Highway congestion prediction and active prevention and control method based on big data

Through big data analysis and model prediction, the accuracy of highway congestion prediction and prevention and control is solved, accurate prediction and early prevention and control of highway congestion conditions are achieved, and the prediction and prevention and control effect is improved.

CN119068670APending Publication Date: 2024-12-03内蒙古交通集团有限公司 +4
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411067309.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology cannot effectively predict and prevent and control highway congestion, and lacks scientific data support, resulting in a lack of accuracy in the guidance orders and a narrow application range, so the prediction and prevention and control effect needs to be improved.

Method used

Using a big data-based method, non-parametric detection and wavelet density estimation are obtained by acquiring real-time highway data, non-parametric distribution model of travel time is established, and short-time interval prediction is carried out by combining the Bootstrap-KNN model, and traffic is analyzed using fuzzy C-means clustering. Operating status, output congestion conditions, and proactive prevention and control are carried out according to meteorological, control, on-site and emergency conditions.

Benefits of technology

It has achieved accurate prediction and early prevention and control of highway congestion conditions, improved the accuracy and effectiveness of prediction and prevention, enriched the connotation of traffic system evaluation, and helped managers and travelers make better decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068670A_ABST
    Figure CN119068670A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of expressway congestion prediction, and particularly relates to a big data-based expressway congestion prediction and active prevention and control method, which comprises the following steps of: firstly, acquiring expressway data; carrying out data quality analysis and data feature analysis on the expressway data distribution characteristics by adopting a non-parametric detection method; establishing a highway travel time nonparametric distribution model by adopting a wavelet density estimation method, and estimating the highway travel time; and then, a Bootstrap-KNN-based highway travel time short-time interval prediction model is adopted to complete highway travel time short-time interval prediction. Then, traffic operation state evaluation is carried out by taking two characteristics of parameter weight and sample capacity difference into consideration by adopting statistical data of travel time and utilizing a fuzzy C-means clustering quantitative analysis method; the congestion condition in the corresponding area range is output; and finally, according to the output congestion condition in the corresponding area range, in combination with objective conditions, carrying out active prevention and control on the road section which is about to be congested.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of highway congestion prediction, and in particular relates to a highway congestion prediction and active prevention and control method based on big data. Background Art

[0002] With the rapid development of the economy, the passenger and freight volumes of highways have also increased year by year. On the one hand, the growth rate of vehicle ownership and vehicle usage frequency far exceeds the rate of increase in road supply, which highlights the contradiction of imbalance between supply and demand; on the other hand, the spatial saturation of the highway network during peak traffic hours continues to rise, so that any small emergency will affect the normal operation of the entire road; these reasons lead to frequent congestion on highways. At present, the congestion on highways is handled through manual intervention: traffic police use their personal work experience to deal with congestion on highways. Manual guidance is performed on the road sections, and at the same time, the command to restrict vehicles from entering the road network is conveyed to the upstream vehicle entrance points to alleviate the congestion of the entire road; the manual intervention method has some disadvantages, for example, it can only guide the road sections where congestion has occurred, and it cannot predict and avoid the occurrence of congestion in advance; the lack of scientific data support makes the entry control command lack of accuracy. In the prior art, congestion prediction is performed through monitoring data on the highway. For example, the Chinese patent with patent number: CN111243279A discloses a method for predicting congestion at a highway toll station. Specifically, a road card is first obtained. The starting point A and the end point B of the center line of the entrance are connected into a line, and the preset distance is extended to both sides as the width to form a rectangular area, which is the electronic fence area E of the toll station; the first electronic fence and the second electronic fence are respectively established on the upper and lower sides of the electronic fence area E; the time when the vehicle n enters the electronic fence area E, the first electronic fence and the second electronic fence for the first time is recorded; through the operator's dedicated line, it is connected with the national key operating vehicle networking and joint control platform to obtain the longitude and latitude coordinates of the vehicle passing through the electronic fence area E, the first electronic fence and the second electronic fence in real time; the time when the vehicle enters the first electronic fence, the second electronic fence and the electronic fence area E for the first time is analyzed by the time difference between entering each electronic fence to obtain the congestion index; the congestion index information is uploaded to the required platform. The above-mentioned prior art only predicts the traffic jam situation within a small area of ​​a toll station by obtaining the time difference between the vehicle entering the two electronic fences, and it also needs to connect to the national key operating vehicle networking and joint control platform, which makes the application scope of the above-mentioned invention narrow, and the prediction and prevention effect needs to be improved. For this reason, we propose a highway congestion prediction and active prevention and control method based on big data. Summary of the invention

[0003] The purpose of the present invention is to provide a highway congestion prediction and active prevention and control method based on big data.

[0004] The technical solution adopted by the present invention is as follows:

[0005] A highway congestion prediction and active prevention and control method based on big data, comprising the following steps:

[0006] Step 1: Obtain real-time highway data;

[0007] Step 2: Use non-parametric detection methods to conduct data quality analysis and data feature analysis on the distribution characteristics of highway data;

[0008] Step 3: Use wavelet density estimation method to establish a non-parametric distribution model of highway travel time and estimate highway travel time;

[0009] Step 4: Use the short-term prediction model of highway travel time based on Bootstrap-KNN to complete the short-term prediction of highway travel time;

[0010] Step 5: Use the fuzzy C-means clustering quantitative analysis method based on the travel time statistics to evaluate the traffic operation status by considering the two characteristics of parameter weight and sample capacity difference; output the congestion status within the corresponding area;

[0011] Step 6: Based on the output of the congestion situation in the corresponding area, combined with the objective conditions of meteorology, control, on-site, and emergency response capabilities, proactively prevent and control the road sections that are about to be congested.

[0012] Preferably, in step 1, the highway data acquired includes vehicle data, road condition information, data, traffic flow information and environmental information data.

[0013] Preferably, in step 2, the asymmetric highway travel time data is preprocessed by combining the quartile method with statistical principles, and non-parametric tests are used to test the distribution characteristics of the data in the time, space and vehicle model dimensions.

[0014] Preferably, in step 3, a wavelet density estimation method is used to establish a non-parametric distribution model of highway travel time to estimate the highway travel time.

[0015] Preferably, the specific steps of step 4 are as follows:

[0016] Step 401: Construct an original sample data set X. Take the highway toll system as the data source, calculate the starting and ending travel time of each historical record, and obtain the original sample X = {x1, x2, ···, x n}, n∈N+;

[0017] Step 402: Bootstrap resampling: extract m samples X from the original sample X with replacement * ={x * 1,x * 2,···,x * m}, m∈N+, usually m=n, this sample is called Bootstrap sample; establish B travel time Bootstrap samples X1 * , X2 * ,···,X B * ;

[0018] Step 403: Build a point prediction model; for each Bootstrap sample x * Perform travel time prediction modeling to obtain t i =f i (x)+ε,i=1,2,···,B;

[0019] Step 404: Calculate the prediction value sequence: Use the B point prediction models in the previous step to predict the test samples respectively, and obtain a sequence θ containing B prediction values * ={θ * 1, θ * 2,…,θ * B};

[0020] Step 405: Bootstrap interval prediction; calculate the sequence θ of B predicted values * ={θ * 1, θ * 2, ..., θ * B}, that is, the travel time prediction interval [low(t), up(t)].

[0021] From the modeling steps, it can be seen that the point prediction model in step 403 and the Bootstrap confidence interval estimation method in step 405 will directly affect the performance and quality of travel time interval prediction, so these two key steps need to be discussed and analyzed in depth.

[0022] Preferably, in step 405, the Bootstrap confidence interval estimation includes an interval estimation of a standard deviation, an interval estimation of a percentile, and an interval estimation of an accelerated bias correction.

[0023] Preferably, in step 5, a sample observation matrix is ​​first given:

[0024]

[0025] Among them, {γ1, γ2,..., γ n} is an n observation sample with p variables, {y i1 ,y i2 , ..., y ip} represents the variable observation value corresponding to the i-th sample, and the sample data {Y1, Y2, ..., Y n} is divided into c categories, n≥c≥2, and C={C1,Y2,...,C c} are c cluster centers, and Define the algorithm objective function:

[0026]

[0027] In the formula, U=(u ik ) c×n ; is the membership matrix, u ik Indicates the membership of the kth sample to the i-th category, 0≤u ik ≤1, d ij =||y j -c i || represents the i-th cluster center C i With the jth sample point y j The distance between them, m is the weighted index, which represents the fuzziness of the membership matrix, m∈[1+∞], and the value of m is proportional to the fuzziness of the membership matrix.

[0028] Preferably, in formula (2), since the Euclidean distance is used as the similarity metric function, the distance function is expressed as:

[0029]

[0030] in, is the weight of the kth traffic parameter in the FCM algorithm, and p is the number of traffic parameters.

[0031] Preferably, a Lagrangian function is constructed for formula (2), as follows:

[0032]

[0033] Among them, λ j (j=1, 2, ...n) are n Lagrange multipliers. The necessary condition for using Lagrange multiplication to solve the minimum value is:

[0034]

[0035] in, According to the two necessary conditions of formula (5) and formula (6), the cluster center C and the membership matrix U are determined by iteration.

[0036] Preferably, for formula (4), ci and ui are respectively j Find the partial derivative and set it to zero, then we have;

[0037]

[0038]

[0039] because Then we have:

[0040]

[0041] Transforming the above formula, we have:

[0042]

[0043] Combining formula (7) and formula (10), we can get:

[0044]

[0045] Through iterative calculation of the iterative algorithm, the cluster center ci and membership degree ui can be obtained j ; By comparing the distance between the real-time data samples and the center of each state class, the traffic operation status is determined based on the principle of the shortest distance.

[0046] The technical effects achieved by the present invention are:

[0047] The present invention first obtains historical highway traffic data, then uses a non-parametric detection method to perform data quality analysis and data feature analysis on the distribution characteristics of highway data; obtains specific highway travel time from the obtained historical highway traffic data; establishes a non-parametric distribution model of highway travel time by using a wavelet density estimation method, and uses the above model to estimate the distribution of highway travel time, and finally completes the training of the non-parametric distribution model of highway travel time through the historical highway traffic data; then continues to use the historical highway traffic data to adopt a short-term prediction model of highway travel time based on Bootstrap-KNN to complete the short-term prediction of highway travel time, specifically The travel time distribution output by the non-parametric distribution model of highway travel time is used as the input of the Bootstrap-KNN highway travel time short-time interval prediction model, and the predicted output travel time interval is completed. The predicted travel time interval is compared with the travel time in the real historical data of the highway, and the Bootstrap-KNN highway travel time short-time interval prediction model is corrected. Finally, the Bootstrap-KNN highway travel time short-time interval prediction model with more accurate output results is obtained; the training of the non-parametric distribution model of highway travel time and the Bootstrap-KNN highway travel time short-time interval prediction model is completed. ; Then obtain the real-time data on the highway, and similarly use the non-parametric detection method to conduct data quality analysis and data feature analysis on the distribution characteristics of highway data, and then use the non-parametric distribution model of highway travel time to complete the estimation and distribution of highway travel time, and use the Bootstrap-KNN highway travel time short-term interval prediction model to predict the travel time in a specific highway interval, and get the prediction interval. Finally, use the fuzzy C-means clustering quantitative analysis method based on the statistical data of travel time to consider the two characteristics of parameter weight and sample capacity difference to evaluate the traffic operation status; output the congestion status within the corresponding area; the congestion status includes completely unblocked and reliability, basic smooth reliability, slight congestion reliability and severe congestion reliability; finally, according to the congestion situation in the corresponding area of ​​the output, combined with the objective conditions of meteorology, control, on-site and emergency handling capabilities, the road sections that will be congested are actively prevented and controlled to avoid road congestion in advance. In summary, when extracting traffic state parameters, the present invention carries out research on wavelet density and kernel density non-parametric estimation of travel time reliability, short-term travel time prediction based on Boostrap-KNN and fuzzy C-means clustering to distinguish traffic states; in the present invention, the concept of reliability is introduced into short-term interval prediction, so that travel time reliability becomes an important measure for evaluating the operation efficiency of the transportation network and the quality of travel services;At the same time, incorporating reliability indicators into the estimation and analysis of highway traffic operation status can enrich the connotation of traffic system evaluation, help managers and travelers make better decisions, and achieve proactive prevention and control of road sections that are about to be congested. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of a highway congestion prediction and active prevention and control method based on big data of the present invention;

[0049] Figure 2 It is a flow chart of constructing a non-parametric distribution model of highway travel time according to the present invention;

[0050] Figure 3 It is a flowchart of constructing a short-term prediction model of highway travel time based on Bootstrap-KNN of the present invention;

[0051] Figure 4 It is a system block diagram of a highway congestion prediction and active prevention and control method based on big data of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is specifically described below in conjunction with embodiments. It should be understood that the following text is only used to describe one or several specific embodiments of the present invention, and does not strictly limit the scope of protection of the specific claims of the present invention.

[0053] like Figure 1 As shown, a highway congestion prediction and active prevention and control method based on big data includes the following steps:

[0054] Step 1: Obtain real-time highway data;

[0055] Step 2: Use non-parametric detection methods to conduct data quality analysis and data feature analysis on the distribution characteristics of highway data;

[0056] Step 3: Use wavelet density estimation method to establish a non-parametric distribution model of highway travel time and estimate highway travel time.

[0057] Step 4: Use the Bootstrap-KNN-based highway travel time short-term interval prediction model to complete the highway travel time short-term interval prediction.

[0058] Step 5: Use the fuzzy C-means clustering quantitative analysis method based on the travel time statistics to evaluate the traffic operation status by considering the two characteristics of parameter weight and sample capacity difference; output the congestion status within the corresponding area; use the complete unblocked reliability, basic unblocked reliability, slight congestion reliability and severe congestion reliability as evaluation indicators;

[0059] Step 6: Based on the output of the congestion situation in the corresponding area, combined with the objective conditions of meteorology, control, on-site, and emergency response capabilities, proactively prevent and control the road sections that are about to be congested.

[0060] It is worth mentioning that, when extracting traffic status parameters, the present invention carries out research on wavelet density and kernel density non-parametric estimation of travel time reliability, short-term travel time prediction based on Boostrap-KNN, and fuzzy C-means clustering to discriminate traffic status; in the present invention, the concept of reliability is introduced into short-term interval prediction, making travel time reliability an important measure for evaluating the operating efficiency of the transportation network and the quality of travel services; at the same time, incorporating reliability indicators into the estimation and analysis of highway traffic operation status can enrich the connotation of transportation system evaluation and help managers and travelers make better decisions.

[0061] In step 1, the highway data obtained includes vehicle data, road condition information data, traffic flow information and environmental information data.

[0062] In the present invention, for example, vehicle data includes the real-time speed, driving trajectory, vehicle model, license plate number, vehicle brand, and fuel gauge reading of the vehicle. Specifically, it should be noted that the real-time speed, license plate track and photo number of the vehicle are obtained through the gantry high-definition monitoring on the highway, or in today's rapidly developing Internet of Things, for networked vehicles, the highway management department directly retrieves the vehicle license plate information and vehicle brand and model information from the Internet of Vehicles platform database corresponding to the vehicle brand when the vehicle enters the highway entrance. For example, for Buick vehicles, the highway management department directly retrieves the vehicle license plate from the iBuick platform used by Buick vehicles, but confidentiality management is required. The obtained vehicle data is only used to obtain vehicle information and prevent unauthorized personnel from maliciously viewing and calling. For vehicles without Internet of Vehicles, vehicle information is obtained through the gantry high-definition monitoring on the highway.

[0063] In the present invention, the highway management department directly retrieves vehicle data from the Internet of Vehicles platform database corresponding to the vehicle brand, and the acquired vehicle data is more accurate.

[0064] In the present invention, the road condition information data includes the road's flatness, width, road surface temperature, humidity, and hardness data. These data are directly acquired when the expressway is built. Similarly, during the subsequent road maintenance process, the road condition information is regularly recollected to complete the update of the road condition information.

[0065] In the present invention, traffic flow information includes the number of vehicles, vehicle speed, and vehicle type ratio in each road section. The traffic flow information is also directly obtained through high-definition monitoring of the gantry on the highway and data collected by toll stations.

[0066] The environmental factor information data in the present invention includes weather conditions, visibility, wind speed and direction, and rainfall. The environmental information data is obtained through a meteorological department or through an environmental factor collection device installed in a highway traffic network.

[0067] In step 2, in order to retain important valuable information, the asymmetric highway travel time data is preprocessed by combining the quartile method with statistical principles, and non-parametric tests are used to test the data distribution characteristics in time, space and vehicle model dimensions.

[0068] It is worth mentioning that in step 2, the quartile method is used to divide the data set into four equal parts, each containing 25% of the data; this method arranges all the values ​​from small to large and divides them into four equal parts; determines the values ​​of the three split point positions, namely the lower quartile (Q1), the median (Q2) and the upper quartile (Q3); the lower quartile is the 25% data point, the median is the 50% data point, and the upper quartile is the 75% data point; the quartiles can be calculated by sorting the data, determining the position of each quartile, and then calculating the value at the position, to help analyze and understand the distribution of the data. The data mainly obtained in the present invention mainly includes the time it takes for a vehicle to pass a certain route, which is used as the travel time of the vehicle along the route, and the travel time of the vehicle through the corresponding section is preprocessed by the quartile method.

[0069] like Figure 2 As shown, in step 3, the travel time distribution estimation is considered to be an important starting point for the quantitative study of travel time reliability; under the premise of uncertainty in probability distribution, the wavelet density estimation method is used to establish a non-parametric distribution model of highway travel time to estimate highway travel time.

[0070] In the present invention, a non-parametric (non-parametric) distribution model of highway travel time is established by using wavelet transform to perform density estimation, and the probability density function h(x) is decomposed into coefficients on each wavelet basis. The probability density function h(x) is estimated by estimating these wavelet decomposition coefficients. This method can provide localized analysis of time, space and frequency, and gradually refine the travel time at multiple scales through telescopic and translation operations, ultimately achieving time subdivision at high frequencies and frequency subdivision at low frequencies, and can automatically adapt to the requirements of time-frequency signal analysis, thereby focusing on any details of the signal.

[0071] like Figure 3As shown, in step 4, it is assumed that the distribution of the highway travel time population is unknown, but it is known that there is a data sample X from the population with a sample size of n = {x1, x2, ···, x n}, n∈N+; the specific steps of step 4 are as follows:

[0072] Step 401: Construct an original sample data set X. Take the highway toll collection system as the data source, calculate the starting and ending time of each historical record, and obtain the original sample X = {x1, x2, ···, x n}, n∈N+;

[0073] Step 402: Bootstrap resampling: extract m samples X from the original sample X with replacement * ={x * 1,x * 2,···,x * m}, m∈N+, usually m=n, this sample is called Bootstrap sample; establish B travel time Bootstrap samples X1 * , X2 * ,···,X B * ;

[0074] Step 403: Build a point prediction model; for each Bootstrap sample X * Perform travel time prediction modeling to obtain t i =f i (x)+ε,i=1,2,···,B;

[0075] Step 404: Calculate the prediction value sequence: Use the B point prediction models in the previous step to predict the test samples respectively, and obtain a sequence θ containing B prediction values * ={θ * 1, θ * 2,…,θ * B};

[0076] Step 405: Bootstrap interval prediction: Calculate the sequence of B prediction values ​​θ≠={θ * 1, θ * 2,…,θ * B}, that is, the travel time prediction interval [low(t), up(t)].

[0077] From the modeling steps, it can be seen that the point prediction model in step 403 and the Bootstrap confidence interval estimation method in step 405 will directly affect the performance and quality of travel time interval prediction, so these two key steps need to be discussed and analyzed in depth.

[0078] In step 405, the Bootstrap confidence interval estimation includes the interval estimation of the standard deviation, the percentile interval estimation and the accelerated bias correction interval estimation.

[0079] In the present invention, the highway data of historical records is first obtained to obtain a point prediction model, and then the highway travel time obtained in step 3 is used for prediction, and the final predicted travel time interval is [1ow(t), up(t)];

[0080] It must be said that, in the present invention, the highway travel time interval prediction uses the original sample X to establish the Bootstrap sample, and adopts the data-driven point prediction model to calculate the travel time prediction value of each Bootstrap sample. The prediction value sequence is used to construct an estimated range that may contain the true value. In the present invention, the interval of the formation time of the vehicle passing through the corresponding road section is predicted for further processing to determine the congestion level of the corresponding road section.

[0081] In step 5, the sample observation matrix is ​​first given:

[0082]

[0083] Among them, {Y1, Y2, ..., Y n} is an n observation sample with p variables, {y i1 ,y i2 , ..., y ip} represents the variable observation value corresponding to the i-th sample. It should be added that the variable observation value is the travel time interval predicted in step 4. n} is divided into c categories, n≥c≥2, and C={C1,Y2,...,C c} are c cluster centers, and Define the algorithm objective function:

[0084]

[0085] In the formula, U=(u ik ) c×n ; is the membership matrix, u ik Indicates the membership of the kth sample to the i-th category, 0≤u ik ≤1, d ij =||yj -c i || represents the i-th cluster center C i With the jth sample point y j The distance between them, m is the weighted index, which represents the fuzziness of the membership matrix, m∈[1,+∞], and the value of m is proportional to the fuzziness of the membership matrix.

[0086] In formula (2), since Euclidean distance is used as the similarity measurement function, the distance function is expressed as:

[0087]

[0088] in, is the weight of the kth traffic parameter in the FCM algorithm, and p is the number of traffic parameters.

[0089] The Lagrangian function is constructed for formula (2), as follows:

[0090]

[0091] Among them, λ j (j=1, 2, ...n) are n Lagrange multipliers. The necessary condition for using Lagrange multiplication to solve the minimum value is:

[0092]

[0093] in, According to the two necessary conditions of formula (5) and formula (6), the cluster center C and the membership matrix U are determined by iteration.

[0094] For formula (4), i and u ij Find the partial derivative and set it to zero, then we have:

[0095]

[0096] because Then we have:

[0097]

[0098] Transforming the above formula, we have:

[0099]

[0100] Combining formula (7) and formula (10), we can get:

[0101]

[0102] Through iterative calculation of the iterative algorithm, the cluster center c can be obtained i and membership u ij ; By comparing the distance between the real-time data samples and the center of each state class, the traffic operation status is determined based on the principle of the shortest distance.

[0103] In the present invention, the data source is the historical toll data of a certain highway from July 1 to 31 of a certain year. Small passenger cars (7 seats or less) are selected as the characteristic vehicle type, and 6:00-0:00 (next day) is selected as the research period with an interval of one hour. The historical data is divided into 31 (days) * 18 (segments) = 558 groups of data. Data cleaning and travel time distribution fitting are performed on each group of data. Considering the travel time and its fluctuation degree, the traffic operation status reliability index is calculated: smooth reliability, basically smooth reliability, light congestion reliability, and moderate congestion reliability. In the clustering algorithm, the number of categories c = 5, the fuzzy index is selected as m = 2, and the maximum number of iterations N is set. max =100, and the traffic operation status corresponding to each cluster center is determined according to the weighted distance between each cluster center and the reference point (1, 1, 1); the traffic operation status corresponding to the cluster center closest to the reference point (1, 1, 1) is unblocked, and the traffic operation status corresponding to the cluster center farthest from it is severely congested. The cluster centers corresponding to different traffic operation statuses can be obtained; finally, the traffic operation status is determined based on the principle of the shortest distance by comparing the distance between the real-time data sample and the center of each state class.

[0104] The working principle of the present invention is as follows:

[0105] like Figure 1-4 As shown, firstly, the historical traffic data of expressways is obtained, and then the non-parametric detection method is used to carry out data quality analysis and data feature analysis on the distribution characteristics of expressway data; the specific expressway travel time is obtained from the obtained historical traffic data of expressways; the non-parametric distribution model of expressway travel time is established by adopting the wavelet density estimation method, and the distribution of expressway travel time is estimated by using the above model, and finally the training of the non-parametric distribution model of expressway travel time is completed through the historical traffic data of expressways;

[0106] Then, the highway historical traffic data is continued to be used to adopt the highway travel time short-time interval prediction model based on Bootstrap-KNN to complete the highway travel time short-time interval prediction. Specifically, the travel time distribution output by the highway travel time non-parametric distribution model is used as the input of the Bootstrap-KNN highway travel time short-time interval prediction model, and the predicted output travel time interval is completed. The predicted travel time interval is compared with the travel time in the real historical data of the highway, and the Bootstrap-KNN highway travel time short-time interval prediction model is corrected, and finally the Bootstrap-KNN highway travel time short-time interval prediction model with more accurate output results is obtained; the training of the highway travel time non-parametric distribution model and the Bootstrap-KNN highway travel time short-time interval prediction model is completed;

[0107] Then, the real-time data on the highway is obtained. Similarly, the non-parametric detection method is first used to conduct data quality analysis and data feature analysis on the distribution characteristics of highway data. Then, the non-parametric distribution model of highway travel time is used to complete the estimation and distribution of highway travel time. The Bootstrap-KNN highway travel time short-term interval prediction model is used to predict the travel time in a specific highway interval to obtain the prediction interval. Finally, the fuzzy C-means clustering quantitative analysis method is used based on the statistical data of the travel time to consider the two characteristics of parameter weight and sample capacity difference to evaluate the traffic operation status; the congestion status in the corresponding area is output; the congestion status includes complete unobstructed reliability, basic unobstructed reliability, slight congestion reliability and severe congestion reliability; finally, according to the output of the congestion status in the corresponding area, combined with the objective conditions of meteorology, control, on-site and emergency handling capabilities, the road sections that are about to be congested are actively prevented and controlled to avoid road congestion in advance.

[0108] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention shall be implemented according to the conventional means in the art unless otherwise specified and limited.

Claims

1. A highway congestion prediction and active prevention and control method based on big data, characterized by: The following steps are involved: Step 1: Obtain real-time highway data; Step 2: Use non-parametric detection methods to conduct data quality analysis and data feature analysis on the distribution characteristics of highway data; Step 3: Use wavelet density estimation method to establish a non-parametric distribution model of highway travel time and estimate highway travel time; Step 4: Use the short-term prediction model of highway travel time based on Bootstrap-KNN to complete the short-term prediction of highway travel time; Step 5: Use the fuzzy C-means clustering quantitative analysis method based on the travel time statistics to evaluate the traffic operation status by considering the two characteristics of parameter weight and sample capacity difference; output the congestion status within the corresponding area; Step 6: Based on the output of the congestion situation in the corresponding area and in combination with objective conditions, proactively prevent and control the road sections where congestion is about to occur.

2. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: In step 1, the highway data obtained includes vehicle data, road condition information data, traffic flow information and environmental information data.

3. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: In the step 2, the asymmetric highway travel time data is preprocessed by combining the quartile method with statistical principles, and non-parametric tests are used to test the distribution characteristics of the data in the time, space and vehicle type dimensions.

4. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: In step 3, a wavelet density estimation method is used to establish a non-parametric distribution model of highway travel time to estimate the highway travel time.

5. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: The specific steps of step 4 are as follows: Step 401: Construct an original sample data set X. Take the highway toll system as the data source, calculate the starting and ending travel time of each historical record, and obtain the original sample X = {x1, x2, ···, x n }, n∈N+; Step 402: Bootstrap resampling: extract m samples X from the original sample X with replacement * ={x * 1,x * 2,···,x * m }, m∈N+, usually m=n, this sample is called Bootstrap sample; establish B travel time Bootstrap samples X1 * , X2 * ,···,X B * ; Step 403: constructing a point prediction model; For each Bootstrap sample X * Perform travel time prediction modeling to obtain t i =f i (x)+ε,i=1,2,···,B; Step 404: Calculate the prediction value sequence: Use the B point prediction models in the previous step to predict the test samples respectively, and obtain a sequence θ containing B prediction values * ={θ * 1, θ * 2,…,θ * B }; Step 405: Bootstrap interval prediction; calculate the sequence θ of B predicted values * ={θ * 1, θ * 2,…,θ * B }, that is, the travel time prediction interval [low(t), up(t)].

6. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: In step 405, the Bootstrap confidence interval estimation includes the interval estimation of the standard deviation, the percentile interval estimation and the accelerated bias correction interval estimation.

7. The method for predicting and actively preventing highway congestion based on big data according to claim 1 is characterized by: In step 5, the sample observation matrix is ​​first given: Among them, {Y1, Y2, ..., Y n } is an n observation sample with p variables, {y i1 ,y i2 , ..., y ip } represents the variable observation value corresponding to the i-th sample, and the sample data {Y1, Y2, ..., Y n } is divided into c categories, n≥c≥2, and C={C1,Y2,...,C c } are c cluster centers, and i=1,2,...,define the algorithm objective function: In the formula, U=(u ik ) c×n ; is the membership matrix, u ik Indicates the membership of the kth sample to the i-th category, 0≤u ik ≤1, d ij =||y j -c i || represents the i-th cluster center C i With the jth sample point y j The distance between them, m is the weighted index, which represents the fuzziness of the membership matrix, m∈[1,+∞], and the value of m is proportional to the fuzziness of the membership matrix.

8. The method for predicting and actively preventing highway congestion based on big data according to claim 7 is characterized by: In formula (2), since Euclidean distance is used as the similarity measurement function, the distance function is expressed as: in, is the weight of the kth traffic parameter in the FCM algorithm, and p is the number of traffic parameters.

9. The method for predicting and actively preventing highway congestion based on big data according to claim 1, characterized in that: The Lagrangian function is constructed for formula (2), as follows: Among them, λ i (j=1, 2, ...n) are n Lagrange multipliers. The necessary condition for using Lagrange multiplication to solve the minimum value is: Where 0≤u ij ≤1, j=1, ..., n; c; j = 1, ..., n; According to the two necessary conditions of formula (5) and formula (6), the cluster center C and the membership matrix U are determined by iteration.

10. The highway congestion prediction and active prevention and control method based on big data according to claim 1 is characterized by: For formula (4), i and u ij Find the partial derivative and set it to zero, and iterate through the iterative algorithm to get the cluster center c i and membership u ij ; By comparing the distance between the real-time data samples and the center of each state class, the traffic operation status is determined based on the principle of the shortest distance.

Citation Information

Patent Citations

  • Method for predicting congestion of expressway toll station

    CN111243279A

  • Method and system for predicting real-time road condition of road in front of logistics vehicle based on V2V communication

    CN113345225A

  • Congestion early warning method based on traffic parameter short-time prediction

    CN118097969A