Congestion pre-control range division method based on local node importance

Abstract

The invention discloses a congestion pre-control range division method based on local node importance, and the method comprises the following steps: firstly, determining a congestion early warning starting threshold value through congestion sample analysis; secondly, for a road section reaching an early warning threshold value, determining a congestion pre-control node set based on an adjacent matrix; then, in combination with a correlation coefficient of a traffic state associated road section, defining a local node importance degree evaluation index capable of representing a congested road section, and performing node importance degree evaluation by using an analytic hierarchy process; and finally, dividing importance levels of local nodes by using a K-Means + + Medoids clustering algorithm, and determining a grading pre-control range. According to the method, the potential congestion points are identified, the graded congestion pre-control range is searched, the road network operation efficiency is improved, the congestion control cost is reduced, and the method has high application value.

Description

Technical Field

[0001] The invention belongs to the field of traffic control, and in particular relates to a method for dividing a congestion pre-control range based on the importance of local nodes. Background Art

[0002] With the continuous improvement of urbanization level and the continuous expansion of urban scale, the travel demand level of urban residents is increasing, resulting in frequent traffic congestion. Congestion not only causes traffic delays, but also aggravates environmental pollution, increases the risk of traffic accidents, and brings huge challenges to urban traffic management. Once traffic congestion is formed, even if congestion relief measures are taken, it is difficult to completely avoid the negative impact brought about by it. In this context, the present invention aims to identify potential congestion points based on road traffic status prediction data, find the congestion classification pre-control range, break through the limitations of traditional congestion relief measures for post-control, and become the core support for active pre-control of congestion. In view of the current research status of congestion control range, the current methods for finding congestion control range mainly have the following problems:

[0003] (1) Congestion warning research is not in-depth enough, and it is easy to mistakenly activate congestion warnings due to accidental traffic flow fluctuations. It fails to distinguish between congestion that can dissipate on its own in a short time and congestion events that require preventive measures, resulting in waste of resources and possible traffic disruptions.

[0004] (2) The congestion control range lacks differentiation. Most studies use clustering methods to divide the cells so that after identifying congestion, regional control can be carried out based on the cells. The cell coverage area delineated by this division method is generally large, and the differences in specific congestion causes and characteristics of different sections within the cells are ignored, which does not meet the needs of refined control.

[0005] (3) The importance of nodes in congestion propagation is not evaluated for congested sections. Some studies have identified important nodes in the entire road network through node importance evaluation, but the selected evaluation indicators usually consider the global level of the road network and cannot characterize the importance of each node for congested sections. Therefore, correctly identifying and distinguishing which nodes are key nodes related to congested sections and taking different control measures for nodes of different levels are important prerequisites for achieving accurate and efficient congestion control. Summary of the invention

[0006] Purpose of the invention: In order to solve the technical problems mentioned in the above background technology, the present invention constructs a method for dividing the congestion pre-control range based on the importance of local nodes based on the road traffic status prediction data. Through the congestion sample analysis, the congestion warning activation threshold is determined; for the road section that reaches the warning threshold, the congestion pre-control node set is determined based on the adjacency matrix; the local node importance evaluation index that can characterize the congested road section is defined, and the node importance is evaluated using the hierarchical analysis method; the local node importance level is divided using the K-Means++Medoids clustering algorithm. Compared with the traditional method, the congestion pre-control range determination and classification method designed by the present invention is advanced mainly in that: through the continuous observation of traffic status labels, in order to avoid the erroneous activation of congestion warnings due to accidental traffic flow fluctuations, a clear trigger condition is defined. When it is detected that the moderate or severe congestion state lasts for more than 10 minutes, the warning is activated to avoid waste of resources and possible traffic interference caused by unnecessary intervention. Compared with the traditional method of finding traffic zones through clustering to determine the control range, the present invention defines local node compactness, local node betweenness, and local node strength for congested sections, combines node degree and node clustering coefficient, quantifies the importance of pre-control nodes to congestion propagation, and designs a K-Means++Medoids algorithm to determine the three-level pre-control range, realizing the transformation of the local importance of nodes from continuous quantitative evaluation to discontinuous qualitative grading. This method provides a scientific basis and method support for the formulation of differentiated congestion pre-control measures, making congestion pre-control more accurate and targeted.

[0007] Technical solution: In order to achieve the above technical objectives, the present invention proposes a method for dividing the congestion pre-control range based on the importance of local nodes, which method includes the following steps:

[0008] (1) Based on the predicted traffic conditions, identify the road sections that meet the congestion warning threshold;

[0009] (2) Determine the preliminary control range according to the adjacency matrix and obtain the pre-control node set;

[0010] (3) Considering the impact of each node on the speed of congestion propagation, define the local node importance evaluation index for congested road sections;

[0011] (4) Calculate the importance evaluation value of local nodes using the hierarchical analysis method;

[0012] (5) Design a K-Means++Medoid clustering algorithm to classify the initially obtained control range based on the node importance value.

[0013] Furthermore, the step (1) specifically includes:

[0014] (101) Observe the traffic status data of all road sections, plot the duration distribution of traffic congestion, set 10 minutes as the congestion warning threshold, and immediately initiate a congestion warning for congestion events that last for 10 minutes or more;

[0015] (102) Identify road sections where traffic conditions are congested and the congestion lasts for more than 10 minutes.

[0016] Furthermore, the specific steps of step (2) are as follows:

[0017] (201) Define the road network as a directed graph G = (V, E), V = {v1, v2, v3, ... v k} represents the set of all nodes in the road network including the starting node and the intermediate nodes, k is the number of nodes, E = {(v i ,v j )|i,j=0,1,2,……k} represents the set of road segments consisting of all nodes, where (v i ,v j ) represents a road segment connected by nodes i and j. Each edge in G is assigned a weight, and the weight is used to represent the distance between the road segments of each road node. The adjacency matrix A[i,j] of the connection relationship between each node and the distance matrix D[i,j] of the road segments between nodes are established. The formula is as follows:

[0018]

[0019] D[i,j]=d ij

[0020] Among them, a ij Represents the element corresponding to the i-th row and j-th column of the A[i,j] matrix; d ij represents the distance between the i-th road segment and the j-th road segment;

[0021] (202) After determining the road section where the congestion detector is located, the congested road section (v m ,v n )’s first-layer adjacent nodes, that is, nodes that satisfy a mo ∪a om =1 and a np ∪a pn = 1, where o,p∈V; extract the second-layer adjacent nodes according to the first-layer adjacent nodes in the same way, that is, a oq ∪a oq =1 and a pr ∪a rp = 1, where r,q∈V, and so on, extracting three layers of adjacent nodes, i.e., a rs ∪a sr =1 and aqt ∪a tq = 1, where s, t∈V. The extracted three-layer adjacent nodes are taken as the pre-controlled node set.

[0022] Furthermore, the specific steps of step (3) are as follows:

[0023] (301) Local node importance identification parameters are defined for congested road sections, and local node importance node evaluation indicators are formed by combining node degree and node compactness. The definitions and calculations of various indicators are as follows:

[0024] (1) Local node density: In the traffic network structure, node density is usually represented by the inverse of its connectivity, which reflects the density of each node itself. In the local node density of congested sections, the inverse of the average path length from each node to the congested node is used to represent the local density of each node to the congested node.

[0025]

[0026] Among them, d im represents the average distance of each path between node i and congested node m; B i Represents the local node closeness of the i-th node.

[0027] (2) Local node strength: Node strength reflects the strength of the node in carrying traffic flow. The connection weight between nodes uses the average traffic flow of the road between the two nodes. In the local node strength for congested sections, in order to reflect the strength of the node carrying traffic flow related to the congested section, the congestion correlation between the traffic state-related sections is used on the basis of the original node strength definition to define the influence of the congested section on the node. That is, when the node and the congested node are not geographically adjacent, the product of the traffic state correlation coefficient and the average traffic flow between the two points is used as the local node strength.

[0028] S i1 =∑ i∈V a im w im / 2

[0029] S i2 =∑r im w im / 2

[0030] S i =S i1 +S i2

[0031] Among them, w im is the weight between nodes, expressed as the average lane flow of nodes i and m; r imis the traffic status correlation coefficient between nodes i and m;

[0032] The traffic state correlation coefficient between road sections is calculated by analyzing the mutual correlation of traffic states between road sections and using the Pearson correlation coefficient calculation formula. This method includes three steps: constructing a traffic state time series considering time lag; calculating the degree of mutual correlation between these time series; determining the traffic state associated road sections according to the Pearson correlation coefficient. The relevant definitions of the road section traffic state mutual correlation analysis method are as follows:

[0033] (a) Traffic status time series: Assuming the lane i , the sequence of its traffic status changing over time is marked as Seq i , the sequence consists of a series of elements, whose values ​​are 0 or 1, 0 represents smooth traffic and 1 represents congestion;

[0034] (b) Time-delayed traffic state time series: when the lane i The congestion condition after time t k Later it affects the lane section j This time interval t k It is defined as the time lag between two road segments; for each lane j , and its time-lagged traffic status time series KSeq j The construction process is as follows: From the original congestion state time series Seq j Delete the first t k data points, and get the truncated sequence Seq j_truncated , in the truncated sequence Seq j_truncated Add k "-1" values ​​to the end of to form a complete time-lagged traffic status time series KSeq j ,Due to the difference in the diffusion speed of traffic between different road sections, the time lag between each pair of road sections is also different. Taking the maximum correlation coefficient as the goal, the optimal time lag is found for each pair of road sections by setting the loop function;

[0035] (c) Inter-link correlation: Assume there are two lanes i and lane j , where lane i The traffic status time series is Seq i , and lane j The traffic status time series considering time lag is KSeq j , when the time lag cross-correlation degree of the two sections, i.e. TLCC(Seq i ,KSeq j ) exceeds the preset threshold r, lanei and lane j There is a causal relationship between them, which is a set of traffic status associated road sections;

[0036] (d) Traffic status correlation coefficient: The traffic status correlation coefficient of two road sections is calculated using the Pearson correlation coefficient formula, which is as follows:

[0037]

[0038] Among them, x i and i Lane i and lane j The congestion status value at the corresponding time; n1 is the total number of time series; Seq i Lane i Traffic status time series; KSeq j For lane j Considering the time series of traffic status with time lag;

[0039] Assume lane i , the sequence of its traffic status changing over time is marked as Seq i ;lane i The congestion condition after time t k Later it affects the lane section j , construct the traffic status time series KSeq considering time lag j According to TLCC (Seq i ,KSeq j ) calculates the traffic state correlation coefficient between road sections; when the time-lagged mutual correlation degree of two road sections, i.e., TLCC(Seq1, NSeq2), exceeds the preset threshold r, it is determined that there is a causal relationship between lane1 and lane2, which is a group of traffic state-related road sections. The corresponding road sections with Pearson correlation coefficient>0.6 are selected to construct the related road section matrix.

[0040] (3) Local node betweenness refers to the frequency with which a node appears on the shortest path between all pairs of nodes. In the local node betweenness for congested road sections, the local node betweenness is defined as the frequency with which a node appears on the shortest path between all pairs of nodes passing through the congested node.

[0041]

[0042] Among them, σ stm is the number of shortest paths from node s to node t via congested node m; σ stm (i) is the number of these paths passing through the i-th node;

[0043] (4) Node degree and node clustering coefficient: The degree k of the i-th node i Represents the total number of edges connected to the node; for the i-th node, the clustering coefficient C i The number of relationships E between the i-th node and its neighboring nodes i The degree k of the i-th node i The relationship between:

[0044]

[0045] Among them, E i is the relationship number of the neighboring nodes of the i-th node, obtained by counting the triangles of the i-th node; k i is the degree of the ith node.

[0046] Furthermore, the specific steps of step (4) are as follows:

[0047] (401) Due to the local node density B i , local node betweenness J i , local node strength S i Considering the impact on the diffusion speed of congested nodes, the node degree k i and node clustering coefficient C i It can more accurately reflect the impact on the diffusion speed of congested nodes, and the node degree has a greater impact on the diffusion speed of congestion than the node clustering coefficient. The judgment matrix B is constructed, and the values ​​in the matrix represent the relative importance of each local node importance node evaluation index;

[0048]

[0049] (402) The judgment matrix B in the hierarchical analysis method is calculated by normalization, and the sum of its j-th column is expressed as sum j , then the element b in the i-th row and j-th column of the matrix ij The normalized value n ij The calculation is as follows:

[0050] n ij =b ij / sum j

[0051] The normalized matrix is ​​represented as N, and each element n ij That is the element b of the original judgment matrix ij The results of column normalization are as follows:

[0052]

[0053] (403) For the normalized matrix N, a vector is obtained by summing up the rows. Let vector S represent the result of summing up the normalized matrix N by rows, and each component in S is the sum of the elements of the corresponding row of the matrix;

[0054] (404) Normalize S to obtain the eigenvector ω = (0.16, 0.05, 0.26, 0.26, 0.26) T , ω is the eigenvector of matrix B, find the maximum eigenvalue λ of the eigenvector max , the calculation formula is as follows:

[0055]

[0056] Where: max is the maximum eigenvalue of the judgment matrix; n is the order of the judgment matrix, ω i is an element in the eigenvector ω;

[0057] (405) A consistency check is performed to evaluate the consistency of the judgment matrix. If the consistency of the judgment matrix is ​​not accepted, the elements of the matrix are re-evaluated. The formula is as follows:

[0058]

[0059] Among them: CI < 0.1, the judgment matrix is ​​considered to have acceptable consistency; CI ≥ 0.1, the consistency is considered insufficient and the judgment matrix needs to be reviewed and adjusted.

[0060] (406) According to the results calculated by the hierarchical analysis method, the corresponding weight coefficients of each eigenvalue index are ω = (0.16, 0.05, 0.26, 0.26, 0.26) T , calculate the local node importance evaluation matrix R according to the following formula:

[0061]

[0062] The characteristic index values ​​of each node and their corresponding weights are comprehensively considered, and the importance of each node is quantitatively scored to obtain the local node importance evaluation value dataset X.

[0063] Furthermore, the specific steps of step (5) are as follows:

[0064] (501) Randomly select a sample point from the local node importance evaluation value data set X as the first center point;

[0065] (502) Calculate the shortest distance D(x) from each point to the current center point, and determine the probability P(x) of selecting the next center point based on this distance. The farther the distance, the higher the probability of the point being selected as the new center.

[0066] (503) Iteratively update the center point until k initial cluster centers are selected;

[0067]

[0068] (504) Within each cluster, calculate an actual sample point x of the cluster s The sum of the squared distances to all points in the cluster, select the point with the smallest sum of squared distances to determine cluster E l The new cluster center u l ;

[0069]

[0070] Among them, x s is the candidate new cluster center, and x is the sample point of the cluster;

[0071] (505) Based on these selected cluster centers, the distance between each sample and all cluster centers is calculated, and it is assigned to the cluster with the closest distance, so that the distance from each class of sample point x to the cluster center μ l The SSE is the smallest, and the SSE calculation formula is as follows:

[0072]

[0073] (406) Repeat steps (404) and (405) until the cluster center is stable or the maximum number of iterations is reached;

[0074] (407) According to the silhouette coefficient and the minimum sum of squared errors, the optimal number of clusters is determined to be 3. For the pre-control range initially obtained, the levels are divided based on the node importance values.

[0075] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0076] Compared with the traditional method, the method for determining and grading the congestion pre-control range designed by the present invention is advanced mainly in that: through continuous observation of traffic status labels, a clear trigger condition is defined to avoid the erroneous activation of congestion warnings due to accidental traffic flow fluctuations. When it is detected that the moderate or severe congestion state lasts for more than 10 minutes, the warning is activated to avoid waste of resources and possible traffic interference caused by unnecessary intervention. Compared with the traditional method of clustering to find traffic cells to determine the control range, the present invention defines the local node density, local node betweenness, and local node strength for congested sections, combines the node degree and the node clustering coefficient, quantifies the importance of the pre-control node to the congestion propagation, and designs the K-Means++Medoids algorithm to determine the three-level pre-control range, realizing the conversion of the local importance of the node from continuous quantitative evaluation to non-continuous qualitative grading. This method provides a scientific basis and method support for the formulation of differentiated congestion pre-control measures, making congestion pre-control more accurate and targeted. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 is a flow chart of the method of the present invention;

[0078] Figure 2 is the distribution of the duration of the congestion event in the present invention;

[0079] Figure 3 It is a comparison diagram of lane flow and traffic status of the inspection station number "G103L221110112" in the present invention;

[0080] Figure 4 is a schematic diagram of a congestion pre-control node set in the present invention;

[0081] Figure 5 is the correlation coefficient matrix of the road sections associated with the detection points in Tongzhou District in the present invention;

[0082] Figure 6 It is a graph showing the change of the silhouette coefficient (SC) with the number of clusters in the present invention;

[0083] Figure 7 is a graph showing the variation of the sum of squared errors (SSE) with the number of clusters in the present invention;

[0084] Figure 8 It is a schematic diagram of the three-level pre-control range division of congestion based on node importance in the present invention. DETAILED DESCRIPTION

[0085] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0086] The present invention proposes a method for dividing the congestion pre-control range based on the importance of local nodes. Figure 1 As shown, the following steps are included:

[0087] (1) Based on the predicted road traffic status data on roads at all levels in Tongzhou District, Beijing, the distribution of congestion event duration is plotted to determine the reasonable threshold for congestion warning, such as Figure 2 As shown in Table 1, the statistical values ​​of congestion frequency distribution indicators are shown:

[0088] Table 1 Statistics of congestion frequency distribution indicators

[0089]

[0090]

[0091] Combination Figure 2 As can be seen from Table 1, since the detector statistical data interval is 5 minutes, the minimum value in the congestion duration sample is 5 minutes, and this type of congestion sample accounts for 69% of the total congestion samples. This type of congestion can dissipate on its own in a short time (5 minutes) and will not cause serious impact, and no congestion prevention measures are required. Therefore, when designing the warning activation threshold, 10 minutes is used as a key time node, and for those congestion events that reach or exceed this threshold, the congestion warning is immediately activated.

[0092] (2) Define the road network as a directed graph G = (V, E), V = {v1, v2, v3, ... v n} represents the set of all nodes in the road network including the starting node and the intermediate nodes, E = {(v i ,v j )|i,j=0,1,2,……n} represents the set of road segments consisting of all nodes, where (v i ,v j ) represents a road segment connected by nodes i and j. Each edge in G is assigned a weight, and the weight is used to represent the distance between each road node. The adjacency matrix A[i,j] of the connection relationship between each node and the distance matrix D[i,j] of the road segment between nodes are established;

[0093]

[0094] D[i,j]=d ij

[0095] Among them, a ij Represents the element corresponding to the i-th row and j-th column of the A[i,j] matrix; d ij Represents the distance between road section i and road section j. Select the detection station number "G103L221110112" with a high frequency of congestion to draw a traffic status distribution map. Its traffic status labels 0 to 3 represent from smooth to severe congestion, respectively. Figure 3As shown. A congestion event is detected at the black dotted line in the figure, and a congestion warning is initiated. After determining that the congested section where the detector "G103L221110112" is located is node 12-23, the first-level adjacent nodes of the congested section are extracted according to the adjacency matrix; then the second-level adjacent nodes are extracted according to the same method based on the first-level adjacent nodes, and so on, the third-level adjacent nodes are extracted. These adjacent sections are used as the preliminary control range to obtain the set of pre-control nodes that need to be calculated. The pre-control range is initially determined as follows Figure 4 shown.

[0096] (3) According to the definition of local node importance evaluation index for congested road sections, calculate the local node importance evaluation value. Obtain the basic parameter node distance of the road network, as shown in Table 2:

[0097] Table 2 is the distance table of some nodes in the road network

[0098]

[0099]

[0100] For each lane i The construction process of the time-lagged congestion state time series NSeqi is as follows: delete the first k data points from the original congestion state time series Seqi to obtain the truncated sequence Seqi_truncated, ensuring that the time-lagged sequence can correctly reflect the changes in the congestion state of the road section after the time lag of k units; add k "-1" values ​​at the end of the truncated sequence Seqi_truncated to form a complete time-lagged congestion state time series NSeqi. Adding "-1" is to ensure that the length of the time series is consistent with the original sequence, and "-1" is used as an identification value to distinguish the original data from the padding data added due to the time lag.

[0101] After constructing the traffic state time series Seq1 of lane1 and the traffic state time series NSeq2 of lane2 with time lag, the degree of mutual correlation between these time series is calculated according to the time lag mutual correlation degree TLCC(Seq1,NSeq2) of the two sections. When TLCC(Seq1,NSeq2) exceeds the preset threshold r, it can be determined that there is a causal relationship between lane1 and lane2, which is a set of traffic state related sections. Draw the correlation coefficient matrix of the traffic state related sections of the detection points in Tongzhou District, as shown in Figure 5 The corresponding road sections with Pearson correlation coefficient>0.6 are selected as traffic status associated road sections.

[0102] The relevant eigenvalues ​​of the importance evaluation index of each local node are calculated, and then normalized and scaled to between 0 and 1 to eliminate the dimensional influence of different eigenvalues ​​to ensure that they are comparable on the same scale. The normalized eigenvalues ​​of the nodes are shown in Table 3:

[0103] Table 3 shows the normalized eigenvalues ​​of nodes

[0104]

[0105]

[0106] According to the pairwise comparison method and the introduction of a nine-point scale, Table 4 shows the quantitative values ​​of the scale and the meaning of the weight ratios they represent:

[0107] Table 4 Quantification values ​​of ratio scale

[0108]

[0109] When the relative importance between the evaluation indicators is 1, it means that the two indicators are equally important; when the value is 3, it means that one indicator is slightly more important than the other; and when the value reaches 9, it means that one indicator is much more important than the other. For those situations between adjacent evaluations, the four intermediate values ​​of 2, 4, 6, and 8 are assigned. According to this rule, the element b in the judgment matrix B is ij It is defined as the importance of the i-th index compared to the j-th index. Since local node strength, local betweenness, and local compactness take into account the impact on the diffusion speed of congested nodes, they are more important than node degree and clustering coefficient. Node degree has a greater impact on the congestion diffusion speed than compactness. According to the hierarchical analysis method, the judgment matrix B is constructed as shown in Table 5:

[0110] Table 5 is the judgment matrix

[0111]

[0112] The summation method is used to normalize the judgment matrix in the hierarchical analysis method. Specifically, if the judgment matrix is ​​B, the sum of its j-th column is expressed as sum j , then the element b in the i-th row and j-th column of the matrix ij The normalized value ni j The calculation is as follows:

[0113] n ij =b ij / sum j

[0114] The normalized matrix is ​​represented by N, and each element n ij That is the element b of the original judgment matrix ij The results of column normalization are as follows:

[0115]

[0116] For a normalized matrix N, summing up by row will result in a vector, each component of which is the sum of the elements in the corresponding row of the matrix. Let vector S represent the result of summing up the normalized matrix N by row, then each component S i The calculation formula is:

[0117]

[0118] Performing the above calculation will obtain the comprehensive weight S of each node.

[0119] Normalize S to get the eigenvector:

[0120] ω=(ω1,ω2,…,ω n ) T

[0121] After calculation, the eigenvector is ω = (0.16, 0.05, 0.26, 0.26, 0.26) T ω is the eigenvector of matrix A, and the maximum eigenvalue λ of the eigenvector is further calculated. max , the calculation formula is as follows:

[0122]

[0123] Calculate the maximum eigenvalue λ max It is 5.01.

[0124] Consistency check is an important step in the hierarchical analysis method, which is used to evaluate the consistency of the judgment matrix. If the consistency of the judgment matrix is ​​not accepted, the elements of the matrix need to be re-evaluated. The formula is as follows:

[0125]

[0126] Where: max is the maximum eigenvalue of the judgment matrix; n is the order of the judgment matrix.

[0127] If CI < 0.1, the judgment matrix is ​​considered to have acceptable consistency; if CI ≥ 0.1, the consistency is considered insufficient and the judgment matrix needs to be reviewed and adjusted. ω = (0.16, 0.05, 0.26, 0.26, 0.26) T The consistency test results are:

[0128]

[0129] CI < 0.1, indicating that the judgment matrix is ​​reasonable and passes the consistency test. According to the results calculated by the hierarchical analysis method, the corresponding weight coefficients of each eigenvalue index are as follows:

[0130]

[0131] The local node importance evaluation matrix R is calculated according to the following formula:

[0132]

[0133] According to the graph structure drawn in the previous article and the characteristic value parameters of each node obtained, substituting them into the importance evaluation matrix R, the importance evaluation values ​​of each node can be obtained as shown in Table 6:

[0134] Table 6 is the local node importance evaluation value table

[0135]

[0136]

[0137] (4) After obtaining the local node importance values ​​calculated for congested nodes, in order to facilitate hierarchical control, the importance of the nodes needs to be converted from quantitative values ​​to hierarchical qualitative ratings, and different levels of control ranges are called according to different degrees of congestion. Based on the local node importance data of the road network, the K-Means++Medoid algorithm is used to determine the node importance classification level. The number of clusters is taken as 2 to 6, and the SC and SSE values ​​are used to analyze the clustering effect. The comparison of the effect evaluation indicators of different cluster numbers is shown in Table 7:

[0138] Table 7 shows the effect evaluation indicators under different cluster numbers

[0139]

[0140] Draw a graph showing the change in the silhouette coefficient index for different numbers of clusters, such as Figure 6 The silhouette coefficient ranges from -1 to 1, and the higher the value, the better the clustering effect. Figure 6 According to the data in , when the number of clusters is 3, the SC value is the highest (0.62), which indicates that the clustering effect is best when the number of clusters is 3.

[0141] Plot the squared error and variation graph for different numbers of clusters, such as Figure 7 As shown in Table 8 and Figure 7According to the data in , SSE decreases as the number of clusters increases. This is because more cluster centers will result in a shorter distance between the sample and its nearest cluster center. According to the change in SSE value from 3 to 4 compared to the change from 2 to 3, the rate of decline begins to slow down, indicating that the number of clusters 3 is the "elbow" of SSE. Combining the analysis results of SC and SSE, the optimal number of clusters is 3. After determining the number of clusters, the node importance level is divided according to the K-Means++Medoids clustering algorithm, and the pre-control nodes are divided into three levels. Draw the three-level pre-control range of congestion based on local node importance, as shown in Figure 8 The node importance level classification is shown in Table 8:

[0142] Table 8 Results of classification of different node importance levels

[0143]

[0144] (5) After predicting that the section of road where nodes 12-23 are located will be severely congested at around 6:30 am, the corresponding level of control range is called according to the evaluation result of the congestion level (moderate congestion or severe congestion). The first-level control range (nodes 29, 8, 9, 22, 11, 10, 41) is used as a high-priority control area, the second-level control range (nodes 15, 14, 20, 6, 36, 31, 27, 21, 3, 24, 34, 25, 32, 13, 37, 43, 53, 38, 46, 45) is used as a medium-level control area, and the third-level control range (nodes 5, 4, 1, 2, 19, 26, 35, 33, 54, 39, 30, 17, 42, 18) is used as an auxiliary control area. Temporary control measures are taken in advance to avoid congestion.

[0145] The embodiments are only for illustrating the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for dividing congestion pre-control range based on local node importance, characterized in that: The method comprises the following steps: (1) Based on the predicted traffic conditions, identify the road sections that meet the congestion warning threshold; (2) Determine the preliminary control range according to the adjacency matrix and obtain the pre-control node set; (3) Considering the impact of each node on the speed of congestion propagation, define the local node importance evaluation index for congested road sections; (4) Calculate the importance evaluation value of local nodes using the hierarchical analysis method; (5) Design a K-Means++Medoid clustering algorithm to classify the initially obtained control range based on the node importance value.

2. The method for dividing the congestion pre-control range based on the importance of local nodes according to claim 1, characterized in that: The step (1) specifically comprises: (101) Observe the traffic status data of all road sections, plot the duration distribution of traffic congestion, set 10 minutes as the congestion warning threshold, and immediately initiate a congestion warning for congestion events that last for 10 minutes or more; (102) Identify road sections where traffic conditions are congested and the congestion lasts for more than 10 minutes.

3. The method for dividing the congestion pre-control range based on the importance of local nodes according to claim 2, characterized in that: The specific steps of step (2) are as follows: (201) Define the road network as a directed graph G = (V, E), V = {v1, v2, v3, ... v k } represents the set of all nodes in the road network including the starting node and the intermediate nodes, k is the number of nodes, E = {(v i ,v j )|i,j=0,1,2,……k} represents the set of road segments consisting of all nodes, where (v i ,v j ) represents a road segment connected by nodes i and j. Each edge in G is assigned a weight, and the weight is used to represent the distance between the road segments of each road node. The adjacency matrix A[i,j] of the connection relationship between each node and the distance matrix D[i,j] of the road segments between nodes are established. The formula is as follows: D[i,j]=d ij Among them, a ij Represents the element corresponding to the i-th row and j-th column of the A[i,j] matrix; d ij represents the distance between the i-th road segment and the j-th road segment; (202) After determining the road section where the congestion detector is located, the congested road section (v m ,v n )’s first-layer adjacent nodes, that is, nodes that satisfy a mo ∪a om =1 and a np ∪a pn = 1, where o,p∈V; extract the second-layer adjacent nodes according to the first-layer adjacent nodes in the same way, that is, a oq ∪a oq =1 and a pr ∪a rp = 1, where r,q∈V, and so on, extracting three layers of adjacent nodes, i.e., a rs ∪a sr =1 and a qt ∪a tq =1, where s,t∈V, and the extracted three-layer adjacent nodes are used as the pre-controlled node set.

4. The method for dividing the congestion pre-control range based on the importance of local nodes according to claim 3 is characterized in that: The specific steps of step (3) are as follows: (301) Local node importance identification parameters are defined for congested road sections, and local node importance node evaluation indicators are formed by combining node degree and node compactness. The definitions and calculations of various indicators are as follows: (1) Local node density In the traffic network structure, node density is usually represented by the inverse of its connectivity, which reflects the density of each node itself. In the local node density of congested sections, the inverse of the average path length from each node to the congested node is used to represent the local density of each node to the congested node. Among them, d im represents the average distance of each path between node i and congested node m; B i Represents the local node closeness of the i-th node; (2) Local node strength Node strength reflects the strength of the node in carrying traffic flow. The connection weight between nodes uses the average traffic flow of the road between the two nodes. In the local node strength for congested sections, in order to reflect the strength of the node carrying traffic flow related to the congested section, the congestion correlation between the traffic state-related sections is used on the basis of the original node strength definition to define the influence of the congested section on the node, that is, when the node and the congested node are not geographically adjacent, the product of the traffic state correlation coefficient and the average traffic flow between the two points is used as the local node strength. S i1 =∑ i∈V a im w im / 2 S i2 =∑r im w im / 2 S i =S i1 +S i2 Among them, w im is the weight between nodes, expressed as the average lane flow of node i and congested node m; r im is the traffic status correlation coefficient between node i and congested node m; The traffic state correlation coefficient between road sections is calculated by analyzing the mutual correlation of traffic states between road sections and using the Pearson correlation coefficient calculation formula. This method includes three steps: constructing a traffic state time series considering time lag; calculating the degree of mutual correlation between these time series; determining the traffic state associated road sections according to the Pearson correlation coefficient. The relevant definitions of the road section traffic state mutual correlation analysis method are as follows: (a) Traffic status time series: Assuming the lane i , the sequence of its traffic status changing over time is marked as Seq i , the sequence consists of a series of elements, whose values ​​are 0 or 1, 0 represents smooth traffic and 1 represents congestion; (b) Time-delayed traffic state time series: when the lane i The congestion condition after time t k Later it affects the lane section j This time interval t k It is defined as the time lag between two road segments; for each lane j , and its time-lagged traffic status time series KSeq j The construction process is as follows: From the original congestion state time series Seq j Delete the first t k data points, and get the truncated sequence Seq j_truncated , in the truncated sequence Seq j_truncated Add k "-1" values ​​to the end of to form a complete time-lagged traffic status time series KSeq j ,The speed of traffic diffusion between different road sections is different, and the time lag between each road section is different. Taking the maximum correlation coefficient as the goal, the optimal time lag is found for each pair of road sections by setting a cyclic function; (c) Inter-link correlation: Assume there are two lanes i and lane j , where lane i The traffic status time series is Seq i , and lane j The traffic status time series considering time lag is KSeq j , when the time lag cross-correlation degree of the two sections, i.e. TLCC(Seq i ,KSeq j ) exceeds the preset threshold r, lane i and lane j There is a causal relationship between them, which is a set of traffic status associated road sections; (d) Traffic status correlation coefficient: The traffic status correlation coefficient of two road sections is calculated using the Pearson correlation coefficient formula, which is as follows: Among them, x i and i Lane i and lane j The congestion status value at the corresponding time; n1 is the total number of time series; Seq i Lane i Traffic status time series; KSeq j For lane j Considering the time series of traffic status with time lag; Assume lane i , the sequence of its traffic status changing over time is marked as Seq i ;lane i The congestion condition after time t k Later it affects the lane section j , construct the traffic status time series KSeq considering time lag j According to TLCC (Seq i ,KSeq j ) calculates the traffic state correlation coefficient between road sections; when the time lag mutual correlation degree of two road sections, i.e., TLCC(Seq1, NSeq2), exceeds the preset threshold r, it is determined that there is a causal relationship between lane1 and lane2, which is a group of traffic state associated road sections. The corresponding road sections with Pearson correlation coefficient>0.6 are selected to construct the associated road section matrix; (3) Local node betweenness refers to the frequency with which a node appears on the shortest path between all pairs of nodes. In the local node betweenness for congested road sections, the local node betweenness is defined as the frequency with which a node appears on the shortest path between all pairs of nodes passing through the congested node. Among them, σ stm is the number of shortest paths from node s to node t via congested node m; σ stm (i) is the number of nodes that the path passes through; (4) Node degree and node clustering coefficient: The degree k of the i-th node i Represents the total number of edges connected to the node; for the i-th node, the clustering coefficient C i The number of relationships E between the i-th node and its neighboring nodes i and the degree k of the i-th node i The relationship between: Among them, E i is the relationship number of the neighboring nodes of the i-th node, obtained by counting the triangles of the i-th node; k i is the degree of the ith node.

5. The method for dividing the congestion pre-control range based on the importance of local nodes according to claim 4, characterized in that: The specific steps of step (4) are as follows: (401) constructing a judgment matrix B, wherein the values ​​in the matrix represent the relative importance between the node evaluation indicators of each local node importance; (402) The judgment matrix B in the hierarchical analysis method is calculated by normalization, and the sum of its j-th column is expressed as sum j , the element b in the i-th row and j-th column of the matrix ij The normalized value n ij The calculation is as follows: n ij =b ij / sum j The normalized matrix is ​​represented as N, and each element n ij That is the element b of the original judgment matrix ij The results of column normalization are as follows: (403) For the normalized matrix N, a vector is obtained by summing up the rows. Let vector S represent the result of summing up the normalized matrix N by rows, and each component in S is the sum of the elements of the corresponding row of the matrix; (404) Normalize S to obtain the eigenvector ω = (0.16, 0.05, 0.26, 0.26, 0.26) T , ω is the eigenvector of matrix B, find the maximum eigenvalue λ of the eigenvector max , the calculation formula is as follows: Among them, λ max is the maximum eigenvalue of the judgment matrix; n is the order of the judgment matrix, ω i is an element in the eigenvector ω; (405) A consistency check is performed to evaluate the consistency of the judgment matrix. If the consistency of the judgment matrix is ​​not accepted, the elements of the matrix are re-evaluated. The formula is as follows: Among them: CI < 0.1, the judgment matrix is ​​considered to have acceptable consistency; CI ≥ 0.1, the consistency is considered insufficient, and the judgment matrix is ​​readjusted; (406) According to the results calculated by the hierarchical analysis method, the corresponding weight coefficients of each eigenvalue index are ω = (0.16, 0.05, 0.26, 0.26, 0.26) T , calculate the local node importance evaluation matrix R according to the following formula: The characteristic index values ​​of each node and their corresponding weights are comprehensively considered, and the importance of each node is quantitatively scored to obtain the local node importance evaluation value dataset X.

6. The method for dividing the congestion pre-control range based on the importance of local nodes according to claim 5, characterized in that: The specific steps of step (5) are as follows: (501) Randomly select a sample point from the local node importance evaluation value data set X as the first center point; (502) Calculate the shortest distance D(x) from each point to the current center point, and determine the probability P(x) of selecting the next center point based on this distance. The farther the distance, the higher the probability of the point being selected as the new center. (503) Iteratively update the center point until k initial cluster centers are selected; (504) Within each cluster, calculate an actual sample point x of the cluster s The sum of the squared distances to all points in the cluster, select the point with the smallest sum of squared distances to determine cluster E l The new cluster center u l ; Among them, x s is the candidate new cluster center, x is the sample point of the cluster, dist(x s , x) is x s The distance between and x; (505) Based on these selected cluster centers, the distance between each sample and all cluster centers is calculated, and it is assigned to the cluster with the closest distance, so that the distance from each class of sample point x to the cluster center μ l The SSE is the smallest, and the SSE calculation formula is as follows: (406) Repeat steps (404) and (405) until the cluster center is stable or the maximum number of iterations is reached; (407) According to the silhouette coefficient and the minimum sum of squared errors, the optimal number of clusters is determined to be 3. For the pre-control range initially obtained, the levels are divided based on the node importance values.

Images