CN110598942A

CN110598942A - Community public transport network and departure frequency synchronous optimization method considering area full coverage for connecting subways

Info

Publication number: CN110598942A
Application number: CN201910880061.4A
Authority: CN
Inventors: 熊杰; 陈彪; 陈艳艳
Original assignee: Beijing University of Technology
Current assignee: Beijing University of Technology
Priority date: 2019-09-18
Filing date: 2019-09-18
Publication date: 2019-12-20
Anticipated expiration: 2039-09-18
Also published as: CN110598942B

Abstract

The invention discloses a community bus network and departure frequency synchronous optimization method for a connection subway considering area full coverage, and belongs to the field of public transport network planning. The invention divides the research area into a plurality of latticed cells, takes the minimization of the total cost as the optimization target, and also considers the problems of the layout of the micro-circulation public transportation network of the community and the setting of the departure frequency of the line, thereby constructing a mixed integer optimization model. In the solving process, the invention provides a series of constraint conditions based on reality, such as: line length constraints, net total service area constraints, vehicle capacity, fleet size constraints, and the like. The invention designs a plurality of crossover operators and mutation operators to optimize in a network scheme population; the invention also provides an adaptive weight adjustment algorithm, which can adjust the probability of executing the next generation of operators according to the execution condition of the genetic operators so as to improve the execution efficiency of the genetic algorithms.

Description

Community public transport network and departure frequency synchronous optimization method considering area full coverage for connecting subways

Technical Field

The invention belongs to the field of public transport network planning, and relates to a community public transport network and departure frequency synchronous optimization method for a connection subway considering area full coverage. The invention is suitable for optimizing the public transport network in medium and small-scale road network areas.

Background

With the development of urban traffic, conventional buses play an important role in urban passenger transport, but due to the limitation of vehicle types, the bus type is difficult to well play a connection function in an area with sparsely distributed subway stations (particularly in suburbs), and the micro-circulation buses in the community can more flexibly run in the area, so that more convenient and faster service can be provided for passengers. In the traveling process, passengers may need to transfer for many times to travel in a public transport-subway mixed transportation mode, and for better coordinating with the subway to complete passenger transport tasks, a bus operator needs to continuously adjust a microcirculation public transport network.

Due to the lack of sufficient basic theory support, the current microcirculation public traffic line network arrangement method also mainly depends on the experience of planners. Therefore, the invention aims to design an optimal community microcirculation public transportation network from a theoretical level, so that the optimal community microcirculation public transportation network plays an important role in subway connection, and simultaneously realizes the advantage complementation of two public transportation modes, reduces the total cost of a public transportation system and improves the total operation benefit.

Disclosure of Invention

A community bus net and departure frequency synchronous optimization method considering area full coverage for connecting subways comprises the following specific steps:

the method comprises the following steps: optimization model construction

Based on realistic constraint conditions, corresponding penalty functions are provided for the line length, the vehicle load and the service level, and a mixed integer optimization model is constructed by taking the minimization of the total cost of the system as an optimization target.

To more fully describe this problem, the present invention proposes the following constraints:

constraint 1: and (4) restricting the full coverage of the cell. In order to improve the accessibility of passengers in the community for traveling, the service range of the public traffic network needs to cover all the square cells. That is, each grid cell is within the maximum walking distance tolerable to passengers, i.e., each cell is within the service range of at least one bus stop or station in the network, and if the station or station closest to the cell exceeds the tolerable maximum walking distance, the tolerable maximum walking distance is regarded as the distance between the cell and the station.

Constraint 2: and (5) full coverage constraint of the subway station. Subway stations in a research area may not be all distributed on one subway line, and each node in the ST appears in the line network at least once to ensure the connection of two public transportation modes. Meanwhile, all nodes in S do not necessarily all appear in the public transportation network.

Constraint 3: total fleet size constraints. For a community, the public transportation line is often operated by the same operator, and the total fleet size of the public transportation network operation cannot exceed a given upper limit due to limited capital investment of the operator.

Constraint 4: vehicle capacity limitations. On each line, the section traffic volume per trip must not exceed the maximum traffic volume of the operating vehicles, which requires that each line must be set to a minimum departure frequency to ensure the corresponding service level.

Constraint 5: a line length constraint. According to the scale of the road network, the length of each line should be within a reasonable range, which is one of the most common constraints in the research.

Constraint 6: and (5) line trend constraint. There cannot be duplicate stations in each line to avoid a turn-around or local loop in the line.

For ease of study, the following assumptions were made: (1) the study area is divided into several equal sized square cells as in fig. 1. The public transportation requirements of each district are different and are uniformly distributed in the district; (2) the starting point/end point of each OD trip is at the center of the cell, and the end point/starting point is at a subway station (in reality, the trip demand among bus stations in a suburb community is negligibly small, especially in peak hours); (3) regular bus speed and station stop time are constant, since planning is the focus of the study, not operation; (4) for convenience of calculation, the walking distance for passenger transfer is 0; (5) all bus stations and subway stations are provided with conditions for regular bus turning; (6) node 0 represents a bus stop where passengers cannot take a ride or transfer, while all other stops can provide ride and transfer services.

The objective function of the present invention mainly includes user cost and operation cost, and can be expressed as:

C_T＝C_U+C_S (1)

in the formula, C_TIs the total cost; c_UThe cost for the user; c_SIs the operating cost.

The user cost can be divided into three parts: cost of passenger walking C_APassenger cost in-vehicle C_IPassenger waiting cost C_W:

C_U＝C_A+C_I+C_W (2)

In the formula, C_A、C_IBoth with respect to the net layout and the path selected by the passenger, they can be expressed as:

in the formula, Z is a square cell set; path_odIs a selectable path set between o and d; d_od ^kThe walking distance from the cell o to the corresponding getting-on/off station on the kth path between o and d; l is_od ^kIs the on-vehicle distance in the kth path between o and d; t is t_sDelaying the time for the station; ns_od ^kThe number of sites of the kth path between o and d; q. q.s_odThe number of passengers between o and d (or d and o); p is a radical of_od ^kProbability of selecting the kth path for the passenger; v_A、V_BThe walking speed and the vehicle speed of the passenger; mu.s_A、μ_IThe walk time value and the vehicle time value.

In formula (3), D_od ^kIndicating the walking distance from the area o to a certain bus stop on the k-th route between o and dAnd the station is within the maximum walking distance range that the residents of the cell can tolerate. If there are a plurality of stations on the route within the range, the residents are matched according to the minimum sum of the walking time and the in-vehicle time. In particular, for a cell close to a subway station, D may be within the range of cell o, which means that the resident will walk directly to the subway station instead of riding a regular bus, for which case the cell is matched directly to the subway station, D_od ^kRepresenting the walking distance from o to d.

The average waiting cost of passengers on a route is mainly related to the departure interval of the related line. That is, if the route passes through two or more lines, the waiting time for transfer is also included in C_WIn (1). Thus, C_WCan be expressed as:

in the formula, R is a line set of a line network scheme; ns_rThe number of bus stations/subway stations (except bus stations) at which regular buses need to stop in one-way driving of the vehicles; HB_rThe departure interval of the line r; nb_s ^rThe number of passengers boarding route r at station s; z is a radical of_wA ratio of arrival time/departure interval desired for the passenger; mu.s_WThe value of waiting time is obtained.

The operating cost is related to the net layout and departure interval for each line, which is equal to the total fleet size multiplied by the bus operating cost, and the fleet size for each line can be further expressed as equation (7).

In the formula, B is the operation cost of the vehicle per hour; f_rIs the fleet size of line r. L is_rIs the unidirectional length of the line r.

A decision variable 0-1 is also determined, and if node j is a node subsequent to node i in line r, j belongs to SST, R belongs to R; if not, then,thus, L_r、ns_rCan be further expressed as:

in the formula I_ij ^rIs the distance on line r from i to j.

In the above equation, x_ij ^r、HB_rAre the decision variables that need to be solved. For a given net plan and departure frequency set, F_r、L_r、ns_rCan be directly obtained by calculation; and D_od ^k、L_od ^k、p_od ^k、Nb_s ^rNeeds to be determined by the passenger flow allocation method.

Because the line length repairing method can not ensure that each unqualified line can meet the length limit through repairing, a line length punishment is made, C_P1As follows:

in the formula, mu_P1Are coefficients.

Considering the vehicle capacity constraint as a flexible constraint, an overload penalty is set. Before calculating the penalty, a penalty function is defined with respect to the vehicle capacity, g (x):

g(x)＝ω(α^x-1) (11)

in the formula, x is the cross section passenger flow volume; and omega and alpha are coefficients. By adjusting ω and α, g (x) approaches 0 when x is 0. ltoreq. P, and x approaches 0>P, rises rapidly. Overload penalty based on equation (21), net scheme, C_P2Expressed as follows:

in the formula, Ld_ij ^rDetermining the section passenger flow quantity between the nodes i and j in the line r per hour according to the passenger flow distribution result; mu.s_P2Are coefficients.

Under each net scheme, the bus service may not meet the travel requirement of each part, that is, there is no route from the starting point to the end point in the current net, or too many transfers are needed in the route (when the route needs more than two transfers, the current travel is regarded as being out of service). Thus, another penalty is built for demand that cannot be serviced, C_P3As follows:

in the equation, if the current net can serve the travel demand, δ_od1, otherwise, δ_od＝0；μ_P3Are coefficients.

C is to be_P1、C_P2、C_P3Added to the initial objective function, the objective function can be expressed as:

min C_T'＝C_A+C_I+C_W+C_S+C_P1+C_P2+C_P3 (14)

step two: net initialization

And generating a net scheme set meeting all the constraint conditions in the model by using a net initialization algorithm, wherein the net scheme set also comprises a sub-algorithm, namely a line length repairing algorithm.

The randomly generated net schemes can all meet the constraint condition. For the constraint of the line length, a length repair algorithm is provided, so that the length of each line is in a reasonable range. When the area full coverage constraint is realized, a corresponding algorithm is designed, namely, a new line is repeatedly generated to serve the current farthest uncovered area until all areas are covered. Finally, the number of subway stations is checked to ensure that all stations are connected into the network.

For the cell Z belonging to Z, the bus station S belonging to S and wd_zsDenote the walking distance from z to s, let wd_min ^z＝min{wd_zs| S ∈ S }, Z0 denotes an uncovered cell, let wd be_max＝max{wd_min ^z| Z ∈ Z0}, ST0 denotes the set of subway stations not connected to the net, NS denotes the currently generated net solution, WD denotes the maximum walking distance tolerable for the passenger, and the matrix RN denotes the connectivity of every two nodes of the road topology based network.

The steps of randomly generating a feasible net scheme are as follows:

in the above algorithm, the shortest path between two nodes always needs to be found (steps 4, 7, 12), and to solve this problem, the Dijkstra algorithm is repeatedly called, and the reason why the shortest path is taken as a path between two nodes is as follows: 1) the generated line is connected with some key nodes, namely target nodes which want to join the network; 2) there is no additional detour on the line and the line length generally does not exceed an upper limit, so line length repair-1 is primarily used to increase the length of shorter lines.

For line r, the length repair algorithm uses two ways to increase the length of the line: 1) for L_r<L_minThe length of the line meets the requirement by extending the line; 2) for L_r≥L_minAccording to a probability p associated with the current line length_insInserting some nodes into the line r:

the detailed steps are as follows:

the invention also provides another line length repairing method, which aims to shorten the longer line, namely line length repairing-2.

For line r, line length repair-2 shortens the line in two ways: 1) for L_r>L_maxThe shortest path between the nodes is used for replacing the sub-line, so that the line length is shortened; 2) for L_r≤L_maxAccording to a probability p associated with the current line length_remDeleting nodes from r:

and p_insDifferent from, p_remThis operation results in longer line lengths, calculated by exponential functions, rather than linear functions, because a net consisting of a large number of short lines cannot be considered a good net solution.

The first line shortening method is that starting from a first node, detecting whether a sub-line between the node and other nodes is the shortest path, if not, replacing the sub-line with the shortest path; the second method for shortening the line is to detect whether the front and rear nodes of each node except the nodes at the two ends of the line can be directly connected (assuming that the triangle inequality rule is effective), and if so, deleting the middle node. Comparing the two methods, it was found that if a line cannot be shortened by the first method, it cannot be shortened by the latter, and that a line which cannot be shortened by the latter method can be tried to be shortened using the first method. Therefore, when L is_r>L_maxThen, the first method is used to shorten the line, and the detailed steps are described as follows:

line length repair-2 is not suitable for the net initialization process because the initial line is generated mainly using the shortest path algorithm and each method of shortening the line may cause the loss of critical nodes on the line. However, it can be applied to the crossover operator, as will be described in detail in section 6.

Two line length repair heuristic algorithms ensure that each line is within a proper length range to the maximum extent, but the connectivity of a receiving network and the L_min，L_maxAnd (4) calculating the penalty of the line length if the line still can not meet the length constraint, and adding the penalty into the objective function.

Step three: fitness evaluation

Determining alternative routes for passenger travel between each pair of ODs based on a public transit network, constructing a Logistic model taking walking time and vehicle time as effect functions, calculating the proportion of passenger flow on each route between the ODs, determining route departure frequency, providing a fleet scale adjustment algorithm to meet the limitation of the maximum fleet, and finally calculating an objective function value of a scheme.

Path_odAnd (4) representing the path set with the least transfer between the o and the d, and determining the boarding station and the transfer station of the passenger through the matching relation between the path and the public traffic network, so that the cross-section passenger flow of each cross section on each line is obtained through calculation. Finding out all paths between each pair of ODs by using the method in graph theory, and judging the transfer times of each Path so as to determine Path_od。

The current problem is how to quickly find all paths with the least number of transfers between each pair of ODs to determine the Path_od. In determining Path_odIn the process of (1), firstly, whether the cell o is in the service range of the subway station d is detected, if yes, the Path between o and d is the shortest walking Path, and the shortest walking Path is recorded in the Path_odSuch a demand does not require a regular bus ride (scenario 0); otherwise, whether a riding path without transfer exists between each pair of ODs (scene 1), a riding path with one transfer (scene 2), a riding path with two transfers (scene 3) is detected respectively,if there is no path not exceeding two transfers (scenario 4), this OD inter-pair demand is considered as an unserviceable demand, and the corresponding C will be calculated_P3And added to the objective function as shown in fig. 1. In fig. 1, there is no case where two scenarios occur simultaneously, and the occurrence of each scenario will terminate the algorithm.

In scenarios 1-3, Path_odThe generation mechanism of (2) is similar, and the most significant difference between the three scenarios is that the number of cycles in the algorithm grows exponentially. In three scenarios, in order to remove too many detours and save the calculation time, the invention makes a screening strategy: will cost the effect more than Path_odThe minimum value of all paths in the Path is 1.5 times from Path_odIs deleted. Scenario 4 is used as a supplement to other scenarios, the routes of the scenarios exceed two transfers, the travel demand of the OD pair is identified as a non-service demand, and corresponding punishment is calculated.

In scenarios 2 and 3, the initial travel demand needs to be allocated to the bidirectional routes, since the routes are directional, which means that the directions of the sub-routes distributed on different routes are also considered while recording each route transfer station, so as to calculate the number of people getting on and off at each station during the round trip of the vehicle.

Due to Path_odThe transfer time of all the paths is equal, the waiting time is not usually the main factor influencing the path selection of the passengers, and therefore, the proportion of the passengers assigned to each path is assumed to depend on the walking time and the time of the vehicles. Based on the above, the invention constructs a Logistic model taking walking time and vehicle time as effect functions, thereby calculating and obtaining the passenger flow proportion on each path between od:

in the formula, V_od ^kThe effect function for the kth path can be expressed as:

in the formula, a₁，a₂Are coefficients.

The cell o belongs to Z, the subway station d belongs to ST, NS represents the current network scheme, SST_oRepresenting bus and subway stations, TRS, within the maximum walking distance of a cell o_odRepresenting transfer stations for all paths.

Suppose that ns exists in a single pass of line r_rThe round trip of a bus station/subway station (except for a station) can be divided into (2 ns)_r-1) section forNb_s ^rIndicates the number of passengers boarding the vehicle at station s, Na_t ^rThe number of passengers getting off the vehicle at the station t can be expressed as:

in the formula (I), the compound is shown in the specification,all are variables from 0 to 1. If a passenger on the k-th route gets on board at station s on route r, thenIf the k-th path passenger transfers at station s on route r, thenIf the passenger on the k-th route stops at a stop t on route r, thenIf the k-th path passenger transfers at a stop t on line r, then

Passenger flow volume Ld of upper section of line r_ij ^rThe calculation method comprises the following steps:

setting a reasonable departure frequency for a route based on a particular fleet size is an important sub-problem in this study. In order to solve the problem, the invention provides a departure frequency setting method, which aims to determine the optimal departure frequency for each line and mainly comprises two stages: a preliminary calculation phase and a fleet adjustment phase.

In the preliminary calculation stage, for the route r, the waiting cost C of the passenger_W ^rAnd an operating cost C_S ^rCan be expressed as:

the departure interval of each line is in positive correlation with the waiting cost of passengers and in negative correlation with the operation cost, and the total cost C of the line r is solved_T ^rRelative to HB_rThe first derivative of (a) yields:

the minimum C can be found by making the first derivative 0_T ^rAnd corresponding departure interval Hb_rTo give formula (25):

upper limit HM of departure interval_rThe calculation method comprises the following steps:

optimized departure intervals, Hb, of line r_r ^*The calculation method comprises the following steps:

round trip time of the line divided by Hb_r ^*The value of (b) is rounded up to be the number of vehicles needed by the line r, namely the scale of the line fleet, and the calculation method comprises the following steps:

corresponding departure frequency f_r ^*The calculation method comprises the following steps:

for all lines s ∈ R, F_r ^*The sum of (a) may exceed the fleet size upper limit TF, and when this occurs, the fleet size of the lines needs to be adjusted, so that the total fleet size meets the constraint, i.e., the fleet adjustment phase, by reducing the number of vehicles on certain lines. This may lead to some cross-sectional overloads, especially for Hb_r ^*＝HM_rThe line of (2), at which time an overload penalty C needs to be calculated_p2And adds it to the objective function.

One net scheme includes a line set R and a fleet size set F ═ F₁，…F_r，...，F_|R|}，ΔC_T ^rRepresents the total cost change of the route r after removal of the vehicle, which can be expressed as Δ C_T ^r＝ΔC_W ^r+ΔC_S ^r+ΔC_P2 ^r(ΔC_W ^r，ΔC_P2 ^r>0，ΔC_S ^r<0) The algorithm is described in detail as follows:

step four: the design of an optimization algorithm is carried out,

the invention optimizes the line network by using the genetic algorithm, provides a plurality of crossover and mutation operators in the genetic algorithm, and embeds an adaptive operator selection strategy into the genetic algorithm, so that the whole iterative process is more intelligent. In addition, various repair operators are proposed to repair the infeasible schemes obtained after crossover and mutation.

The genetic algorithm optimization framework is shown in fig. 2. At the beginning of each iteration, the selection operator randomly selects a group of individuals to perform the following operation, and the individuals with higher fitness are considered as the better individuals and have higher probability of being selected. Therefore, the inverse of the objective function is taken as the fitness function: f 1/C_T'。

The probability that each individual is selected can be expressed as:

in the formula, p_iProbability of being selected for individual i; f. of_iThe fitness function value of the individual i. Then, using roulette, NIND' individuals (NIND × GGAP) were selected as initial populations, and prepared for crossover and mutation.

The invention provides three kinds of crossover operators, namely site crossover, line crossover and scheme crossover.

In site crossing, Parent1, Parent2 represent two parents for crossing, as shown in fig. 3. The operator first randomly selects a line from the two individuals, respectively, then cuts each line at a single site and swaps subsequences of the two lines to generate two new lines. The purpose of this operator is to explore better site order, changing routes from a microscopic perspective.

This operator may be ineffective when two lines are split and switched at randomly selected sites, and this crossover operation is considered an effective crossover only when at least one feasible line is generated that is different from the other lines, and therefore the emphasis of operator design is on how quickly to find two reasonable split points to accomplish an effective crossover of lines. In order to solve the problem, the invention provides two single-point crossing methods as the sub-heuristic algorithms of the operator, namely single-point crossing-1 and single-point crossing-2.

The single point cross-1 is described in detail as follows:

the single point cross-2 is described in detail as follows:

the loop cancellation algorithm is used to cancel a local loop present in a line, and is described in detail as follows:

to compare the performance of the two intersection methods, the overlap ratio of the two lines is quantified by the ratio of the number of nodes of intersection to the number of nodes of union, and is expressed by equation (31):

in the formula, c_r1,r2 ¹∈[1,1]Indicating the degree of coincidence of r1 and r 2. c. C_r1,r2 ¹A larger value of (a) indicates a higher coincidence degree of the two lines.

Curve of a single point crossing-1 with c_r1,r2 ¹Gradually decreases with increasing; while the curve of the single-point cross-2 has a fitting interval, which can be expressed as f_s，f_e]. In [ f_s，f_e]Inner, single point cross-2 stable performance, but the curve drops sharply outside the interval, f_s，f_eThe value of (c) may vary with the scale of the problem.

In terms of efficiency, single-point cross-2 operates at a slightly faster speed than single-point cross-1 because single-point cross-1 repeatedly runs the shortest path search requiring a large amount of computation time. The detailed steps of the site intersection operator are as follows:

unlike site crossing, line crossing aims to explore different line combinations from the macro level to create a new net scenario. In operator design, a line intersection separates an individual from one or more pairs of split points, which are all the end points of the line, as shown in fig. 4:

in fig. 4, the line crossing operator generates a new individual by swapping the same number of lines from two individuals, the present invention limits the number of swapped lines to:

in the formula, n_sIs the number of lines switched; and | P1|, and | P2| are the number of lines of two individuals.

Notably, after selecting the route r1 from parent generation 1, the route r2 selected from parent generation 2 is not randomly selected. In order to ensure the coverage cell and the number of subway stations connected by the generated new individual to the maximum extent, a route with the maximum correlation coefficient r1 is selected from the parent generation 2 as r2, and the correlation coefficient calculation method comprises the following steps:

in the formula, c² _r1,r2Is the correlation coefficient of r2 and r1, Z_r1 ^*Set of cells covered by the service area of r1, Z_r2Set of cells covered by r2 service area, ST_r1 ^*Set of subway stations connected in r1, ST_r2For the set of subway stations connected in r2, | r2| is the number of nodes of r 2.

The line crossing operator is described in detail as follows:

a solution crossover is to split an individual from a single point and swap subsequences between two parents to form a pair of new individuals. The split points may be end points or intermediate points of the line. This operator is easy to implement and it can explore different line combinations and site sequences simultaneously, as shown in fig. 5.

Similar to a site crossing, if the splitting point in the solution crossing is located in the middle of a line, the operator involves the exchange of the site sequence between two lines, which can be done by the mixed use of single-point crossing-1 and single-point crossing-2, and after the new line is generated, the loop elimination algorithm and line length repair 1/2 need to be performed.

It is noted that the two lines used for crossing are not randomly selected, as shown in fig. 5, the selected lines divide the whole net scheme into two line groups, one before the other, and if the selected line position order is biased to one side, the number of lines included in the two line groups is unbalanced, which will make the descendants include too many or too few lines, which is not favorable for generating a competitive scheme. Therefore, the invention makes a weight coefficient for each line in the net scheme based on the gaussian function:

in the formula, wc_kIs the weight coefficient of line k, where 'k' is the sequential index of the line; the | R | is the number of lines in the line network; ps is_kIs the probability that line k is selected.

The invention provides three mutation operators, namely deletion mutation, insertion mutation and shortening mutation. Deleting a variation is to change the scheme by deleting a line, and inserting a variation and shortening a variation can change the scheme by changing the node sequence of a line, as shown in fig. 6.

In fig. 6(a), the deletion variant is to remove one line from the parent to generate the child, and the deleted line is the line with the smallest importance coefficient, which can be quantified according to the number of coverage cells and connection sites:

in the formula, c_r ³Is the importance coefficient of the line r; i Z_r ^*L is the number of cells covered in the service area of the line r, | ST_r ^*And | is the number of subway stations connected by the line r. The line selection strategy reduces the damage to individual feasibility caused by line deletion, and the purpose of the mutation operator is to stabilize the number of lines in the scheme, because new lines can be added to the scheme during the scheme repair process.

The insertion mutation operator selects a line R (L) from the individual_R≤L_max) And attempts to insert a node (which is not present in the line R and can simultaneously connect adjacent nodes (bus stations or subway stations) in the line R) into the line R, as shown in fig. 6 (b). If all nodes can not be inserted into the line R, replacing one line and repeating the insertion operation until the nodes are successfully inserted or the line is selectedThe number reaches the maximum number of attempts max rs1, wherein | NS_fL to satisfy the line length constraint (L)_R≤L_max) The detailed steps are as follows:

the probability of success of inserting a node using the shortest path search method is higher than the probability of success by detecting the connectivity of the node with the node in the line. When searching for the shortest path, some unexpected nodes may be introduced into the line, as shown in fig. 6(b), and node 3 is the unexpected node introduced when node 4 is inserted, and since the line has a local loop, node 4 is deleted later, and node 3 is still retained. This allows this to happen since the purpose of this operation is to accomplish as efficient an insert operation as possible without the need to insert the specified node.

As shown in FIG. 6(c), the shortening mutation is to select a line R (L) from the individuals_R≥L_min) And attempts to replace a particular sequence of nodes with the shortest path. If the entire route is the shortest route from the start point to the end point, the process is repeated for a route replacement until the route replacement is successfully performed or the maximum number of attempts for route selection max _ rs2 is reached. Similarly, letWherein | NS_fI satisfies the length constraint (L) in individuals_R≥L_min) The number of lines of (2).

In the scheme after the crossover and mutation operators, infeasible lines still exist, and the infeasibility is mainly expressed in the following aspects:

(1) two or more identical lines exist in the line network scheme;

(2) there are areas not covered by the current line service area;

(3) not all subway stations are connected to the network.

In order to ensure the feasibility of the scheme, the invention provides a repair operator, which comprises the following detailed steps:

the repair algorithm detects the condition (1) first, and eliminates the repeated line (step 2); then detecting the condition (2), inserting some sites which can expand the service range of the net into the net scheme (step 4-7); similarly, the algorithm inserts unconnected subway stations into the current net (steps 8-10). It should be noted that finally, if there are still uncovered cells or unconnected subway stations, new lines will be generated to ensure the feasibility of the solution.

The basic idea of the adaptive operator selection strategy is: the score is given to each previous stage operator, and measures the performance of the operators so that the probability of selecting crossover and mutation operators is adaptively adjusted in the next stage according to their previous performance.

In the design of this strategy, the whole iterative process is divided into several stages, each stage being defined as a certain algebra. Each stage is the basic unit for updating the score of an operator according to the performance of the operator, and each stage can be further divided into multiple iterations of intersection and variation (and intersection probability P)_cAnd the mutation probability P_mCorrelated), that is to say, at the end of each phase, each newly generated individual is detected according to the different cases shown in table 1 and passes through σ₁-σ₆And scoring each operator.

TABLE 1 fraction adjustment parameters and description

When the operator is given the adjustment parameter, two aspects are mainly considered: one is the objective function value of the generation scheme; the other is whether it will generate a new wire net scheme. This strategy is designed to use as many operators as possible that can improve the solution, promoting solution diversity. If a new solution is generated after the crossover operator and mutation operator are executed, the two operators will get the same score because there is no way to distinguish how effective the two operators are for generating the new solution.

The weight calculation method of the operator i at the next stage comprises the following steps:

in the formula, w_ijFor the weights of algorithm i in run phase j,w_i1＝0；λ∈[0,1]a coefficient for adjusting the recent expression change reaction speed of an operator; beta is a_ijIs the total fraction, gamma, obtained after the operating phase j for the operator i_ijThe number of times operator i is used at the end of run phase j. At the beginning of the run phase j +1, the beta values of all operators are set_i,j+1，γ_i,j+1Reset to 0. In the operating phase j +1, the probability that the operator is selected is:

in the formula, p_i,j+1For the probability that operator i is selected in operating phase j +1,p_i,1＝1/3。

to avoid premature trapping of genetic algorithms into suboptimal, the present invention proposes an evolutionary stagnation detection strategy, i.e. if the optimal solution does not improve significantly in multiple successive operational phases, the population initialization algorithm is used to update the selected population at the beginning of the next operational phase to ensure that genetic algorithms can search for optimized net solutions in a wider solution space.

Step five: wire mesh optimization scheme

The method comprises the steps of optimizing a practical public traffic network by using a genetic algorithm of a plurality of intersection and mutation operators and a self-adaptive operator selection strategy combination, obtaining an optimized public traffic network scheme, and providing information such as the trend, departure interval and fleet scale of each line in the network scheme.

Drawings

FIG. 1 is a graph of path conversion discrimination.

FIG. 2 is a flow chart of a genetic algorithm.

Fig. 3 is a schematic cross-site diagram.

Fig. 4 is a schematic cross-line diagram.

FIG. 5 is a schematic cross-sectional view of the protocol.

FIG. 6 is a schematic diagram of a mutation operator.

Fig. 7 studies regional road networks and cell maps.

Fig. 8 studies a regional topology network map.

Fig. 9 studies traffic zone division and travel demand of a zone.

Figure 10 current bus net.

Figure 11 optimizes a bus network scheme.

FIG. 12 is a flow chart of an embodiment of the method of the present invention.

Detailed Description

The real case study area is the Tokyo Community in Beijing, which has an area of about 4.8km × 2.2km and is located in the suburban area of Beijing, as shown in FIG. 7. The area is divided into 264 grid cells, the side length of each cell is 200 meters, 106 nodes are totally arranged in a road network, the nodes comprise 1 station (node 0), 55 bus stations (nodes 1-55) and 6 subway stations (nodes 56-61), and the topological network is shown in fig. 8.

In order to measure and calculate the travel demand between each grid cell and each subway station, the public transportation IC card data is used for identifying a 'public transportation-subway' combined travel chain of each bus station in an early peak period in a research area, then the demand at the station is distributed to the cells covered by a service range according to a certain proportion, the proportion can be determined by the total travel demand in the current traffic cell, and the calculation formulas (39) - (40) are given out for the demand of each grid cell.

In the formula, D_o,dThe public transportation trip demand (bidirectional demand) from the grid cell o to the subway station d; SD_s,dThe current bus demand from the bus stop s to the subway stop d is obtained through the IC card data; s_oA station set within the maximum walking distance range of the grid cell o; TD_oThe total travel demands of various traffic modes in the grid cell o are obtained; epsilon_s,zAs a coefficient, if the grid cell z is within the walking distance of the station s, then e_s,z1, otherwise, ∈_s,z＝0；TTD_kThe total travel demand of various traffic modes in the traffic cell k is obtained through the mobile phone signaling data; eta_o,kThe ratio of the overlapping area of the grid cell o and the traffic cell k to the area of the traffic cell k.

Traffic cell division in the research area is shown in fig. 9, the number in each cell is the number of the traffic cell, and the demand of all traffic modes in each traffic cell is obtained through mobile phone signaling data.

Currently, there are seven lines serving the community during peak hours, each of which is connected to at least one subway station, as shown in fig. 10. Lines 101 and 102 are private lines serving community passengers, while the other lines are public lines that only partially pass through the area under study.

In calculating the fleet size and current operating costs, a factor defined by L '/L is considered, L representing the total link length and L' representing the length of the portion of the link within the study area, and the information for each link is shown in table 2.

Table 2 present information per line

^aConverted fleet size, F ═ F × (L'/L);

^bthe operating cost of the conversion, CS ═ CS × (L'/L).

The method comprises the steps that various costs and punishments are measured and calculated according to a given passenger flow distribution method, the converted fleet scale and the operation cost can be used for evaluating the current network scheme, and in the process of solving the optimization scheme, the converted fleet scale is used as the upper limit of the fleet scale, namely TF-57 veh. Other parameter settings in the model: v_A＝5.8km/h；V_B＝25km/h；t_s1/90 h; p is 60 persons; l is_min＝5km；L_max14 km; $ 100/hour; mu.s_A$ 10/hour; mu.s_W$ 10/hour; mu.s_I$ 5/hour; mu.s_P1500 yuan/km; mu.s_P2＝1；μ_P3$ 10/person; z is a radical of_w＝0.5，g(x)＝0.01×(<0.001). The parameters of the genetic algorithm are set as: NIND-20; MAXGEN 1000; p_c＝0.9；P_m0.1; the length of the operation stage is 20 generations; sigma₁＝8，σ₂＝6，σ₃＝3，σ₄＝2，σ₅＝1，σ₆1 is ═ 1; λ is 0.5. For simplicity, the present disclosure treats a demand that requires more than two transfers as an unserviceable demand.

The data in table 3 are the costs and penalties of the current net and the costs and penalties of the optimized net.

TABLE 3 optimization scheme information

^*The experiment was repeated 5 times to obtain results;

^arun the number of detected protocols 5 times;

as shown in table 3, the optimized net scheme is significantly better than the existing net scheme, and the total cost of the searched net scheme is reduced by 57.09%. In practical application, by redesigning the community public transportation network under the scale of the existing fleet, the travel cost of passengers can be reduced by 62.8%, and the operation cost can be reduced by 18.3%. The optimization scheme is shown in table 4, and the route trend is shown in fig. 11.

TABLE 4 optimization scheme

Claims

1. A community public transport network and departure frequency synchronous optimization method for a connection subway considering area full coverage is characterized by comprising the following steps:

step one, constructing a bus network optimization model

Providing a constraint condition based on reality, a line length constraint, a net total service range constraint, a vehicle capacity and a fleet scale constraint, taking the total cost, namely the minimum value of user cost and operation cost, as an optimization target, giving consideration to community microcirculation public traffic net layout and line departure frequency setting, and constructing a mixed integer optimization model;

step two, initializing the public traffic network

For the constraint of the line length, a length repair algorithm is provided, so that the length of each line is in a reasonable range; when the area full coverage constraint is realized, designing a corresponding algorithm, namely, repeatedly generating a new line to serve the current farthest uncovered area until all areas are covered; finally, detecting the number of the subway stations to ensure that all the subway stations are connected to the network;

step three, evaluating the fitness of the wire mesh

Fitness evaluation is used for evaluating the overall performance of a wire mesh scheme and comprises two parts: passenger flow distribution and fleet scale setting; the passenger flow distribution aims at distributing travel demands in the grid cells to bus stops or route sets of a bus network, so that the getting-on and getting-off passenger flow of each stop and the section passenger flow of each line are obtained; the motorcade size determination method is that a given number of vehicles are accurately distributed to each line under the condition of considering the total motorcade size and the vehicle capacity; finally, the fitness of the net scheme, namely the quality degree, is reflected by calculating a model objective function value;

step four, optimizing algorithm design

Optimizing the network scheme by using a genetic algorithm, and designing a plurality of crossover operators, mutation operators and repair operators to ensure that a new feasible scheme is obtained based on the original scheme;

step five, network optimization scheme

Based on the characteristics of different algorithm combinations, the final bus network optimization scheme is obtained by using the algorithm combination with the best performance.

2. The method for optimizing the community public transportation network and departure frequency of the connected subway in consideration of the full coverage of the area as claimed in claim 1, wherein said step one,

considering various constraint conditions based on reality, providing corresponding penalty functions for the line length, the vehicle load and the service level, and constructing a mixed integer optimization model by taking the minimization of the total system cost as an optimization target;

the proposed constraints are as follows:

constraint 1: cell full coverage constraint; in order to improve the accessibility of passengers in the community during traveling, the service range of a public traffic network needs to cover all grid cells;

constraint 2: the full coverage of the subway station is restricted; in the network, each subway station appears at least once to ensure the connection of two public transportation modes;

constraint 3: total fleet scale constraints; the fleet size of the line network cannot exceed the upper limit of the total fleet size;

constraint 4: a vehicle capacity limit; on each line, the section passenger flow of each journey cannot exceed the maximum passenger capacity of the operating vehicle;

constraint 5: line length constraints; the length of each line should be within a reasonable range;

constraint 6: line trend constraint; repeated stations cannot exist in each line, so that a turning back or a local loop is avoided;

the objective function contains user cost and operation cost, and is expressed as:

C_T＝C_U+C_S (1)

in the formula, C_TIs the total cost; c_UThe cost for the user; c_SIs the operating cost;

the user cost is divided into three parts: cost of passenger walking C_APassenger cost in-vehicle C_IPassenger waiting cost C_W:

C_U＝C_A+C_I+C_W (2)

In the formula, C_A、C_IBoth related to the net layout and the passenger selected route are represented as:

in the formula, Z is a square cell set; path_odIs a selectable path set between o and d; d_od ^kThe walking distance from the cell o to the corresponding getting-on/off station on the kth path between o and d; l is_od ^kIs the on-vehicle distance in the kth path between o and d; t is t_sDelaying the time for the station; ns_od ^kThe number of sites of the kth path between o and d; q. q.s_odThe number of passengers between o and d (or d and o); p is a radical of_od ^kProbability of selecting the kth path for the passenger; v_A、V_BThe walking speed and the vehicle speed of the passenger; mu.s_A、μ_ITo the walk time value and the time-on-vehicle value;

in formula (3), D_od ^kOn the k-th path between o and dThe walking distance from the cell o to a bus stop within the maximum walking distance range tolerable by the residents of the cell; if a plurality of stations exist on the path within the range, matching the residents according to the minimum sum of the walking time and the time in the vehicle; for a cell close to a subway station, D may be within the range of cell o, meaning that the resident will walk directly to the subway station instead of riding a regular bus, matching the cell directly to the subway station, D_od ^kRepresents the walking distance from o to d;

the average waiting cost of passengers on one path is related to the departure interval of the related line; c_WExpressed as:

in the formula, R is a line set of a line network scheme; ns_rIn the one-way running of the vehicles, the number of bus stations/subway stations needing stopping of regular buses is except for bus stations; HB_rThe departure interval of the line r; nb_s ^rThe number of passengers boarding route r at station s; z is a radical of_wA ratio of arrival time/departure interval desired for the passenger; mu.s_WThe value of waiting time is obtained;

the operation cost is related to the net layout and the departure interval of each line, the operation cost is equal to the total fleet size multiplied by the bus operation cost, and the fleet size of each line is further expressed as formula (7);

in the formula, B is the operation cost of the vehicle per hour; f_rIs the fleet size of line r; l is_rIs the unidirectional length of the line r;

0-1 decision variable if node j is subsequent to node i in line rWhen node, x_ij ^r＝1，j belongs to SST, R belongs to R; otherwise, x_ij ^r＝0；L_r、ns_rExpressed as:

in the formula I_ij ^rIs the distance on line r from i to j;

line length penalty C_P1The calculation method is as follows:

in the formula, mu_P1Is a coefficient;

penalty function for vehicle capacity, g (x):

g(x)＝ω(α^x-1) (11)

in the formula, x is the cross section passenger flow volume; omega and alpha are coefficients;

overload penalty C for net scheme_P2The calculation method is as follows:

in the formula, Ld_ij ^rDetermining the section passenger flow quantity between the nodes i and j in the line r per hour according to the passenger flow distribution result; mu.s_P2Is a coefficient;

unserviceable demand penalty C_P3The calculation method is as follows:

where if the current net serves the travel demand, δ_od1, otherwise, δ_od＝0；μ_P3Is a coefficient;

c is to be_P1、C_P2、C_P3Added to the initial objective function, which is expressed as:

min C_T'＝C_A+C_I+C_W+C_S+C_P1+C_P2+C_P3 (14)。

3. the method for optimizing the community public transportation network and departure frequency of the connected subway in consideration of the full coverage of the area as claimed in claim 1, wherein in the second step,

generating a net scheme set meeting all constraint conditions in the model by utilizing a net initialization algorithm, wherein the net scheme set also comprises a sub-algorithm, namely a line length repairing algorithm;

all the randomly generated wire net schemes can meet constraint conditions;

for the cell Z belonging to Z, the bus station S belonging to S and wd_zsDenote the walking distance from z to s, let wd_min ^z＝min{wd_zs| S ∈ S }, Z0 denotes an uncovered cell, let wd be_max＝max{wd_min ^z-Z ∈ Z0}, ST0 denotes the set of subway stations not connected to the wire network, NS denotes the currently generated wire network solution, WD denotes the maximum walking distance tolerable for the passenger, and the matrix RN denotes the connectivity of every two nodes of the road topology based network; the steps of randomly generating a feasible net scheme are as follows:

for line r, the length repair algorithm uses two ways to increase the length of the line: 1) to pairIn L_r<L_minThe length of the line meets the requirement by extending the line; 2) for L_r≥L_minAccording to a probability p associated with the current line length_insInserting some nodes into the line r:

the detailed steps are as follows:

the detailed steps are described as follows:

two line length repair heuristic algorithms ensure that each line is within a proper length range to the maximum extent, but the connectivity of a receiving network and the L_min，L_maxAnd (4) influence of the value, if the line cannot meet the length constraint, calculating the line length penalty, and adding the line length penalty into the objective function.

4. The method for optimizing the community public transportation network and departure frequency of the connected subway in consideration of the full coverage of the area as claimed in claim 1, wherein said step three,

determining alternative paths for passenger travel between each pair of ODs based on a public traffic network, constructing a Logistic model taking walking time and vehicle time as effect functions, calculating the proportion of passenger flow on each path between the ODs, determining route departure frequency, providing a fleet scale adjustment algorithm to meet the limitation of the maximum fleet, and finally, calculating a target function value of a scheme;

Path_odrepresenting the path set with the least transfer between the o and the d, and determining the boarding station and the transfer station of the passenger through the matching relation between the path and the public traffic network, thereby calculating the cross-section passenger flow of each cross section on each line; finding out all paths between each pair of ODs by using the method in graph theory, and judging the transfer times of each Path so as to determine Path_od；

In determining Path_odIn the process of (1), firstly, whether the cell o is in the service range of the subway station d is detected, if yes, the Path between o and d is the shortest walking Path, and the shortest walking Path is recorded in the Path_odSuch a demand does not require a regular bus ride — scenario 0; otherwise, whether a riding path without transfer, namely scene 1, a riding path with one transfer, namely scene 2, and a riding path with two transfers, namely scene 3 exist between each pair of ODs is respectively detected, if the path without more than two transfers, namely scene 4, does not exist, the demand between the OD pairs is regarded as the non-service demand, and the corresponding C is calculated_P3And adding it to the objective function; moreover, there is no case where two scenarios occur simultaneously, and the occurrence of each scenario will terminate the algorithm;

in scenarios 1-3, Path_odThe generation mechanism of (2) is similar, and the most obvious difference among the three scenes is that the cycle number in the algorithm increases exponentially; in three scenarios, in order to remove too many detours, save the calculation time, a screening strategy is formulated: will cost the effect more than Path_odThe minimum value of all paths in the Path is 1.5 times from Path_odDeleting; scene 4 is used as a supplement of other scenes, the paths of the scenes exceed twice transfer, the travel demand of the OD pair is identified as the unserviceable demand, and corresponding punishment is calculated；

In scenarios 2 and 3, the initial travel demand needs to be allocated to the bidirectional routes, and since the routes are directional, while recording the transfer station of each route, the directions of the sub-routes distributed on different routes need to be considered, so as to calculate the number of people getting on or off at each station during the round trip of the vehicle;

constructing a Logistic model taking walking time and vehicle time as effect functions, and calculating to obtain the passenger flow proportion on each path between the od:

in the formula, V_od ^kThe effect function for the kth path is expressed as:

in the formula, a₁，a₂Is a coefficient;

suppose that ns exists in a single pass of line r_rThe round trip of a bus station/subway station (except for a station) can be divided into (2 ns)_r-1) section forNb_s ^rIndicates the number of passengers boarding the station s (see equation (5)),the number of passengers getting off the vehicle at the station t can be expressed as:

in the formula (I), the compound is shown in the specification,all are variables from 0 to 1; if a passenger on the k-th route gets on board at station s on route r, thenIf the k-th path passenger transfers at station s on route r, thenIf the passenger on the k-th route stops at a stop t on route r, thenIf the k-th path passenger transfers at a stop t on line r, then

a departure frequency setting method, which aims to determine the optimal departure frequency for each line, comprises two stages: a preliminary calculation stage and a motorcade adjustment stage;

in the preliminary calculation stage, the departure interval Hb of the line r_r ^*The calculation method comprises the following steps:

Hb_rthe calculation method comprises the following steps:

HM_rthe calculation method comprises the following steps:

divide the round trip time of the line by Hb_r ^*The value of (b) is rounded up to be the number of vehicles needed by the line r, namely the fleet scale of the line, and the calculation method comprises the following steps:

frequency f of departure_r ^*The calculation method comprises the following steps:

for all lines s ∈ R, F_r ^*The sum of (2) may exceed the fleet scale upper limit TF, when this occurs, the fleet scale of the line needs to be adjusted, and the total fleet scale meets the constraint condition by reducing the number of vehicles of some lines, i.e. the fleet adjustment stage;

5. the method for optimizing the community public transportation network and departure frequency of the connected subway in consideration of the full coverage of the area as claimed in claim 1, wherein said step four,

optimizing the line network by using a genetic algorithm, providing a plurality of crossover and mutation operators in the genetic algorithm, and embedding a self-adaptive operator selection strategy into the genetic algorithm, so that the whole iterative process is more intelligent; in addition, a plurality of repair operators are provided to repair the infeasible schemes obtained after intersection and mutation;

when each iteration starts, the genetic algorithm randomly selects a group of individuals to perform the following operations, the individuals with higher fitness are considered as the better individuals, and the selected probability is higher; taking the inverse of the objective function as the fitness function: f 1/C_T'；

The probability that each individual is selected can be expressed as:

in the formula, p_iProbability of being selected for individual i; f. of_iA fitness function value of the individual i; then, selecting NIND' individuals (NIND multiplied by GGAP) individuals to form an initial population by using a roulette method, and preparing for crossing and mutation;

three kinds of crossover operators are provided, namely site crossover, line crossover and scheme crossover;

in site crossing, a line is randomly selected from two individuals respectively, then each line is cut at a single site, and subsequences of the two lines are exchanged to generate two new lines; the operator aims at exploring a better site sequence and changing a line from a microscopic angle;

when two lines are split and switched at randomly selected sites, the operator may be invalid, and this crossover operation is considered to be an effective crossover only when at least one feasible line different from the other lines is generated, so the operator design is focused on how to quickly find two reasonable split points to complete an effective crossover of lines; in order to solve the problem, two single-point crossing methods are provided as the sub-heuristic algorithms of the operator, namely single-point crossing-1 and single-point crossing-2;

single point cross-1, as follows:

single point cross-2, as follows:

the loop cancellation algorithm is used to cancel the local loop present in the line, as follows:

the site intersection operator comprises two single-point intersection operators, as follows:

unlike site crossing, line crossing aims at exploring different line combinations from the macro level to generate a new line network scheme; in operator design, line crossing separates an individual from one or more pairs of split points, which are all end points of a line, and a new individual is generated by exchanging the same number of lines from two individuals; the number of lines switched is limited to:

in the formula, n_sIs the number of lines switched; p1 and P2 are two individualsThe number of lines;

it should be noted that, when selecting a route, after selecting the route r1 from the parent1, the route r2 selected from the parent2 is not randomly selected, but a route with the largest correlation coefficient with r1 is selected from the parent2 as r2, and the correlation coefficient can be expressed as:

in the formula, c² _r1,r2Is the correlation coefficient of r2 and r1, Z_r1 ^*Set of cells covered by the service area of r1, Z_r2Set of cells covered by r2 service area, ST_r1 ^*Set of subway stations connected in r1, ST_r2The number of nodes is r2, | r2| is a subway station set connected in r 2;

the operation of the line crossing operator is described as follows:

a solution crossover is to split an individual from a single point and exchange subsequences between two parents to form a pair of new individuals, the split points can be end points or intermediate points of a line, which can simultaneously explore different line combinations and site sequences; similar to the site crossing, if the splitting point in the solution crossing is located in the middle of the line, the operator includes the exchange of site sequences between the two lines, the operation can be completed by the mixed use of the single-point crossing-1 and the single-point crossing-2, and after a new line is generated, a loop elimination algorithm and line length repair 1/2 need to be executed;

the two lines in the solution intersection are not randomly selected, because the selected lines divide the whole net solution into two line groups, one before the other, and if the position sequence of the selected lines is biased to one side, the number of lines included in the two line groups is not balanced, which will make the filial generation include too many or too few lines, which is not favorable for generating a competitive solution; therefore, a weighting factor is formulated for each line in the net scheme based on the gaussian function:

in the formula, wc_kIs the weight coefficient of line k, where 'k' is the sequential index of the line; the | R | is the number of lines in the line network; ps is_kProbability of being selected for line k;

three mutation operators are provided, namely deletion mutation, insertion mutation and shortening mutation; deletion of a variation is to change the scheme by deleting a line, and insertion of a variation and shortening of a variation can change the scheme by changing the node sequence of a line;

deletion variant is to remove a line from a parent to generate a child, the deleted line being the line with the smallest importance coefficient, which can be quantified according to the number of coverage cells and connection sites:

in the formula, c_r ³Is the importance coefficient of the line r; i Z_r ^*L is the number of cells covered in the service area of the line r, | ST_r ^*L is the number of subway stations connected by the line r;

the insertion mutation operator selects a line R (L) from the individual_R≤L_max) And try to insert a node (not present in the line R and capable of connecting adjacent nodes (bus stations or subway stations) in the line R at the same time) into the line R; if all nodes cannot be inserted into the line R, the insertion operation is repeated by replacing one line until the node is successfully inserted or the number of line selections reaches the maximum number of attempts max _ rs1,wherein | NS_fL to satisfy the line length constraint (L)_R≤L_max) The detailed steps are as follows:

the shortened variants are selected from the individuals for the line R (L)_R≥L_min) And attempting to replace the particular sequence of nodes with the shortest path; if the whole line is the shortest path from the starting point to the end point, replacing one line and repeating the process until the path replacement is successfully realized or the maximum attempt time max _ rs2 of line selection is reached; similarly, letWherein | NS_fI satisfies the length constraint (L) in individuals_R≥L_min) The number of lines of (a);

the crossover and mutation operations result in unfeasible lines, which are expressed in the following aspects:

(1) two or more identical lines exist in the line network scheme;

(2) there are areas not covered by the current line service area;

(3) not all subway stations are connected into a wire network;

in order to ensure the feasibility of the scheme, a repair operator is provided, and the detailed steps are as follows:

the repair algorithm detects the condition (1) first, and eliminates the repeated line (step 2); then detecting the condition (2), inserting some sites which can expand the service range of the net into the net scheme (step 4-7); similarly, the algorithm inserts unconnected subway stations into the current net (steps 8-10); if there are still uncovered cells or unconnected subway stations, a new line is generated and added to the scheme (step 11);

the basic idea of the adaptive operator selection strategy is to score the scores of operators in each previous stage, the scores measuring the performance of the operators, so that the probability of selecting each pair of crossover and mutation operators is adaptively adjusted in the next stage iteration according to their previous performance;

in the design of this strategy, the whole iterative process is divided into several stages, each stage being defined as a certain algebra; each stage is a basic unit for updating the score of an operator according to the performance of the operator, and each stage can be further divided into multiple iterations of intersection and variation (and intersection probability P)_cAnd the mutation probability P_mCorrelated), that is to say, at the end of each phase, each newly generated individual is detected according to the different cases shown in table 1 and passes through σ₁-σ₆Scoring each operator;

TABLE 1 fraction adjustment parameters and description

The operator is endowed with an adjusting parameter, and two aspects are considered: one is the objective function value of the generation scheme; the other is whether it will generate a new wire net scheme;

the operator i weight of the next stage is updated as:

in the formula, w_ijFor the weights of algorithm i in run phase j,w_i1＝0；λ∈[0,1]a coefficient for adjusting the recent expression change reaction speed of an operator; beta is a_ijIs the total fraction, gamma, obtained after the operating phase j for the operator i_ijThe number of times operator i is used at the end of run phase j; at the beginning of the run phase j +1, the beta values of all operators are set_i,j+1，γ_i,j+1Reset to 0; in the operating phase j +1, the operatorThe probability of being selected is:

6. the method for optimizing the community public transportation network and departure frequency of the connected subway in consideration of the full coverage of the area as claimed in claim 1, wherein said step five,

a practical public transport network is optimized by using a genetic algorithm of a plurality of intersection and mutation operators and a self-adaptive operator selection strategy combination, an optimized public transport network scheme is obtained, and the trend, departure interval and fleet scale information of each line in the network scheme are given.