vector machine is used as a core algorithm of the model, and a gray-wolf optimization algorithm (GWO) is used for searching the optimal parameters of the Support Vector Machine (SVM), and when GWO is used for optimizing the SVM parameters, the parameters of the SVM are generally used as an objective function of the optimization problem. These parameters may include penalty factor C, parameters of the kernel function, and other super parameters. GWO algorithm determines the optimal parameter value by the position and fitness of the individual wolf;
s442: training a ponding road classification model based on a support vector machine, wherein the method mainly comprises the following steps of:
step 1: and extracting features from the preprocessed data, and constructing feature vectors. According to the multidimensional feature index introduced in step S22, a plurality of features including road surface material, land use type, road length, etc. are included. Combining the features into a feature vector, wherein the dimension of the feature vector is 13;
Step 2: a portion of the data is selected as training sample input and a Support Vector Machine (SVM) is searched for appropriate parameters using a wolf optimization algorithm (GWO): firstly, initializing the position and fitness of a wolf group, determining Alpha, beta and Delta wolves in the group according to the fitness, respectively representing the current optimal solution, the suboptimal solution and the suboptimal solution, updating the position of the wolves, and simulating the cooperative behavior of the wolf group. And the positions of other wolves are adjusted according to the positions of Alpha, beta and Delta wolves. And then calculating the fitness of the updated wolf individuals, and updating Alpha, beta and Delta wolves according to the fitness. The above steps are repeated until a stop condition is reached (e.g., a maximum number of iterations is reached). Finally, obtaining optimized SVM kernel function type and regularization parameters according to the final Alpha gray wolf position；
Step 3: determining a weight coefficient of a training sample to construct a classification recognition model, and in a support vector machine, adjusting the quantity difference between a positive example and a negative example by setting the weight of the sample, wherein when the proportion value of the positive example and the negative example is greater than 4, the reciprocal of the proportion value is used as the weight to be given to the class of the positive example sample; when the proportion value of the positive sample and the negative sample is smaller than 0.25, the reciprocal of the proportion value is used as weight to be assigned to the class of the counterexample sample;
S443: and (5) evaluating a model. After the training of the model is completed, the accuracy of the training model is evaluated by using two indexes of the precision and the recall. For the typical classification problem of the ponding road classification recognition, the method divides the prediction obtaining conditions of a GWO-SVM ponding classification model into four types: correct identification (TP), incorrect identification (FP), missing identification (TN), and incomplete identification (FN). Precision ratio ofAnd recall->Respectively defined as:
，
，
further, the step S45 includes the following substeps:
s451: and (5) hierarchical expression of the characteristics. The invention divides the indirect exposure computing system into a target layer, a criterion layer and an index layer;
s452: and (5) indirectly exposing characteristic weight analysis. To better assess the extent of impact of different features on variable variation, the invention scores all features based on a gradient-lifted tree (Gradient Boosting Tree) machine learning method.
Gradient boosting trees are an integrated learning method by combining multiple weak learners (typically decision trees) to build a more powerful model. The method is a serial method that progressively improves the predictive performance of the model by iteratively training a decision tree to minimize the loss function.
Step 1: data set partitioning. For the gradient lifting tree, constructing a plurality of training subsets by adopting replaced random sampling;
step 2: training a gradient lifting tree model. The model is first initialized to a simple weak learner. In the iterative training process, residuals (differences between predicted values and true values) are calculated according to the predicted results and the actual labels of the current model, and then these residuals are used to train a new decision tree, which is called a "residual tree". Combining the new residual tree with the previous model to obtain an updated model. Finally, continuously adding new residual trees through repeated iteration until stopping conditions are met;
step 3: an importance metric score for the indirect exposure feature is calculated. I.e. a weighted average of the information gain or splitting criterion of the indirect exposition features when constructing the decision tree is calculated. Finally, the features are ranked according to the importance measurement scores of the features, and the features with higher scores have larger correlation with the change of the strain.
Further, the step S5 includes the following substeps:
s51: constructing a road traffic exposure intelligent body model;
s52: making a host behavior rule and an object behavior rule of the vehicle intelligent agent;
S53: dynamic simulation of urban road traffic exposure;
s54: road traffic exposure at multiple scales is predicted based on the agent.
Further, the S51 agent model includes three parts, namely a disaster recovery environment subsystem, a disaster-bearing body subsystem and a risk assessment subsystem:
s511: and constructing a disaster-tolerant environment subsystem. Through the disaster-tolerant environment subsystem, an environment causing disaster is constructed, and an activity space is provided for an intelligent body. The model abstracts the real urban environment into a virtual space-time, and deduces disaster variation trend and influence in the same environment by simulating a storm waterlogging and road traffic system. Through the association of time and space, the model describes the disaster process, reveals the influence on urban traffic, and simulates the space-time change of road traffic state;
s512: and constructing a disaster-bearing body subsystem. Urban road traffic and vehicles are the main components of road traffic exposure. Through feature engineering and weight analysis, the traffic exposure of the urban traffic road can be quantitatively calculated, and the weighting calculation is carried out by using a formula. In addition, the number of vehicle passes may indirectly reflect road traffic exposure;
s513: and (5) constructing a risk analysis subsystem. The road and vehicle risks are the basis of a risk analysis subsystem, and the storm waterlogging is associated with the disaster area, so that the road traffic exposure degree is calculated quantitatively. The system has the following functions: and calculating the exposure degree of the road and analyzing the road with high exposure degree. And (3) traffic optimization is carried out according to the result, and the road with high exposure is closed, so that loss is avoided.
Further, the S52 vehicle agent can make decisions and react to flood disasters and can actively complete behavioral activities. The method specifically comprises the following substeps:
s521: and (5) designing behavior rules of the main body of the running vehicle. The design of the vehicle behavior rules is critical to the urban traffic system. The vehicle is used as an intelligent agent, and the time-space coordinates of the vehicle are critical to the calculation of traffic exposure. In storms and inland inundations, vehicles need to perceive road changes and take appropriate strategies. The spatial position difference is caused by different vehicles responding to disasters and different daily behavior modes. The invention thus uses a probabilistic finite state machine (Probabilistic Finite State Machine, PFSM) to describe possible states and activities of the vehicle. The PFSM consists of a set of states, transition probabilities, and output probabilities. In this case, the state may represent a possible behavioral state or a response state of the agent. Transition probabilities represent the probability of transitioning from one state to another and can be determined based on the awareness, perception, and decision mechanisms of the agent to the waterlogging disaster. The output probabilities represent the probabilities of producing a particular behavior or activity in each state. In a storm flood disaster, each agent evaluates its own position, if there is a risk of blockage, it needs to stop to avoid flameout. If part of the road is accumulated but not jammed, the road is driven according to the original plan; if the road condition cannot pass, re-planning the route;
S522: and (5) designing a behavior rule of the object of the driving vehicle. Because the running of the vehicle is limited by the geographical environment, the speeds of roads with different grades are different, and the speed of a ponding road is reduced, the behavior rule of the vehicle and the object is required to be designed. According to the method, the behavior rules of the vehicle and the object are indirectly defined through the behavior rules of the road intelligent body, each road in the road set is traversed firstly, the passing speed of the road is set according to the number of lanes and the road type information, and finally the passing state at the moment is set according to the road grade and the risk, so that the design of the behavior rules of the vehicle and the object is completed, and the road intelligent body can rapidly cope with the influence of urban waterlogging.
Further, the step S53 includes the following substeps:
s531: and (5) dynamically calculating the road traffic exposure. Traffic exposure dynamic calculation refers to the calculation and evaluation of the exposure of road traffic under different conditions. In the case of stormwater and waterlogging, the exposition of road traffic is affected directly and indirectly. Directly impact environmental and road conditions from disasters, including rainfall and waterlogging. The indirect effect is a propagation effect caused by traffic anomalies, mainly traffic congestion spreading, resulting in increased road traffic exposure. The road traffic dynamic calculation utilizes an intelligent system to acquire the current environment state and queries a vehicle information database to acquire the vehicle position. By the mechanism, the affected condition of road traffic in each area can be counted, and the change of the road traffic exposure is reflected;
S532: exposure dynamically simulates a scene design. The exposure dynamic simulation scene is mainly divided into two scenes, namely a normal traffic scene and a flood disaster scene:
normal traffic scenario. According to the behavior rule and the feature sharing result of the intelligent agent, the invention designs a normal traffic scene;
flood disaster scenarios. The scene mainly simulates the dynamic change condition of road traffic exposure when urban storm water-logging disasters occur.
Further, the step S54 takes a future rainstorm and waterlogging situation as input, and predicts urban road traffic exposure under multiple scales based on the intelligent agent.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention designs a quantitative calculation model for the direct and indirect exposure of road traffic, and considers multiple factors such as rainfall, ponding, roads, topography and the like. And a ponding road classification recognition model is put forward and trained, so that the direct exposure quantification is realized. The quantitative calculation of the indirect exposure of the road traffic is realized by calculating the weight parameters;
2. aiming at the defect of single view angle of the current urban waterlogging traffic exposure research, the invention regards urban waterlogging as a natural disaster of natural-society composite type, considers hydrologic process and human activity, and analyzes factors influencing road traffic exposure from multiple dimensions;
3. Aiming at the defect that the input of the current waterlogging model mainly depends on rainfall data of a ground rainfall station, the invention develops multi-scale urban road traffic exposure prediction based on the waterlogging scene simulation based on a meteorological hydrographic coupling technology;
4. aiming at the defect of insufficient consideration of dynamic simulation of traffic exposure under the constraint of the current geographic environment, an intelligent behavior rule and an exposure dynamic simulation method are designed, and multi-scale road traffic exposure prediction based on intelligent agents is realized.
Drawings
The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention. In the drawings:
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph showing exposure variation trend of a part of road traffic under normal conditions according to an embodiment of the present invention;
fig. 3 shows exposure variation trend of partial road traffic in flood disaster situations according to the embodiment of the invention.
Detailed Description
For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.
The main urban area of Nanjing, jiangsu province was selected as an example area, and the average annual precipitation days of the area was 117 days and the average precipitation amount was 1200 mm according to data statistics. The main flood season is 6 to 10 months each year, thunderstorm and severe convection weather occur frequently, and strong precipitation is easily caused by typhoons, so that extreme weather conditions such as heavy rain, heavy rain or local heavy rain are caused. The data of the examples are shown in Table 1:
table 1 example data table
As shown in fig. 1, the method for predicting the traffic exposure of the multi-scale urban waterlogged roads based on the intelligent agent disclosed by the embodiment of the invention mainly comprises the following steps:
s1: road exposure influence characteristic analysis under waterlogging situations;
s2: dividing a road traffic space multi-scale exposure unit;
s3: a multi-scale urban road traffic exposure prediction coupled with future climate data;
s4: carrying out quantitative calculation on the road traffic exposure of rainstorm and waterlogging;
s5: urban road traffic exposure in future climate is predicted based on the agent.
Further, the step S1 includes the following substeps:
s11: analyzing road traffic exposure influence factors;
s12: analyzing the road traffic exposure forming process;
S13: dominant and recessive impact feature recognition.
Further, the S11 urban waterlogging disaster is mainly generated by the comprehensive action of three parts, and is specifically as follows:
s111: and analyzing disaster causing factors. Disaster causing factors are direct causes of disasters, and refer to dynamic factors which can cause disasters and subsequent socioeconomic losses. Heavy rain, which is a fast-occurring and widely existing inducing factor, has an important influence on the traffic network, resulting in an increase in travel delay, a decrease in road vehicle capacity, and an imbalance in travel mode. The impact of different rainfall intensities on urban traffic is shown in table 2:
table 2 influence of rainfall intensity on road traffic;
s112: and analyzing the pregnant disaster environment. The regional differences determine the spatial-temporal distribution characteristics of the disaster-causing factors. The pregnant disaster environment refers to natural and artificial environments for the pregnant urban waterlogging disasters. Natural environments include earth surface topography and the like, while artificial environments include urban drainage capacity, urban land function area distribution and the like;
s113: and analyzing the disaster bearing body. The disaster-bearing body is an urban road and is characterized by comprising the following components:
road speed. Since the road speed is mainly detected by a sensor such as a loop detector, it is difficult to acquire the value. In order to obtain road speed data at different time points, the invention utilizes vehicle GPS data to match vehicle track points with corresponding roads by using a map matching method. The average speed of the road is the average speed of all track points passing through the road section in the time unit, and the calculation formula is as follows:
，
Wherein,the number of target road track points per unit time, +.>For the track point on the road section +.>Instantaneous speed;
road elevation. In order to correlate raw elevation (Digital Elevation Model, DEM) data with urban road data, the invention assigns grid-type elevation data to the corresponding vector roads;
road intermediation centrality. The intermediation centrality is used to quantify the importance of nodes in the network graph, itReflecting the number of shortest paths through the node in the graph;
land use type. Land use types are classified into park, residential and commercial uses. Typically, commercial land traffic congestion is more frequent;
road surface material. There is a certain relationship between the road surface material and the road traffic exposure. Road surface materials refer to construction materials for road surfaces, such as asphalt road surfaces, cement road surfaces, etc. The manner in which different pavement materials affect traffic exposure varies. The roughness of the road surface material may affect the running safety and traffic exposure of the vehicle. The rough pavement material can increase the friction between the vehicle and the pavement, improve the stability and the grip of the vehicle and reduce the traffic exposure. The smooth road surface material can cause the vehicle to easily slip on a rainy day or a wet road surface, so that the risk of traffic accidents is increased, and the traffic exposure is further improved;
Road grade. Roads are classified into expressways, urban main roads, secondary main roads and branch roads according to grades. The road division criteria of different grades are shown in Table 3:
table 3 road class division criteria table;
number of lanes. The number of lanes refers to the number of lanes on a road that a vehicle can travel. Since the road class is not directly related to the number of lanes, it is necessary to study the road traffic exposure by means of the feature of the number of lanes. Normally, a road with a small number of lanesThe higher the traffic exposure, the more susceptible is to the problem of water accumulation caused by rainfall;
road length. The length of a link refers to the distance between two end points and is generally used to measure the buffering capacity provided by the link when it is congested. The shorter the road length, the lower the congestion buffering capacity of the road, and the higher the probability of traffic congestion.
Further, the step S12 includes the following substeps:
s121: and detecting rainfall abnormal roads. The rainfall abnormal road means that after rainy days, the speed change of part of vehicles on the road causes the creep or congestion of the surrounding road. In order to detect rainfall abnormal roads, the embodiment performs statistical comparison on two types of road speed data in a normal scene of the area and in a rainstorm waterlogging scene;
S122: road extraction is affected by rainfall. According to the statistical result, the road types are divided into three types: a road with a significantly rising speed, a road with no significant change in speed, and a road with a significantly falling speed. In order to analyze more accurately, the invention combines the rainfall monitoring point data of the embodiment area to carry out statistical analysis and calculate the average change rate of the road speed. The proportions of different types of roads in a certain section of area in the embodiment are shown in table 4:
table 4 different category road proportion statistics;
s123: and (5) extracting the rainfall abnormal road. Road speed time series data during rainfall is analyzed using a Mann-Kendall abrupt change detection method to detect a speed change situation. According to the method, all roads in the embodiment area are analyzed, the roads with speed mutation are counted, and the positions are represented through a map;
s124: road traffic congestion spread analysis. In urban traffic, sporadic road congestion usually initially occurs at a specific point, such as an intersection, and only directly affects roads associated with the space. In order to analyze the influence of an abnormal road caused by initial heavy rain on the surrounding road conditions, the invention adopts a local Morgan index to analyze the spatial autocorrelation of the abnormal road caused by the heavy rain. The spatial autocorrelation analysis method can be used for researching the spatial distribution characteristics of the geographic elements, namely analyzing the potential correlation existing between the geographic elements in the same research area, and further analyzing the road traffic jam diffusion phenomenon caused by rainfall and waterlogging. The local molan index is calculated as follows:
，
In the method, in the process of the invention,is a variable->Is the variable mean value->Is the number of geographic elements->Is a weight matrix.
Further, the step S13 includes the following substeps:
s131: dominant impact feature recognition. The dominant impact features refer to the dominant impact features of the occurrence of puddles and the like on the road when the exposure is calculated. Puddles may cause traffic anomalies, resulting in traffic jams. Sometimes, the blocking situation is propagated through the adjacent roads, so that the exposure degree of the adjacent roads is increased;
s132: implicit impact feature recognition. Implicit impact characteristics refer to the potential impact on the exposure of a road when calculating the exposure. The present invention calculates the expositivity by quantifying the indirect expositivity features.
Further, the spatial multi-scale effect of the S2 urban waterlogging disaster generates obvious multi-scale characteristics on the exposure distribution pattern and change mechanism of road traffic. The invention designs four road traffic space scale exposure units, which specifically comprise: primary large scale exposition units, i.e. only including urban expressways; a second-level larger exposure unit, namely only comprising urban expressways and main roads; three-level mesoscale exposure units, namely, urban expressways, arterial roads and secondary arterial roads; four-level small-scale exposition units, namely, urban expressways, main roads, secondary main roads and branches. Specifically, the road traffic space multi-scale exposure unit dividing steps are as follows:
S21: by analyzing the space scale effects of the four scales, researching geometric form characteristics of the space scale effects and summarizing the difference and similarity rules of the space scale effects, the space association and constraint relation of exposure units among different scales are obtained;
s22: carrying out attribute feature analysis by adopting a correlation analysis method, exploring the rule of the road traffic exposure characterization along with the change of the scale, researching the conversion and change mechanism of the exposure elements, and obtaining attribute association and constraint relation of the exposure units among different scales;
s23: the geometry of the multiscale exposed cell is constructed. The geometry of the primary large scale exposed cells is obtained first. And then, based on the spatial association and constraint relation between the S21 primary exposure unit and the secondary scale, combining higher-precision data and water flow motion characteristics, and obtaining the geometric form of the secondary larger-scale exposure unit under a specific space-time range and organization mode. By a similar method, spatial information of other exposing units is acquired, and geometric construction of the exposing units from large scale to small scale is realized;
s24: and designing attribute characteristics of the multi-scale exposition unit according to the exposition simulation prediction requirements of different scales. And designing proper attribute characteristics for the exposure units of different scales by taking into consideration attribute association and constraint relation among the scales S22.
Further, the S3 utilizes a weather hydrographic coupling technology to respectively take weather forecast rainfall forecast data (Coupled Model Intercomparison Project Phase 6, CMIP6) and high-resolution short-time adjacent rainfall forecast data (The Integrated Nowcasting throughComprehensive Analysis, INCA) as rainfall input of a waterlogging model, and simulate and forecast urban road exposition under different scales. Specifically, the method comprises the following steps:
s31: obtaining and preprocessing CMIP6 climate estimated data;
s32: acquiring and processing INCA high-resolution short-time adjacent precipitation forecast data;
s33: and (5) predicting the multi-scale road traffic exposure based on the meteorological waterlogging coupling technology.
Further, the step S31 includes the following substeps:
s311: CMIP6 predicts data acquisition. According to the invention, the urban waterlogging model is driven by the recently released rainfall estimated data of the future situation in the CMIP6, so that the waterlogging condition of the target area of the 5 years in the future is estimated. The climate pattern used was a CNRM climate pattern developed by the french national weather research center, which provided lattice data as simulation data. By the method, the situation of future urban inland inundation can be simulated, and references are provided for related decisions and planning. The predicted time period is 2021-2026, and the meteorological elements include: daily precipitation, daily gas temperature maximum and daily gas Wen Junzhi;
S312: CMIP6 predicts the downscaling of the data. The invention downscales CMIP6 mode data to the resolution of observed data as follows:
(1) Interpolating the average value of the observation data in each month of years to the mode resolution to obtain the deviation of the observation field and the simulation field;
(2) The resolution ratio of the deviation field data is kept consistent with that of the original observed data through interpolation, and the deviation field data and the original observed data are summed;
(3) When the resolution ratio of the observed data is higher than 0.25 DEG, the CMIP6 data can not accurately reflect the spatial heterogeneity of urban rainfall forecast, and the climate estimated grid point data is interpolated into the values；
S313: and (5) extremely precipitating water. The invention researches the extreme of strong rainfall which causes waterlogging, and selects 80 percentiles as the threshold value of the extreme rainfall. By analyzing the total extreme precipitation of 2021-2026, samples with estimated daily precipitation of more than or equal to 0.2mm are extracted, and the samples are arranged according to the precipitation amount from large to small. The extreme precipitation threshold is the daily precipitation at the 80 th percentile, and precipitation exceeding the threshold is identified as an extreme precipitation event.
Further, the step S32 includes the following substeps:
s321: INCA data acquisition. INCA is developed by the Austrian weather bureau, and is a short-time proximity prediction system with the time resolution of 1 hour and the spatial resolution of 0.01 degrees;
S322: and forecasting quality assessment by INCA data. In order to evaluate the quality of INCA rainfall forecast products, the invention utilizes the rainfall observation data of 2016-2018 of Meteorological departments in Nanjing city to evaluate the forecast accuracy of the short-time proximity forecast system in the embodiment region.
Further, the step S33 adopts the urban waterlogging disaster exposure multiscale expression partitioning criteria in step S2, including the following four spatio-temporal scales:
s331: under the future climate warming background, predicting road traffic exposure of only urban expressways after 5 years of extreme rainfall by adopting CMIP6 climate estimated data in S31, wherein the time scale is year, and the space scale is a primary large-scale exposure unit;
s332: under the future climate warming background, predicting road traffic exposition which only comprises urban expressways and trunk roads in the future under the extreme rainfall of 1 year by adopting the CMIP6 climate estimated data in S31, wherein the time scale is a month, and the space scale is a secondary larger scale exposition unit;
s333: within the near aging of 0-24 hours in the future, predicting the road traffic exposition including urban expressways, arterial roads and secondary arterial roads by adopting short-time near precipitation forecast (INCA) data with high resolution in S32, wherein the time scale is hour, and the space scale is a three-level mesoscale exposition unit;
S334: and in the short time adjacent aging of 0-3h in the future, predicting the road traffic exposure comprising the urban expressway, the arterial road, the secondary arterial road and the branches by adopting short time adjacent precipitation forecast (INCA) data with high resolution in S32, wherein the time scale is minute, and the space scale is a four-level small-scale exposure unit.
Further, the step S4 includes the following substeps:
s41: preprocessing track data;
s42: selecting an exposure computing index;
s43: characteristic index quantization treatment;
s44: identifying the direct exposure characteristic of road traffic based on a gray wolf optimization-support vector machine;
s45: and calculating the indirect exposure characteristic weight of the road traffic by using the gradient lifting tree.
Further, the step S41 includes the following substeps:
s411: and (5) converting a coordinate system. The invention relates to a coordinate system conversion method, which is used for converting longitude and latitude coordinates of a track point from a hundred-degree coordinate system to a Mars coordinate system, and converting the coordinate system into a world geodetic coordinate system 1984 unified with road network data through coordinate inverse solution, so that the unification and the accuracy of the data are realized;
s412: and (5) spatial filtering. The method needs to delete the data of the track points outside the embodiment region, and the spatial positions of the data are distributed in Chuzhou, zhenjiang and other cities around Nanjing. In order to remove redundant data, the invention needs to delete the track point data outside the Nanjing urban area;
S413: and removing the trace abnormal points. In the process of sampling track data, partial abnormal points may appear, and need to be removed in data preprocessing. These outliers can be categorized into two categories:
multi-point problem for bicycle. It means that at the same point in time, a vehicle leaves a repeated track point. This may be because the sensor sends multiple signals in a short time, resulting in a net jockey track point number that far exceeds 1 at many points in time;
vehicle parking problems. It is characterized by that the speed and displacement of the vehicle are not zero, but the displacement value is very small, so that it can be considered that the vehicle has not undergone actual movement.
S414: map matching. In order to study the space-time law of road traffic under the condition of urban inland inundation and further analyze the exposition of urban roads, the invention carries out data matching on the GPS track data of vehicles and an electronic map. In order to associate the track points with the embodiment road network, a corresponding map matching process is required.
Further, the step S414 specifically includes the following substeps:
step 1: setting the threshold value of the azimuth angle difference value to be 25 degrees;
step 2: searching all roads by taking each track point as a center and taking 30m as a searching radius;
step 3: making a vertical line from the track point to the matched road, and taking the foot drop as the position of the matched track point;
Step 4: and (3) repeating the steps 2-3 until all the track point data are matched with the map.
Further, the step S42 selects the road as the disaster-bearing body, and performs the exposure research by analyzing the characteristics related to the exposure, and the specific characteristic indexes according to the step S113 include: road intermediary center line, road elevation, land use type, road class, number of lanes, road surface material, road length, and road speed;
further, the step S43 includes the following substeps:
s431: and (5) value range scaling processing. In order to unify standards, the invention performs value domain scaling on the original data of various numerical type indexes. A common value range scaling method is to scale with the highest value to ensure that each index value is between 0 and 1. The invention adopts a range normalization method to carry out value range scaling treatment on the running speed, the intermediation centrality, the DEM, the road length and the lane number of the road. After the treatment, the values of all the characteristics are in the range of 0 to 1, and are shown in the following table 5:
table 5 value range scaling results;
s432: one-hot Encoding (One Encoding) is a commonly used data Encoding technique for converting classification variables into a format that can be processed by machine learning algorithms. Because the attribute fields such as pavement materials, road types and the like are not numerical values, the invention converts the category characteristic values of the areas based on the independent-heat coding method so as to avoid the problem of directly using integer digital conversion. The idea of one-hot encoding is to encode M states using M state vectors, where only one position is 1 at each point in time and the other positions are 0. The invention adopts a single-heat coding technology, taking an example of an embodiment road grade, 4 roads with different grades are represented by 4-bit binary codes, 1 on each bit corresponds to different road grade names respectively, and the obtained codes are classified features, namely 0001, 0010, 0100 and 1000.
Further, the step S44 of identifying the water accumulation road section in the area through the machine learning model specifically includes the following substeps:
s441: and (3) constructing a ponding road classification recognition model based on a gray wolf optimization-support vector machine (Grey Wolf Optimizer-Support Vector Machine, GWO-SVM). Since road water accumulation is closely related to the existence of direct exposure, and the degree of water accumulation is mainly reflected by road speed. In order to accurately and efficiently evaluate the direct exposure of the road, the invention trains a model for specially classifying and identifying the ponding road. Therefore, a support vector machine is used as a core algorithm of the model, and a gray-wolf optimization algorithm (GWO) is used for searching the optimal parameters of the Support Vector Machine (SVM), and when GWO is used for optimizing the SVM parameters, the parameters of the SVM are generally used as an objective function of the optimization problem. These parameters may include penalty factor C, parameters of the kernel function, and other super parameters. GWO algorithm determines the optimal parameter value by the position and fitness of the individual wolf;
s442: training a ponding road classification model based on a support vector machine, wherein the method mainly comprises the following steps of:
step 1: and extracting features from the preprocessed data, and constructing feature vectors. According to the multidimensional feature index introduced in step S22, a plurality of features including road surface material, land use type, road length, etc. are included. Combining the features into a feature vector, wherein the dimension of the feature vector is 13;
Step 2: selecting a portion of dataAs training sample inputs, a gray wolf optimization algorithm (GWO) is used to search for the appropriate parameters of a Support Vector Machine (SVM): firstly, initializing the position and fitness of a wolf group, determining Alpha, beta and Delta wolves in the group according to the fitness, respectively representing the current optimal solution, the suboptimal solution and the suboptimal solution, updating the position of the wolves, and simulating the cooperative behavior of the wolf group. And the positions of other wolves are adjusted according to the positions of Alpha, beta and Delta wolves. And then calculating the fitness of the updated wolf individuals, and updating Alpha, beta and Delta wolves according to the fitness. The above steps are repeated until a stop condition is reached (e.g., a maximum number of iterations is reached). Finally, obtaining optimized SVM kernel function type and regularization parameters according to the final Alpha gray wolf position；
Step 3: determining a weight coefficient of a training sample to construct a classification recognition model, and in a support vector machine, adjusting the quantity difference between a positive example and a negative example by setting the weight of the sample, wherein when the proportion value of the positive example and the negative example is greater than 4, the reciprocal of the proportion value is used as the weight to be given to the class of the positive example sample; when the proportion value of the positive sample and the negative sample is smaller than 0.25, the reciprocal of the proportion value is used as weight to be assigned to the class of the counterexample sample;
S443: and (5) evaluating a model. After the training of the model is completed, the accuracy of the training model is evaluated by using two indexes of the precision and the recall. For the typical classification problem of the ponding road classification recognition, the method divides the prediction obtaining conditions of a GWO-SVM ponding classification model into four types: correct identification (TP), incorrect identification (FP), missing identification (TN), and incomplete identification (FN). Precision ratio ofAnd recall->Respectively defined as:
，
，
in order to verify the accuracy of the model in classifying and identifying the ponding roads, all data sets in the time period of 20:00-22:00 of 2017, 8 months and 1 day in the embodiment area are selected, 70% of the data are selected as training sets, and 30% of the data are selected as test sets. The model evaluation after training is shown in table 6:
table 6 training model evaluation;
further, the step S45 includes the following substeps:
s451: and (5) hierarchical expression of the characteristics. The invention divides the indirect exposure computing system into a target layer, a criterion layer and an index layer. The specific hierarchy is shown in table 7 below:
table 7 hierarchical expression of exposure features;
s452: and (5) indirectly exposing characteristic weight analysis. To better assess the extent of impact of different features on variable variation, the invention scores all features based on a gradient-lifted tree (Gradient Boosting Tree) machine learning method.
Gradient boosting trees are an integrated learning method by combining multiple weak learners (typically decision trees) to build a more powerful model. The method is a serial method that progressively improves the predictive performance of the model by iteratively training a decision tree to minimize the loss function.
Step 1: data set partitioning. For the gradient lifting tree, constructing a plurality of training subsets by adopting replaced random sampling;
step 2: training a gradient lifting tree model. The model is first initialized to a simple weak learner (e.g., a shallow decision tree). In the iterative training process, residuals (differences between predicted values and true values) are calculated according to the predicted results and the actual labels of the current model, and then these residuals are used to train a new decision tree, which is called a "residual tree". Combining the new residual tree with the previous model to obtain an updated model. Finally, continuously adding new residual trees through multiple iterations until stopping conditions (such as reaching the preset number of trees or reaching a certain performance index) are met;
step 3: an importance metric score for the indirect exposure feature is calculated. I.e. a weighted average of the information gain or splitting criterion of the indirect exposition features when constructing the decision tree is calculated. Finally, the features are ranked according to the importance measurement scores of the features, and the features with higher scores have larger correlation with the change of the strain. In this embodiment, the weight calculation results of the indirect exposure feature are shown in table 8:
Table 8 indirect exposure characteristic weight table;
further, the step S5 includes the following substeps:
s51: constructing a road traffic exposure intelligent body model;
s52: making a host behavior rule and an object behavior rule of the vehicle intelligent agent;
s53: dynamic simulation of urban road traffic exposure;
s54: road traffic exposure at multiple scales is predicted based on the agent.
Further, the S51 agent model includes three parts, namely a disaster recovery environment subsystem, a disaster-bearing body subsystem and a risk assessment subsystem:
s511: and constructing a disaster-tolerant environment subsystem. Through the disaster-tolerant environment subsystem, an environment causing disaster is constructed, and an activity space is provided for an intelligent body. The model abstracts the real urban environment into a virtual space-time, and deduces disaster variation trend and influence in the same environment by simulating a storm waterlogging and road traffic system. Through the association of time and space, the model describes the disaster process, reveals the influence on urban traffic, and simulates the space-time change of road traffic state;
s512: and constructing a disaster-bearing body subsystem. Urban road traffic and vehicles are the main components of road traffic exposure. Through feature engineering and weight analysis, the traffic exposure of the urban traffic road can be quantitatively calculated, and the weighting calculation is carried out by using a formula. In addition, the number of vehicle passes may indirectly reflect road traffic exposure;
S513: and (5) constructing a risk analysis subsystem. The road and vehicle risks are the basis of a risk analysis subsystem, and the storm waterlogging is associated with the disaster area, so that the road traffic exposure degree is calculated quantitatively. The system has the following functions: and calculating the exposure degree of the road and analyzing the road with high exposure degree. And (3) traffic optimization is carried out according to the result, and the road with high exposure is closed, so that loss is avoided.
Further, the S52 vehicle agent can make decisions and react to flood disasters and can actively complete behavioral activities. The method specifically comprises the following substeps:
s521: and (5) designing behavior rules of the main body of the running vehicle. The design of the vehicle behavior rules is critical to the urban traffic system. The vehicle is used as an intelligent agent, and the time-space coordinates of the vehicle are critical to the calculation of traffic exposure. In storms and inland inundations, vehicles need to perceive road changes and take appropriate strategies. The spatial position difference is caused by different vehicles responding to disasters and different daily behavior modes. The invention thus uses a probabilistic finite state machine (Probabilistic Finite State Machine, PFSM) to describe possible states and activities of the vehicle. The PFSM consists of a set of states, transition probabilities, and output probabilities. In this case, the state may represent a possible behavioral state or a response state of the agent. Transition probabilities represent the probability of transitioning from one state to another and can be determined based on the awareness, perception, and decision mechanisms of the agent to the waterlogging disaster. The output probabilities represent the probabilities of producing a particular behavior or activity in each state. In a storm flood disaster, each agent evaluates its own position, if there is a risk of blockage, it needs to stop to avoid flameout. If part of the road is accumulated but not jammed, the road is driven according to the original plan; if the road condition cannot pass, re-planning the route;
S522: and (5) designing a behavior rule of the object of the driving vehicle. Because the running of the vehicle is limited by the geographical environment, the speeds of roads with different grades are different, and the speed of a ponding road is reduced, the behavior rule of the vehicle and the object is required to be designed. According to the invention, the behavior rules of the vehicle and the object are indirectly defined through the behavior rules of the road intelligent agent, each road in the road set is traversed firstly, the traffic speed of the road is set according to the number of lanes and the road type information, and finally the traffic state at the moment is set according to the road grade and the risk, so that the behavior rule design of the running vehicle and the object is completed; so that the road intelligent agent can rapidly cope with the influence of urban inland inundation.
Further, the step S53 includes the following substeps:
s531: and (5) dynamically calculating the road traffic exposure. Traffic exposure dynamic calculation refers to the calculation and evaluation of the exposure of road traffic under different conditions. In the case of stormwater and waterlogging, the exposition of road traffic is affected directly and indirectly. Directly impact environmental and road conditions from disasters, including rainfall and waterlogging. The indirect effect is a propagation effect caused by traffic anomalies, mainly traffic congestion spreading, resulting in increased road traffic exposure. The road traffic dynamic calculation utilizes an intelligent system to acquire the current environment state and queries a vehicle information database to acquire the vehicle position. By the mechanism, the affected condition of road traffic in each area can be counted, and the change of the road traffic exposure is reflected;
S532: exposure dynamically simulates a scene design. The exposure dynamic simulation scene is mainly divided into two scenes, namely a normal traffic scene and a flood disaster scene:
normal traffic scenario. According to the behavior rules and feature sharing results of the intelligent agent, the invention designs a normal traffic scene.
In this embodiment, by combining road traffic data and agent rules, the 20:00 point is selected as the starting simulation time. Normal simulation of the driving situation of the vehicle on the road is performed in the period of 20:00 to 22:00 points. Taking a road section as an example, the change of the road traffic condition is analyzed, and the change trend of the road traffic exposure is plotted as fig. 2. It can be observed from the figure that under normal traffic scenarios, road traffic exposure remains at a low level at all times. This means that the road in this area remains relatively clear during the simulated time period;
flood disaster scenarios. The scene mainly simulates the dynamic change condition of road traffic exposure when urban storm water-logging disasters occur.
Examples based on the exposure calculation and the stormwater logging data, the stormwater onset time was set at 20:00 to 22:00 for a period of time, and the rainfall intensity increased rapidly, resulting in a gradual increase in road traffic exposure. Since there are many vehicles traveling at this time, the number of roads exposed to the risk also increases greatly at this stage. Taking a road section as an example, the dynamic change of the road traffic exposure is analyzed, and the change trend is plotted as fig. 3.
Further, the step S54 takes a future rainstorm and waterlogging situation as input, and predicts urban road traffic exposure under multiple scales based on the intelligent agent. Calculated, the partial road exposure at different scales for the examples is shown in table 9 below:
table 9 exposure of part of the road at different scales;
the above description of embodiments is only for aiding in the understanding of the method of the present invention and its core ideas; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (14)
1. The method for predicting the traffic exposure of the multi-scale urban waterlogged roads based on the intelligent agent is characterized by comprising the following steps of:
s1: road exposure influence characteristic analysis under waterlogging situations;
s2: dividing a road traffic space multi-scale exposure unit;
s3: a multi-scale urban road traffic exposure prediction coupled with future climate data;
s4: carrying out quantitative calculation on the road traffic exposure of rainstorm and waterlogging;
s5: predicting urban road traffic exposure in future climate based on the intelligent agent;
wherein step S1 comprises the sub-steps of:
S11: analyzing road traffic exposure influence factors;
s12: analyzing the road traffic exposure forming process;
s13: dominant and recessive impact feature recognition;
four road traffic space scale exposure units are designed in the step S2, and specifically include: primary large scale exposition units, i.e. only including urban expressways; a second-level larger exposure unit, namely only comprising urban expressways and main roads; three-level mesoscale exposure units, namely, urban expressways, arterial roads and secondary arterial roads; a four-level small-scale exposition unit, namely, an urban expressway, a main road, a secondary trunk road and a branch road; the method comprises the following specific steps:
s21: by analyzing the space scale effects of the four scales, researching geometric form characteristics of the space scale effects and summarizing the difference and similarity rules of the space scale effects, the space association and constraint relation of exposure units among different scales are obtained;
s22: carrying out attribute feature analysis by adopting a correlation analysis method, exploring the rule of the road traffic exposure characterization along with the change of the scale, researching the conversion and change mechanism of the exposure elements, and obtaining attribute association and constraint relation of the exposure units among different scales;
s23: constructing the geometric form of a multi-scale exposure unit, firstly obtaining the geometric form of a primary large-scale exposure unit, then, based on the spatial association and constraint relation between the primary exposure unit and a secondary scale of S21, combining higher-precision data and water flow motion characteristics to obtain the geometric form of a secondary large-scale exposure unit, and sequentially acquiring the spatial information of other exposure units to realize the geometric form construction of the exposure units from large scale to small scale;
S24: according to the exposure simulation prediction requirements of different scales, designing attribute characteristics of the multi-scale exposure units, and designing suitable attribute characteristics for the different-scale exposure units by considering attribute association and constraint relations among the S22 scales;
wherein, the step S3 is specifically as follows,
step S3, using a weather hydrographic coupling technology, respectively taking weather pre-estimated precipitation prediction data Coupled Model Intercomparison Project Phase, CMIP6 and high-resolution short-time adjacent precipitation prediction data The Integrated Nowcasting through Comprehensive Analysis, and INCA as rainfall inputs of a waterlogging model, and simulating and predicting urban road exposition under different scales, wherein the method comprises the following steps:
s31: obtaining and preprocessing CMIP6 climate estimated data;
s32: acquiring and processing INCA high-resolution short-time adjacent precipitation forecast data;
s33: multiscale road traffic exposure prediction based on meteorological waterlogging coupling technology;
wherein step S4 comprises the sub-steps of:
s41: preprocessing track data;
s42: selecting an exposure computing index;
s43: characteristic index quantization treatment;
s44: identifying the direct exposure characteristic of road traffic based on a gray wolf optimization-support vector machine;
S45: calculating the indirect exposure characteristic weight of the road traffic by using the gradient lifting tree;
wherein, the step S5 comprises the following substeps:
s51: constructing a road traffic exposure intelligent body model;
s52: making a host behavior rule and an object behavior rule of the vehicle intelligent agent;
s53: dynamic simulation of urban road traffic exposure;
s54: road traffic exposure at multiple scales is predicted based on the agent.
2. The method for predicting the urban waterlogging road traffic exposure based on the intelligent agent according to claim 1, wherein the urban waterlogging disaster in the step S11 is mainly generated by three parts of comprehensive actions, specifically:
s111: disaster causing factor analysis, wherein the disaster causing factor is a direct cause of disasters and refers to dynamic factors possibly causing disasters and subsequent socioeconomic losses;
s112: analysis of disaster-pregnant environment, wherein regional differences determine the space-time distribution characteristics of disaster-causing factors, and the disaster-pregnant environment refers to natural and artificial environments for breeding urban waterlogging disasters; the natural environment comprises ground surface landforms, and the artificial environment comprises urban drainage capacity and urban land function area distribution;
s113: and analyzing the disaster bearing body.
3. The agent-based multi-scale urban inland inundation road traffic exposure prediction method according to claim 1, wherein the step S12 exposure forming process analysis comprises the following sub-steps:
S121: detecting a rainfall abnormal road, namely after rainy days, slowly moving or blocking surrounding roads caused by the speed change of part of vehicles on the road, and carrying out statistical comparison on two types of road speed data in a normal scene and a rainstorm waterlogging scene of a target area in order to detect the rainfall abnormal road;
s122: road extraction affected by rain fall, road types are divided into three categories: the method comprises the steps of carrying out statistical analysis by combining rainfall monitoring point data of a target area and calculating the average change rate of the speed of a road, wherein the speed of the road is obviously increased, the speed of the road is not obviously changed, and the speed of the road is obviously reduced;
s123: extracting rainfall abnormal roads, analyzing road speed time sequence data during rainfall by using a Mann-Kendall mutation detection method to detect the speed change condition, analyzing all roads in a target area, counting the roads with speed mutation and representing the positions through a map;
s124: and (3) analyzing road traffic jam diffusion, wherein a local Morgan index is adopted to analyze the spatial autocorrelation of a storm abnormal road, and the calculation formula of the local Morgan index is as follows:
in the method, in the process of the invention,is a variable->Is the variable mean value->Is the number of geographic elements- >Is a weight matrix.
4. The agent-based multi-scale urban inland inundation road traffic exposure prediction method according to claim 1, wherein the step S13 of influencing feature recognition comprises the sub-steps of:
s131: dominant impact feature identification, namely when the exposure degree is calculated, the dominant impact feature of a puddle appears on a road, and the puddle causes traffic abnormality, so that traffic jam is caused;
s132: and identifying implicit influence features, wherein the implicit influence features have potential influence on the exposure degree of the road when the exposure degree is calculated.
5. The agent-based multi-scale urban inland inundation road traffic exposure prediction method according to claim 1, wherein step S31 comprises the following sub-steps:
s311: the method comprises the steps that the CMIP6 estimated data is obtained, a city waterlogging model is driven by precipitation estimated data of a newly released future situation in the CMIP6 to estimate waterlogging conditions of a target area of 5 years in the future, a climate mode is a CNRM climate mode developed by a French national weather research center, grid point data is provided as simulation data, the situation of the future city waterlogging is simulated by the method, and references are provided for related decisions and planning, and weather elements comprise: daily precipitation, daily gas temperature maximum and daily gas Wen Junzhi;
S312: the CMIP6 estimated data downscaling process comprises the following specific steps:
(1) Interpolating the average value of the observation data in each month of years to the mode resolution to obtain the deviation of the observation field and the simulation field;
(2) The resolution ratio of the deviation field data is kept consistent with that of the original observed data through interpolation, and the deviation field data and the original observed data are summed;
(3) When the resolution ratio of the observed data is higher than 0.25 DEG, the CMIP6 data can not accurately reflect the spatial heterogeneity of urban rainfall forecast, and the climate estimated grid point data is interpolated into the values；
S313: and calculating extreme precipitation, namely, selecting an 80 percentile as a threshold value of the extreme precipitation, extracting samples with estimated daily precipitation amount not less than 0.2mm, arranging the samples according to the precipitation amount from large to small, wherein the extreme precipitation threshold value is the daily precipitation amount at the 80 th percentile, and the precipitation exceeding the threshold value is identified as an extreme precipitation event.
6. The agent-based multi-scale urban inland inundation road traffic exposure prediction method according to claim 1, wherein step S32 comprises the following sub-steps:
s321: INCA data acquisition, INCA is developed by the Austrian weather bureau, and is a short-time proximity prediction system with the time resolution of 1 hour and the spatial resolution of 0.01 degrees;
S322: and (3) forecasting quality assessment of INCA data, namely, assessing forecasting precision of the short-time proximity forecasting system in the embodiment area by utilizing rainfall observation data of a meteorological department in order to assess quality of INCA rainfall forecasting products.
7. The agent-based multi-scale urban inland inundation road traffic exposure prediction method according to claim 1, wherein the step S33 multi-scale expression partitioning criteria comprises the following four spatio-temporal scales:
s331: under the future climate warming background, predicting road traffic exposure of only urban expressways after 5 years of extreme rainfall by adopting CMIP6 climate estimated data in S31, wherein the time scale is year, and the space scale is a primary large-scale exposure unit;
s332: under the future climate warming background, predicting road traffic exposition which only comprises urban expressways and trunk roads in the future under the extreme rainfall of 1 year by adopting the CMIP6 climate estimated data in S31, wherein the time scale is a month, and the space scale is a secondary larger scale exposition unit;
s333: within the near aging of 0-24 hours in the future, predicting the road traffic exposition including urban expressways, main roads and secondary roads by adopting short-time near precipitation forecast INCA data with high resolution in S32, wherein the time scale is hour, and the space scale is a three-level mesoscale exposition unit;
S334: and in the short time adjacent aging of 0-3h in the future, predicting the road traffic exposition comprising urban expressways, main roads, secondary trunk roads and branches by adopting short time adjacent precipitation forecast INCA data with high resolution in S32, wherein the time scale is minute, and the space scale is a four-level small-scale exposition unit.
8. The method for predicting the traffic exposure of the multi-scale urban inland inundation road based on the intelligent agent according to claim 1, wherein,
step S41 comprises the following sub-steps:
s411: coordinate system conversion, namely converting longitude and latitude coordinates of the track points from a hundred-degree coordinate system to a Mars coordinate system, and converting the coordinate system into a world geodetic coordinate system 1984 unified with road network data through coordinate inverse solution;
s412: spatially filtering, namely deleting the track point data outside the target area;
s413: trace outlier removal, wherein the S413 outlier is classified into the following two categories:
(1) A single car multipoint problem, which refers to the fact that at the same point in time, one car leaves a repeated track point, probably because the sensor sends multiple signals in a short time, resulting in a net car track point number far exceeding 1 at many points in time;
(2) The vehicle parking problem, which is represented by the fact that the vehicle speed and displacement are not zero, but the displacement value is very small, the vehicle can be considered to have no actual movement,
S414: map matching comprising the sub-steps of:
step 1: setting the threshold value of the azimuth angle difference value as；
Step 2: searching all roads by taking each track point as a center and taking 30m as a searching radius;
step 3: making a vertical line from the track point to the matched road, and taking the foot drop as the position of the matched track point;
step 4: repeating the steps 2-3 until all the track point data are matched with the map,
wherein the S42 selects the road as a disaster-tolerant body, and the exposure research is performed by analyzing the characteristics related to the exposure.
9. The method for predicting the traffic exposure of the multi-scale urban inland inundation road based on the intelligent agent according to claim 1, wherein,
wherein S43 comprises the following sub-steps:
s431: the value range scaling treatment adopts a range standard method to carry out the value range scaling treatment on the running speed, the intermediation centrality, the DEM, the road length and the lane number of the road, namely, the value range of the characteristics is between 0 and 1;
s432: one-hot Encoding, the idea of which is to encode M states using M state vectors, where only One position is 1 at each point in time and the other positions are 0.
10. The method for predicting the traffic exposure of the multi-scale urban inland inundation road based on the intelligent agent according to claim 1, wherein,
Wherein, S44 identifies the ponding road section in the area through a machine learning model, and specifically comprises the following substeps:
s441: the method is characterized in that a ponding road classification recognition model is built based on a wolf optimization-support vector machine, GWO-SVM, because road ponding is closely related to direct exposure, the ponding degree is mainly reflected by road speed, the support vector machine is adopted as a core algorithm of the model, the optimal parameters of the support vector machine are searched by using the wolf optimization algorithm GWO, when parameters of the SVM are optimized by using GWO, the parameters of the SVM are taken as objective functions of optimization problems, and the parameters comprise penalty coefficientsThe parameters of the kernel function and other super parameters, and the GWO algorithm determines the optimal parameter value through the position and the adaptability of the gray wolf individuals;
s442: training a ponding road classification model based on a support vector machine, wherein the method mainly comprises the following steps of:
step 1: extracting features from the preprocessed data, constructing feature vectors, and combining the features into a feature vector according to the multidimensional feature indexes introduced in the step S22, wherein the multidimensional feature indexes comprise a plurality of features of pavement materials, land utilization types and road lengths, and the dimension of the feature vector is 13;
step 2: selecting part of the data as training sample input and searching the proper parameters of the support vector machine SVM by using a Grey wolf optimization algorithm GWO: firstly, initializing the position and fitness of a wolf group, and determining Alpha, beta and Delta wolves in the group according to the fitness, wherein the Alpha, beta and Delta wolves respectively represent a current optimal solution, a suboptimal solution and a suboptimal solution; updating the positions of the wolves, simulating the cooperative behavior of the wolves, and adjusting the positions of other wolves according to the positions of Alpha, beta and Delta wolves; Calculating the adaptability of the updated wolf individuals, and updating Alpha, beta and Delta wolves according to the adaptability; repeating the steps until a stopping condition is reached; finally, obtaining optimized SVM kernel function type and regularization parameters according to the final Alpha gray wolf position，
Step 3: determining a weight coefficient of a training sample to construct a classification recognition model, and in a support vector machine, adjusting the quantity difference between a positive example and a negative example by setting the weight of the sample, wherein when the proportion value of the positive example and the negative example is greater than 4, the reciprocal of the proportion value is used as the weight to be given to the class of the positive example sample; when the ratio value of the positive sample and the negative sample is smaller than 0.25, the reciprocal of the ratio value is used as the weight to be assigned to the class of the counterexample sample,
s443: after model evaluation and model training are completed, the accuracy of the training model is evaluated by using two indexes of precision and recall ratio, and for the typical classification problem of ponding road classification and identification, the prediction obtaining conditions of a GWO-SVM ponding classification model are divided into four types: correct TP, incorrect FP, missing TN and incomplete FN, and precision rateAnd recall->Respectively defined as:
，
；
step S45 comprises the following sub-steps:
s451: the characteristic layering expression divides an indirect exposure computing system into a target layer, a criterion layer and an index layer;
S452: the indirect exposure characteristic weight analysis is carried out, and importance scoring is carried out on all the characteristics based on a gradient lifting tree Gradient Boosting Tree machine learning method;
wherein, the S452 indirect exposure characteristic weight analysis comprises the following substeps:
step 1: dividing a data set, and constructing a plurality of training subsets by adopting replaced random sampling for a gradient lifting tree;
step 2: training a gradient lifting tree model, firstly initializing the model into a simple weak learner, calculating residual errors according to a prediction result of a current model and an actual label in each training in the iterative training process, and then training a new decision tree by using the residual errors; this new decision tree is called "residual tree", and the new residual tree is combined with the previous model to obtain an updated model; finally, continuously adding new residual trees through repeated iteration until stopping conditions are met;
step 3: and calculating an importance metric score of the indirect exposure feature, namely calculating a weighted average of information gain or splitting criteria of the indirect exposure feature when constructing a decision tree, and finally sorting the features according to the importance metric score of the feature, wherein the features with higher scores have larger correlation with strain change.
11. The method for predicting the traffic exposure of the multi-scale urban inland inundation road based on the intelligent agent according to claim 1, wherein,
step S51, an agent model comprises three parts of a disaster recovery environment subsystem, a disaster bearing body subsystem and a risk assessment subsystem:
s511: constructing a disaster-tolerant environment subsystem, namely firstly constructing an environment causing disaster, providing an active space for an intelligent body, abstracting a real urban environment into a virtual space-time, deducing disaster change trend and influence, revealing influence on urban traffic, and finally simulating space-time change of road traffic state to finish the construction of the subsystem;
s512: the disaster-bearing subsystem is built, namely the traffic exposure of the road is reflected by the traffic quantity and the running speed of the vehicle on the road within a set time range, so that the disaster-bearing subsystem is built;
s513: the risk analysis subsystem is built, firstly, roads and vehicle risks are set as the basis of the risk analysis subsystem, the rainstorm waterlogging is associated with the disaster area, the road traffic exposure degree is calculated in a quantification mode, and finally, the high-exposure-degree roads are screened and sealed.
12. The method for predicting the traffic exposure of the multi-scale urban inland inundation road based on the intelligent agent according to claim 1, wherein,
In step S52, the vehicle agent makes decisions and reactions to the flood disasters, and can actively complete the behavior activities, which specifically includes the following sub-steps:
s521: the method comprises the steps that a behavior rule design of a main body of a running vehicle is carried out, a probability finite state machine Probabilistic Finite State Machine is adopted, a PFSM (pulse frequency per minute) describes the state and the activity of the vehicle, the PFSM consists of a group of states, transition probabilities and output probabilities, the output probabilities represent the probability of generating the activity in each state, each intelligent body evaluates the position of the intelligent body in a storm flood disaster, and if part of roads are accumulated but not jammed, the intelligent body runs according to an original plan; if the road condition cannot pass, re-planning the route;
s522: the method comprises the steps of designing a driving vehicle object behavior rule, namely, designing the vehicle object behavior rule because the driving of a vehicle is limited by a geographic environment, indirectly defining the vehicle object behavior rule through the behavior rule of a road intelligent body, traversing each road in a road set, setting the passing speed of the road according to the number of lanes and the road type information, and finally setting the passing state at the moment according to the road grade and the risk to finish the driving vehicle object behavior rule design; and S54, taking a future rainstorm waterlogging situation as input, and predicting urban road traffic exposure under multiple scales based on the intelligent agent.
13. The method for predicting the exposure of urban waterlogging road traffic based on intelligent agents according to claim 1, wherein the dynamic simulation of the exposure of S53 comprises the following substeps:
s531: the dynamic calculation of the road traffic exposure, which means the calculation and evaluation of the exposure of the road traffic under different conditions, and the exposure of the road traffic is directly and indirectly influenced under the condition of rainstorm and waterlogging, and the direct influence is from disaster environments and road conditions, including rainfall and waterlogging; the indirect influence is a propagation effect caused by traffic abnormality, mainly traffic congestion is diffused, the road traffic exposure is increased, the road traffic dynamic calculation utilizes an intelligent system to acquire the current environment state, a vehicle information database is queried to acquire the vehicle position, and the mechanism is used for counting the affected condition of the road traffic in each area and reflecting the change of the road traffic exposure;
s532: the exposition dynamic simulation scene design comprises the following two types:
(1) A normal traffic scene is designed according to the behavior rules and feature sharing results of the intelligent agent;
(2) The flood disaster scene mainly simulates the dynamic change condition of road traffic exposure when urban storm and waterlogging disasters occur.
14. The method for predicting the traffic exposure of the multi-scale urban waterlogging road based on the intelligent agent according to claim 2, wherein the S113 disaster-bearing body is an urban road, and the method is characterized by comprising the following steps:
(1) Road speed, in order to obtain road speed data of different time points, vehicle GPS data is utilized to match vehicle track points with corresponding roads by using a map matching method, the average speed of the road is the average speed of all track points passing through the road section in the time unit, and the calculation formula is as follows:
，
wherein,is a target of unit timeRoad track number, +.>For the track point on the road section +.>Instantaneous speed;
(2) Road elevation, in order to correlate the original elevation Digital Elevation Model, DEM data with urban road data, elevation data of grid type is assigned to the corresponding vector road;
(3) Road intermediacy, which quantifies the importance of a node in a network graph, reflects the number of shortest paths through the node in the graph;
(4) Land use types, which are classified into park, residential and commercial uses;
(5) The road surface materials have a relation with road traffic exposure, different road surface materials have different modes for influencing the traffic exposure, and the roughness of the road surface materials can influence the running safety and the traffic exposure of the vehicle;
(6) Road grades, wherein the roads are divided into expressways, urban main roads, secondary main roads and branch roads according to the grades;
(7) The number of lanes refers to the number of lanes on a road for vehicles to travel, and the road class and the number of lanes are not directly related, so that the road traffic exposure needs to be studied by means of the characteristic of the number of lanes;
(8) The length of a link refers to the distance between two end points and is used to measure the buffering capacity of the link in congestion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311637750.5A CN117332909B (en) | 2023-12-01 | 2023-12-01 | Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311637750.5A CN117332909B (en) | 2023-12-01 | 2023-12-01 | Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117332909A true CN117332909A (en) | 2024-01-02 |
| CN117332909B CN117332909B (en) | 2024-03-08 |
Family
ID=89277855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311637750.5A Active CN117332909B (en) | 2023-12-01 | 2023-12-01 | Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117332909B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117854285A (en) * | 2024-03-07 | 2024-04-09 | 南京邮电大学 | Storm water road section identification method considering urban hydrology and traffic flow characteristics |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010050300A1 (en) * | 2008-10-30 | 2010-05-06 | 株式会社ブリヂストン | Method of estimating road surface condition |
| WO2015148887A1 (en) * | 2014-03-28 | 2015-10-01 | Northeastern University | System for multivariate climate change forecasting with uncertainty quantification |
| WO2020215117A1 (en) * | 2019-04-24 | 2020-10-29 | South Australian Water Corporation | Method and system for detecting a structural anomaly in a pipeline network |
| CN111915158A (en) * | 2020-07-15 | 2020-11-10 | 云南电网有限责任公司带电作业分公司 | Rainstorm disaster weather risk assessment method, device and equipment based on Flood Area model |
| CN112733337A (en) * | 2020-12-28 | 2021-04-30 | 华南理工大学 | Method for evaluating urban road traffic efficiency under influence of rainstorm and waterlogging |
| CN113177737A (en) * | 2021-05-26 | 2021-07-27 | 南京恩瑞特实业有限公司 | Urban rainstorm disaster risk assessment method and system based on GA (genetic algorithm) optimization BP (back propagation) neural network |
| CN113191582A (en) * | 2021-03-15 | 2021-07-30 | 西南石油大学 | Road torrential flood vulnerability evaluation method based on GIS and machine learning |
| CN113450027A (en) * | 2021-08-30 | 2021-09-28 | 南京师范大学 | Dynamic exposure quantification method and device for urban inland inundation disasters |
| CN113610437A (en) * | 2021-08-24 | 2021-11-05 | 南京信息工程大学 | Disaster-bearing body dynamic exposure degree evaluation method and system |
| CN114372625A (en) * | 2021-12-30 | 2022-04-19 | 华南理工大学 | Urban waterlogging rapid forecasting method based on multi-output machine learning algorithm |
| CN114372685A (en) * | 2021-12-28 | 2022-04-19 | 长江生态环保集团有限公司 | Urban rainstorm waterlogging risk assessment method based on SWMM model |
| CN115619213A (en) * | 2022-09-27 | 2023-01-17 | 中国地质大学（武汉） | Highway traffic rainfall meteorological disaster risk assessment method, device and equipment |
| CN115936490A (en) * | 2022-11-23 | 2023-04-07 | 华南师范大学 | SHAP-based urban rainstorm waterlogging influence factor quantitative analysis method |
-
2023
- 2023-12-01 CN CN202311637750.5A patent/CN117332909B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010050300A1 (en) * | 2008-10-30 | 2010-05-06 | 株式会社ブリヂストン | Method of estimating road surface condition |
| WO2015148887A1 (en) * | 2014-03-28 | 2015-10-01 | Northeastern University | System for multivariate climate change forecasting with uncertainty quantification |
| WO2020215117A1 (en) * | 2019-04-24 | 2020-10-29 | South Australian Water Corporation | Method and system for detecting a structural anomaly in a pipeline network |
| CN111915158A (en) * | 2020-07-15 | 2020-11-10 | 云南电网有限责任公司带电作业分公司 | Rainstorm disaster weather risk assessment method, device and equipment based on Flood Area model |
| CN112733337A (en) * | 2020-12-28 | 2021-04-30 | 华南理工大学 | Method for evaluating urban road traffic efficiency under influence of rainstorm and waterlogging |
| CN113191582A (en) * | 2021-03-15 | 2021-07-30 | 西南石油大学 | Road torrential flood vulnerability evaluation method based on GIS and machine learning |
| CN113177737A (en) * | 2021-05-26 | 2021-07-27 | 南京恩瑞特实业有限公司 | Urban rainstorm disaster risk assessment method and system based on GA (genetic algorithm) optimization BP (back propagation) neural network |
| CN113610437A (en) * | 2021-08-24 | 2021-11-05 | 南京信息工程大学 | Disaster-bearing body dynamic exposure degree evaluation method and system |
| CN113450027A (en) * | 2021-08-30 | 2021-09-28 | 南京师范大学 | Dynamic exposure quantification method and device for urban inland inundation disasters |
| CN114372685A (en) * | 2021-12-28 | 2022-04-19 | 长江生态环保集团有限公司 | Urban rainstorm waterlogging risk assessment method based on SWMM model |
| CN114372625A (en) * | 2021-12-30 | 2022-04-19 | 华南理工大学 | Urban waterlogging rapid forecasting method based on multi-output machine learning algorithm |
| CN115619213A (en) * | 2022-09-27 | 2023-01-17 | 中国地质大学（武汉） | Highway traffic rainfall meteorological disaster risk assessment method, device and equipment |
| CN115936490A (en) * | 2022-11-23 | 2023-04-07 | 华南师范大学 | SHAP-based urban rainstorm waterlogging influence factor quantitative analysis method |
Non-Patent Citations (3)
| Title |
|---|
| 朱净萱等: "基于多智能体的城市洪涝灾害动态脆弱性计算模型构建", 《地球信息科学学报》, vol. 23, no. 10, pages 1787 - 1797 * |
| 朱雪虹: "顾及居民出行的城市洪涝灾害暴露性计算方法研究", 《中国优秀博硕士学位论文全文数据库（硕士）工程科技Ⅱ辑》, no. 05, pages 1 - 107 * |
| 江游: "基于轨迹数据的城市暴雨内涝道路交通暴露性时空模拟研究", 《中国优秀博硕士学位论文全文数据库（硕士）工程科技Ⅱ辑》, no. 03, pages 1 - 95 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117854285A (en) * | 2024-03-07 | 2024-04-09 | 南京邮电大学 | Storm water road section identification method considering urban hydrology and traffic flow characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117332909B (en) | 2024-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tang et al. | Flood susceptibility assessment based on a novel random Naïve Bayes method: A comparison between different factor discretization methods | |
| CN110570651B (en) | Road network traffic situation prediction method and system based on deep learning | |
| CN101354757B (en) | Method for predicting dynamic risk and vulnerability under fine dimension | |
| CN109448361B (en) | Resident traffic travel flow prediction system and prediction method thereof | |
| CN111665575B (en) | Medium-and-long-term rainfall grading coupling forecasting method and system based on statistical power | |
| CN117332909B (en) | Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent | |
| CN101673369A (en) | Projection pursuit-based method for evaluating flooding risk of drainage pipe network | |
| Chen et al. | Probabilistic forecasting of drought: a hidden Markov model aggregated with the RCP 8.5 precipitation projection | |
| Valverde et al. | Neural network and fuzzy logic statistical downscaling of atmospheric circulation-type specific weather pattern for rainfall forecasting | |
| Deng et al. | Risk assessment and prediction of rainstorm and flood disaster based on henan province, China | |
| CN113836808A (en) | PM2.5 deep learning prediction method based on heavy pollution feature constraint | |
| Lin et al. | Applications of cluster analysis and pattern recognition for typhoon hourly rainfall forecast | |
| CN117556197B (en) | Typhoon vortex initialization method based on artificial intelligence | |
| CN113779113B (en) | Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation | |
| CN115129802A (en) | Population spatialization method based on multi-source data and ensemble learning | |
| CN117010726B (en) | Intelligent early warning method and system for urban flood control | |
| Yuan et al. | A novel method based on deep learning model for national-scale landslide hazard assessment | |
| CN117494034A (en) | Air quality prediction method based on traffic congestion index and multi-source data fusion | |
| CN116110210A (en) | Data-driven landslide hazard auxiliary decision-making method in complex environment | |
| CN115600047A (en) | Small watershed surface average rainfall measurement and calculation method and system based on grid analysis | |
| Pandey et al. | Hybrid deep learning model for flood frequency assessment and flood forecasting | |
| Huynh et al. | An optimal rain-gauge network using a GIS-based approach with spatial interpolation techniques for the Mekong River Basin | |
| Zhou et al. | Real‐time prediction and ponding process early warning method at urban flood points based on different deep learning methods | |
| Halim et al. | Flash flood prediction in selangor using data mining techniques | |
| Zhang et al. | Urban waterlogging prediction and risk analysis based on rainfall time series features: A case study of Shenzhen |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |