Optimization method of coordinated freight train flow allocation between technical stations based on the neighborhood search method of arrival and departure flow continuity difference
The neighborhood search algorithm optimizes freight train flow distribution by addressing uncertainty in disassembly times and flexible departure times, enhancing the reliability and efficiency of train dispatch in railway stations.
Patent Information
- Application Number
- CN202211022034.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-24
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-08-24
AI Technical Summary
Existing methods for railway freight train flow distribution at technical stations fail to account for the uncertainty in train disassembly times and the flexibility of departure times, leading to inefficiencies and potential train delays, especially in stations with abundant railway transport capacity.
A method utilizing a neighborhood search algorithm that considers the randomness of train disassembly times and flexible departure times, constructing a random chance-constrained programming model to optimize the coordinated flow distribution between adjacent technical stations, adjusting train disassembly and departure sequences to minimize vehicle stay time and ensure timely train dispatch.
The method reduces vehicle stay time by 86.96 hours and increases the number of successfully dispatched trains by 2, providing a highly reliable distribution scheme that converges to the optimal solution quickly without manual iteration adjustments.
Smart Images

Figure CN115438845B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of railway transportation, and in particular to an inter-station freight train coordinated flow allocation optimization method based on a neighborhood search method for the difference between arrival and departure train flows. Background Art
[0002] Railway technical stations are important nodes in the organization of traffic flow. A large number of transit traffic flows are reorganized and transferred here. The quality of their work organization will largely affect the transportation efficiency of the entire railway freight network. Railway traffic flow organization is a collaborative operation of multiple units. The flow distribution results of a technical station will inevitably affect the flow distribution operations of the next technical station. At the same time, the traffic flow continuation plan of the technical station will inevitably change due to some uncertain information during the execution process, such as train arrival and departure times, de-marshalling operation time, train formation composition, etc. These factors will affect the specific implementation of the traffic flow continuation plan. Therefore, it is particularly important to consider the impact of these uncertain factors to improve the reliability of the traffic flow continuation plan of the technical station.
[0003] At present, many scholars at home and abroad have studied the preparation of technical station traffic flow continuation plans. Niu Huimin et al. considered the various operating capabilities of the marshaling yard and the number of tracks in the marshaling yard, and took the vehicle's running distance, assembly time, and exchange vehicle costs as the objective function to comprehensively optimize the traffic flow continuation of the two-way marshaling station. Wang Ciguang transformed the flow distribution problem into a transportation problem and solved it using the table operation method. Zhao Jun et al. constructed a mixed integer programming model for the flow distribution and shunting operation of a one-way marshaling station based on the parallel machine scheduling and resource allocation theory. Boysen et al. studied the train marshaling problem and expanded the problem model by considering the use of shunting yard tracks and different full axle conditions of trains. During this period, many scholars also studied the preparation of robust traffic flow continuation plans for technical stations. He Shiwei et al. pointed out that in actual transportation organization work, the conditions such as the number of full axles of departure trains, departure time, and station technical operation time are all fuzzy interval numbers. At the same time, these concepts were fuzzily defined, and a single-machine fuzzy mixed 0-1 programming model and a multi-machine fuzzy mixed 0-1 programming model for stage planning were constructed. Subsequently, scholars from his research team conducted research on specific issues. Li Haodong et al. considered the randomness of train arrival times, constructed a random-dependent opportunity planning model for stage planning, and compiled a scheduling plan with high robustness based on it. In addition, the train traction quality and length change constraints were also considered at the same time, and different weights were assigned according to different train levels, which effectively avoided the phenomenon of train suspension due to a single constraint, alleviated the traffic congestion and shortened the station time; Liu Ting et al. considered the randomness of train disassembly time, and assumed that the time obeyed the normal distribution, and used the opportunity constraint planning method to clarify the disassembly time, thereby converting it into a stage planning problem under deterministic conditions; Jing Yun et al. regarded the disassembly time as a fuzzy variable, and used the pessimistic value of the variable to represent the disassembly operation time under a certain confidence level, and then studied it, and used the non-deterministic tree search algorithm to give the optimal disassembly solution for the flow distribution problem. Ma Liang considered the uncertain abnormal events that occurred during the execution of the stage plan, and established a dynamic adjustment model for the robustness of the marshaling station stage based on the minimum perturbation dynamic adjustment theory. Xue Feng et al. comprehensively considered information such as train delays and traffic status, and used coupling to measure the impact of actual arriving traffic on station scheduling. By optimizing and adjusting the planned traffic sources, they reduced the impact of traffic fluctuations on station operations. There are relatively few studies on robust traffic flow continuation plans for technical stations abroad, mainly because train delays are not common in stations with abundant railway transport capacity in Europe and the United States, and usually occur mostly in busier stations in Europe.Representative studies include: Bühsing et al. considered the uncertainty of train arrival times, transformed the flow allocation problem into a sorting problem, and obtained a flow allocation solution that is robust to train arrival delays by inserting additional sorting steps; Brady et al. used random and robust modeling techniques to reduce the allocation conflicts of adjacent trains on the same track to achieve robust optimization of train arrival delays; Cicerone et al. considered the uncertain situations that train sets may encounter on the shunting yard tracks during the flow allocation process, including three types of uncertain conditions: train set parking position, train set composition, and shunting yard track failures, designed an efficient robust optimization algorithm and gave corresponding solutions.
[0004] Most of the above studies are based on a single technical station, and the train departure time is considered a rigid constraint. In fact, the departure time constraints of my country's railway freight trains are flexible to a certain extent. For example, if the operation diagram or daily schedule stipulates that a train departs at 13:25, it is acceptable to delay it to 13:26. At the same time, delaying the train departure time within an acceptable range will also allow more traffic to meet the connection time constraints. Summary of the invention
[0005] In order to solve the problems existing in the prior art, the purpose of the present invention is to provide a method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method of the connection difference between arrival and departure train flows. The present invention takes into account the uncertainty of the unmarshaling operation time, and at the same time gives a certain flexibility to the train departure time, studies the problem of coordinated flow distribution of freight trains between adjacent technical stations, and designs an efficient neighborhood search algorithm based on the connection difference between arrival and departure train flows according to the characteristics of the flow distribution problem, which provides a feasible way for the rapid preparation of train flow connection plans, thereby improving the fulfillment rate of train flow connection plans.
[0006] To achieve the above object, the technical solution adopted by the present invention is: a method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method of the difference between arrival and departure flow, comprising:
[0007] Model construction: Set the objective function and construct a random chance constrained programming model between adjacent technical stations that takes into account the randomness of train disassembly time and flexible constraints on departure times;
[0008] Model solution: transform the opportunity constraints into deterministic equivalence classes, and then realize the coordinated flow distribution of adjacent technical stations based on the neighborhood search algorithm of the connection difference between the departure and arrival flows.
[0009] As a further improvement of the present invention, the objective function is as follows:
[0010] minG=G1+G2 (1)
[0011]
[0012]
[0013] As a further improvement of the present invention, the constraints of the constraint programming model include: train disassembly time constraints, train formation time constraints, vehicle flow connection constraints, train full axle constraints, coordinated flow distribution constraints and departure train formation content adjustment constraints.
[0014] As a further improvement of the present invention, the constraint conditions are specifically as follows:
[0015] (A) Train dismantling time constraints:
[0016] For the disintegration order of Y h The earliest dismantling time of the arriving train is the end time of the train inspection, and the actual dismantling time of the train shall not be earlier than the end time of the train inspection. Since the train dismantling time is a random variable subject to normal distribution, the chance constraint (6) is established with a certain probability α. At the same time, the actual dismantling time of the train must also be earlier than the end time of the dismantling of the previous dismantling train:
[0017]
[0018]
[0019]
[0020]
[0021] (B) Train marshaling time constraints:
[0022] For grouping order Z l For a departing train, the latest marshaling end time is the departure time of the train minus the departure technical operation inspection time, and the actual marshaling end time of the train shall not be later than its latest marshaling end time; since the train marshaling time is a random variable that obeys the normal distribution, the chance constraint (10) is established with a certain probability β. At the same time, the actual marshaling end time of the train must also be earlier than the start marshaling time of the next train:
[0023]
[0024]
[0025]
[0026]
[0027] (C) Traffic flow constraints:
[0028] ① In a single technical station, if the arriving train i can provide a connecting train flow for the departing train j, the marshaling start time of the departing train j must be greater than the disassembly end time of the arriving train i:
[0029]
[0030] Where M is a large constant;
[0031] ② The number of trains with the vehicle group number k provided by the arriving train i to the departing train cannot exceed the number of corresponding vehicle groups contained in the arriving train:
[0032]
[0033] ③ The traffic provided by the arriving train i to the departing train j must be the train set included in the j train:
[0034]
[0035] In this formula, M can be taken as the number of cars in the departure train j;
[0036] (D) Train full axle constraint:
[0037] For the departure train j to be successfully dispatched, at least one of the full-axle constraints, full-weight or full-length, must be satisfied;
[0038]
[0039]
[0040]
[0041] (E) Coordinated flow constraints:
[0042] If the departure train j of the previous technical station corresponds to the arrival train i'' of the next technical station and can provide a connecting train flow for the departure train j' of the next technical station, then the following conditions shall be met:
[0043]
[0044] (F) Constraints on adjustment of departure train composition:
[0045] If the departure train j' of the next technical station does not meet the full axle constraint of train departure, and there is a lost train j at the previous technical station, which corresponds to the arrival train i' of the next technical station, and there is a train flow with the same marshaling content as the departure train j' of the next technical station in the unsuccessful traffic flow of the previous technical station, then consider adjusting the marshaling content of the lost train at the previous technical station, and the adjustment should meet the following requirements:
[0046] Before the content of the j marshaling of the train departing from the previous technical station is adjusted:
[0047]
[0048] After the content of the j marshaling of the train departing from the previous technical station is adjusted:
[0049]
[0050]
[0051]
[0052]
[0053]
[0054]
[0055]
[0056]
[0057] Wherein, O represents a technical station, where O=1 is the previous technical station, O=2 is the next technical station, i is the arriving train, i=1,2,…,n; j is the departing train, j=1,2,…,m; is the arrival time of train i at technical station O; is the scheduled departure time of the departure train j; The end time of the stage of technical station O; is the arrival technical operation time of technical station O; The departure technical operation time of technical station O; is the dismantling operation time of the arriving train i at the technical station O; is the marshaling operation time of the departure train j from the technical station O; Y1, Y2, Y h , …, Y n is the train disassembly sequence, corresponding to the train arrival sequence i = 1, 2, ..., n; Z1, Z2, Z l , …, Z m is the train marshaling sequence, corresponding to a permutation of the train departure sequence j = 1, 2, …, m; is the deviation of the departure time of train j from technical station O from the on-time schedule. A positive value indicates a late departure, while a negative value indicates an early departure of the train. is the earliest disassembly start time of the arriving train i at the technical station O; is the actual disassembly start time of the arriving train i at the technical station O; is the disassembly end time of the arriving train i at the technical station O; is the marshaling start time of the departure train j from the technical station O; is the latest marshaling end time of the departure train j from the technical station O; is the actual marshaling end time of the departure train j at the technical station O; is a 0-1 variable. When the arriving train i in the technical station O can provide a connecting train flow for the departing train j, its value is 1, otherwise it is 0; k represents the train flow group number, and K is the set of all train flow group numbers. represents the number of vehicles with the vehicle flow group number k provided by the arriving train i at the technical station O to the departing train j; represents the total number of vehicles with vehicle group number k in the arriving train i at technical station O; It is a 0-1 variable, indicating whether the departure train j contains the destination of the k-th train. If so, it is 1, otherwise it is 0; It is a 0-1 variable, indicating whether the departure train j contains the destination of k' after adjusting the marshaling content. If it does, it is 1, otherwise it is 0; and The maximum and minimum number of vehicles corresponding to the departure train j of technical station O; and is the maximum and minimum weight change corresponding to the departure train j of technical station O; w k Indicates the corresponding weight change of the vehicle group with vehicle group number k; It is a 0-1 variable, indicating whether the train j departing from technical station O meets the minimum number of trains required; is a 0-1 variable, indicating whether train j departing from technical station O meets the minimum weight change requirement; is a 0-1 variable, indicating whether the train j from technical station O is successfully dispatched, and is 1; L CD is the distance between adjacent technical stations; v is the running speed of the freight train.
[0058] As a further improvement of the present invention, the neighborhood search algorithm based on the difference in the connection between the arrival and departure traffic flows specifically includes the following steps:
[0059] Step 1: Adjust the departure train value based on the train arrival and departure times Based on the time of various technical operations, the ideal arrival and departure vehicle flow matrix of two adjacent technical stations is calculated;
[0060] Step 2: Select the initial solution, use the current train arrival and departure sequence of each technical station as the train disassembly sequence, and perform flow distribution to obtain the initial solution Z. Let the initial Z best =Z,Z now = Z, the train arrival and departure sequence is taken as the initial optimal disassembly sequence, the current optimal disassembly sequence;
[0061] Step 3: Calculate the actual arrival and departure traffic flow matrix and arrival and departure traffic flow difference matrix under the optimal disassembly sequence, and find all positions greater than 0 in the arrival and departure traffic flow difference matrix to generate a candidate set of disassembly adjustment points U. At the same time, u1 is the first adjustment point in the candidate set, and the total number of adjustment points is N. TZ ;
[0062] Step 4: Neighborhood construction: select adjustment points u from the unmarshaled adjustment candidate set U in sequence i , each time only one of the points is constructed for the neighborhood, initially i = 1, the selected disassembly adjustment point is calculated, and the disassembly order adjustment corresponding to the arrival train at this point under the optimal disassembly order is obtained, and the result is a column vector with n rows and 1 column, and the marshaling time delay corresponding to the marshaling order adjustment corresponding to the departure train at this point is calculated, and the result is a row vector with 1 row and m columns; at the same time, the connection time difference CZ between the arrival train and the departure train corresponding to the disassembly adjustment point is calculated, and the value is negative, and all suitable train disassembly order adjustment values that satisfy the change value just greater than the absolute value of the train connection time difference at this point are found from the time change matrix, thereby generating the domain of the current solution;
[0063] Step 5: Update the current solution. Compare the flow distribution results of all neighboring solutions with the current solution in turn. If the neighboring solution result is better than the current solution, assign the neighboring solution result to the current solution Z. now , the corresponding solution order is used as the current solution order. If no neighbor solution result is better than the current solution, the current result remains unchanged;
[0064] Step 6: If i<N TZ , let i=i+1, return to step 4, if i=N TZ , judge Z now and the previous optimal solution Z best Size, if Z now <Z best , let Z best =Z now , the optimal solution sequence is updated to Z now Corresponding to the decompilation order, return to step 3; otherwise, output Z best , Z best Corresponding to the train disassembly sequence, the technical station flow distribution results are obtained;
[0065] Step 7: Based on the flow distribution results of a single technical station, optimize the trains that failed to be dispatched from the next station, consider the coordinated flow distribution constraints and the train departure time adjustment constraints, and determine whether the departure time and marshaling content of the train departing from the previous station can be adjusted to achieve coordinated flow distribution between the two technical stations. If not, output the separate flow distribution results of the two technical stations and the algorithm ends; if so, calculate the corresponding train departure adjustment time. The departure train formation is adjusted, and the algorithm returns to step 1 to recalculate and obtain the coordinated flow distribution result.
[0066] The beneficial effects of the present invention are:
[0067] The present invention uses the opportunity constrained programming method to study the problem of coordinated flow distribution of freight trains between adjacent technical stations considering the randomness of train disassembly time and the flexible constraints of departure time. Taking the minimum vehicle stay time at the station as the goal, an opportunity constrained programming model for coordinated flow distribution between adjacent technical stations is constructed. Starting from the essence of the train disassembly sequence adjustment, the vehicle flow connection difference matrix between the ideal connecting vehicle flow and the actual connecting vehicle flow is calculated. According to the connection time difference between trains corresponding to the vehicle flow connection difference, the neighborhood is constructed in combination with the time variation corresponding to the train disassembly sequence adjustment, and the neighborhood search algorithm is designed for solution. The example results show that considering the flexibility of train departure time and the coordinated optimization between stations, the total stay time of vehicles at the two technical stations is reduced by 86.96h, and the number of successfully dispatched trains increases by 2; compared with the existing intelligent algorithm, the algorithm does not need to manually set the number of iterations, can avoid invalid search, converge to the optimal solution in a short time, and obtain a highly reliable distribution scheme. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 This is a schematic diagram of the location of the technical station in the embodiment of the present invention;
[0069] Figure 2 This is an example diagram of the redundant waiting time for train disassembly in an embodiment of the present invention;
[0070] Figure 3 This is an example diagram of the train marshaling advance time in an embodiment of the present invention;
[0071] Figure 4 This is a graph showing the change of the vehicle's stay time at the station versus the number of iterations before the technical station D is adjusted in an embodiment of the present invention;
[0072] Figure 5 This is a graph showing the variation of the Now solution with the number of iterations before the technical station D is adjusted in an embodiment of the present invention;
[0073] Figure 6 This is a graph showing the change of the best solution with the number of iterations before the technical station D is adjusted in an embodiment of the present invention. DETAILED DESCRIPTION
[0074] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0075] Example
[0076] like Figure 1As shown in the figure, in the technical station, the trains that need to be reorganized must go through the five basic operation processes of "arrival, disassembly, assembly, compilation, and departure" in sequence. Among them, the arrival technical operation time and the departure technical operation time are relatively stable, while the disassembly operation time has a large uncertainty, especially the disassembly operation, which will directly affect the assembly of freight cars. If it is simply regarded as a fixed value, it is very likely that the compiled traffic flow connection plan will be difficult to fully implement. At the same time, most existing studies regard the departure time of the train as a fixed value. If the departure time of the train is adjusted within the acceptable range of scheduling, the lost line train that originally could not meet the traffic flow connection time constraint may meet the constraint condition and thus depart normally. In a single technical station, the adjustment of the train departure time only has the situation of giving a certain delay to the departure time of the lost line train. In adjacent technical stations, there are many possibilities for adjusting the departure time of the departure train, such as Figure 1 Take the example to explain this situation in detail, considering the coordinated flow distribution of stations C and D:
[0077] (1) When the train departure time remains unchanged, the trains departing from Technical Station C and Technical Station D can complete the dispatching smoothly, and the train departure time is not adjusted at this time.
[0078] (2) When the train departure time remains unchanged, only the lost train exists at technical station D. Meanwhile, some trains from technical station C can provide connecting traffic for it in the direction, but the connecting time is insufficient. At this time, try to advance the departure time of these trains at technical station C, while maintaining or delaying the departure time of the lost train at technical station D, so that the lost train at station D can be successfully allocated. If the departure time adjustment value of the two technical stations reaches the maximum and still fails, the departure time adjustment will not be performed.
[0079] (3) When the train departure time remains unchanged, there are departure trains that have failed to be assigned at both technical stations C and D. Consider the following four situations:
[0080] ① The lost train at technical station C cannot provide follow-up traffic for the lost train at technical station D. This situation is the same as (2);
[0081] ② The train successfully dispatched by technical station C cannot provide follow-up traffic for the lost line train of technical station D, that is, the only train that can provide follow-up traffic for the lost line train of station D is the lost line train of technical station C. At this time, try to give the lost line train of technical station C a certain departure delay to make its traffic allocation successful, thereby providing follow-up traffic for technical station D.
[0082] ③ The lost line train of Technical Station C may provide follow-up traffic for the lost line train of Technical Station D by delaying its departure time. At the same time, some trains successfully dispatched from Technical Station C may also provide follow-up traffic for the lost line train of Technical Station D by advancing its departure time. At this time, try to comprehensively adjust the departure times of the two types of trains at Technical Station C, so that the lost line train of Technical Station D can be successfully allocated.
[0083] ④ It is impossible for the lost train and the successfully dispatched train at Technical Station C to provide follow-up traffic for Technical Station D by adjusting the departure train time. At this time, no departure time adjustment is made.
[0084] It should be noted that the adjustment of the departure time of the Technical Station C train is only for the trains that can provide connecting traffic for the lost trains at the Technical Station D and have insufficient connecting time. No adjustment will be made for the trains that can originally provide connecting traffic for the Technical Station D.
[0085] Accordingly, this embodiment provides a method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method for the difference between arrival and departure train flows, including:
[0086] Model construction:
[0087] Since the time of train unbundling operation at technical stations is random, chance-constrained programming is an effective method to describe and solve uncertainty optimization. Therefore, this embodiment takes adjacent technical stations C and D as examples to construct a random chance-constrained programming model that considers the randomness of train unbundling time and the flexible constraints of departure time.
[0088] Model assumptions and symbol description:
[0089] (1) Model assumptions:
[0090] ① Assume that both adjacent technical stations adopt the single-push and single-slide operation mode;
[0091] ② Assume that the two technical stations have sufficient train inspection equipment and personnel to meet the requirements of on-demand inspection;
[0092] ③ Assume that each yard in the two technical stations can meet the requirements of train reception and dispatching and shunting operations;
[0093] ④ Regardless of the fault, preparation and maintenance status of the shunting machine, the shunting machine can only dismantle or marshal one train at any time, and once the operation starts, it will not be stopped;
[0094] ⑤ The acceptable train departure time adjustment limit for dispatching at the two technical stations is set to 5min.
[0095] (2) Model parameters and symbols:
[0096] O represents a technical station, where O=1 is technical station C and O=2 is technical station D. i is an arriving train, i=1,2,…,n; j is a departing train, j=1,2,…,m; is the arrival time of train i at technical station O; is the scheduled departure time of the departure train j; The end time of the stage of technical station O; is the arrival technical operation time of technical station O; The departure technical operation time of technical station O; is the dismantling operation time of the arriving train i at the technical station O; is the marshaling operation time of the departure train j from the technical station O; Y1, Y2, Y h , …, Y n is the train disassembly sequence, corresponding to the train arrival sequence i = 1, 2, ..., n; Z1, Z2, Z l , …, Z m is the train marshaling sequence, corresponding to a permutation of the train departure sequence j = 1, 2, …, m; is the deviation of the departure time of train j from technical station O from the on-time schedule. A positive value indicates a late departure, while a negative value indicates an early departure of the train. is the earliest disassembly start time of the arriving train i at the technical station O; is the actual disassembly start time of the arriving train i at the technical station O; is the disassembly end time of the arriving train i at the technical station O; is the marshaling start time of the departure train j from the technical station O; is the latest marshaling end time of the departure train j from the technical station O; is the actual marshaling end time of the departure train j at the technical station O; is a 0-1 variable. When the arriving train i in the technical station O can provide a connecting train flow for the departing train j, its value is 1, otherwise it is 0; k represents the train flow group number, and K is the set of all train flow group numbers. represents the number of vehicles with the vehicle flow group number k provided by the arriving train i at the technical station O to the departing train j; represents the total number of vehicles with vehicle group number k in the arriving train i at technical station O; It is a 0-1 variable, indicating whether the departure train j contains the destination of the k-th train. If so, it is 1, otherwise it is 0; It is a 0-1 variable, indicating whether the departure train j contains the destination of k' after adjusting the marshaling content. If it does, it is 1, otherwise it is 0; and The maximum and minimum number of vehicles corresponding to the departure train j of technical station O; and is the maximum and minimum weight change corresponding to the departure train j of technical station O; w k Indicates the corresponding weight change of the vehicle group with vehicle group number k; It is a 0-1 variable, indicating whether the train j departing from technical station O meets the minimum number of trains required; is a 0-1 variable, indicating whether train j departing from technical station O meets the minimum weight change requirement; is a 0-1 variable, indicating whether the train j from technical station O is successfully dispatched, and is 1; L CD is the distance between technical stations C and D; v is the running speed of the freight train.
[0097] The parameters and symbols involved in the model are shown in Table 1:
[0098] Table 1 Model parameters and symbols
[0099]
[0100]
[0101] Objective function:
[0102] Since the train departure time adjustment is taken into account, if the minimum vehicle stay time at the station is directly used as the objective function, most trains may depart early, which is obviously unreasonable. This embodiment aims to adjust the departure time of the train that may provide the connecting traffic according to the departure train loss situation of the technical station D. For this purpose, the objective function is set as follows:
[0103] minG=G1+G2 (1)
[0104]
[0105]
[0106] Constraints:
[0107] (1) Train dismantling time constraints:
[0108] For the disintegration order of Y h The earliest dismantling time of the arriving train is the end time of the train inspection, and the actual dismantling time of the train shall not be earlier than the end time of the train inspection. Since the train dismantling time is a random variable that obeys the normal distribution, the chance constraint (6) is established with a certain probability α. At the same time, the actual dismantling time of the train must also be earlier than the end time of the dismantling of the previous dismantling train.
[0109]
[0110]
[0111]
[0112]
[0113] (2) Train marshaling time constraints:
[0114] For grouping order Z l The latest marshaling end time of a departing train is the departure time of the train minus the departure technical operation inspection time, and the actual marshaling end time of the train shall not be later than its latest marshaling end time. Since the train marshaling time is a random variable that obeys the normal distribution, the chance constraint (10) is established with a certain probability β. At the same time, the actual marshaling end time of the train must also be earlier than the start marshaling time of the next train.
[0115]
[0116]
[0117]
[0118]
[0119] (3) Traffic flow constraints:
[0120] ① Within a single technical station, if the arriving train i can provide a connecting train flow for the departing train j, the marshaling start time of the departing train j must be greater than the disassembly end time of the arriving train i.
[0121]
[0122] Where M is a large constant.
[0123] ② The number of trains with vehicle group number k provided by the arriving train i to the departing train cannot exceed the number of corresponding vehicle groups contained in the arriving train.
[0124]
[0125] ③The traffic provided by the arriving train i to the departing train j must be the train set included in the j train.
[0126]
[0127] In this formula, M can be taken as the number of cars in the departure train j.
[0128] (4) Train full axle constraints:
[0129] For the departure train j to be successfully dispatched, at least one of the full-axle constraints, full weight or full length, must be met.
[0130]
[0131]
[0132]
[0133] (5) Collaborative flow constraints:
[0134] If the departure train j of technical station C corresponds to the arrival train i'' of technical station D and can provide a connecting train flow for the departure train j' of technical station D, then the following conditions should be met:
[0135]
[0136] (6) Constraints on adjustment of departure train composition:
[0137] If the departure train j' of technical station D does not meet the full axle constraint of train departure, and there is a lost line train j at station C, which corresponds to the arrival train i' of technical station D, and there is a train flow with the same composition as the departure train j' of technical station D in the unsuccessful traffic flow at station C, then consider adjusting the composition of the lost line train at technical station C, and the adjustment should meet the requirements.
[0138] Before the content of train j departing from technical station C is adjusted:
[0139]
[0140] After the content of train j departing from technical station C is adjusted:
[0141]
[0142]
[0143]
[0144]
[0145]
[0146]
[0147]
[0148]
[0149] Model solution:
[0150] Chance constraints are transformed into deterministic equivalent classes:
[0151] One method of solving the opportunity constraint programming problem is to convert the opportunity constraint programming into a corresponding deterministic equivalent class, thereby converting it into a deterministic problem for solution. The train disassembly and marshaling operation time in the technical station obeys the normal distribution. Combining equations (6) and (10), only the disassembly time is a random variable in the opportunity constraint formula. Therefore, the opportunity constraint discussed in this embodiment meets the conversion conditions. The opportunity constraint formula in this embodiment is converted into a deterministic equivalent class.
[0152] Neighborhood search algorithm based on the difference in the connection between arrival and departure traffic:
[0153] The core difficulty of the technical station flow distribution problem lies in the adjustment and determination of the disassembly order, and the optimization and adjustment of the train disassembly order is essentially to adjust the train disassembly end time and the assembly start time, so that the train that cannot meet the vehicle flow connection time constraint under the original disassembly order meets the vehicle flow connection constraint. In other words, if there is a possibility of optimizing the train disassembly order, then there must be some vehicle flows that do not meet the vehicle flow connection constraints under the current disassembly order, and after adjustment, these vehicle flows meet the vehicle flow connection constraints. Therefore, if it is possible to find these vehicle flows that can be adjusted through the disassembly order, so as to meet the vehicle flow constraints from being dissatisfied with the vehicle flow connection constraints, they can be adjusted in a targeted manner to achieve the optimization of the train disassembly order. Based on this idea, this embodiment proposes a neighborhood search algorithm based on the difference in the connection between arrival and departure vehicles. To facilitate the description of the algorithm, some concepts of the method are explained as follows:
[0154] Definition 1: Arrival-departure traffic flow connection difference: the difference in the number of vehicles between the ideal arrival-departure traffic flow and the actual arrival-departure traffic flow.
[0155] Definition 2 Ideal arrival and departure connecting traffic flow: the maximum connecting traffic flow value obtained by only considering the arrival and departure times of trains and ignoring the conflicts in train equipment occupancy;
[0156] Definition 3 Actual arrival and departure connecting traffic: the maximum connecting traffic value is obtained by considering the current train disassembly sequence and the conflict of equipment occupancy in the technical station.
[0157] Calculation of the difference between the arrival and departure traffic flow:
[0158] (1) Calculation of ideal continuous traffic flow:
[0159] ①Construct the matrix A of the arriving traffic flow, It represents the total number of vehicles with vehicle group number k among the arriving train i at technical station O, which has the same symbol as above.
[0160]
[0161] ②Construct the departure vehicle flow grouping and destination matrix B, It is a 0-1 variable, indicating whether the departure train j contains the destination of train k. If so, it is 1, otherwise it is 0.
[0162]
[0163] In the above-mentioned arrival traffic composition matrix and departure traffic grouping destination matrix, the traffic group numbers represented by each column have the same direction.
[0164] ③ Without considering the time constraint, the arrival traffic is the maximum traffic value that the departure traffic can provide. Using the linear algebra matrix operation rules, multiply the arrival traffic composition matrix A with the transposed matrix of the departure traffic grouping direction composition matrix B. That is, the maximum number of subsequent traffic matrix CAR_JX that the arrival train can provide for each departure train without considering the time constraint is obtained.
[0165] CAR_JX=AB T
[0166] The matrix That is, the number of vehicles that the i-th arriving train can provide for the j-th departing train, without considering the time connection constraints and only considering the destination of the train.
[0167] ④Construct the ideal disassembly end time matrix of the arriving train
[0168] Without considering equipment occupancy conflicts, the train arrival times are The train arrival technical inspection operation time is Then the corresponding train technical inspection operation end time is This time is also the earliest possible dismantling time of each arriving dismantling train (the earliest possible dismantling time, that is, no matter how the dismantling order of the train is adjusted, the dismantling time will not be earlier than this time), so the earliest possible dismantling end time is also determined, that is, the dismantling time of each arriving train plus the corresponding dismantling time, that is Construct a matrix of n rows and 1 column according to the arrival order of the ideal train disassembly end time. At the same time, construct a 1-row, m-column all-1 matrix, that is, L = [1, 1, ..., 1] m , using the linear algebra matrix operation rules, multiply the two matrices to obtain a matrix M with n rows and m columns LXEJT =M lxejt L.
[0169]
[0170] ⑤Construct the ideal starting time matrix of the departure train
[0171] Without considering equipment occupancy conflicts, the train departure times are The time for technical inspection of train departure is Then the latest departure technical inspection start time for the corresponding train is This time is also the latest possible marshaling end time for each departure train (the latest possible marshaling end, that is, no matter how the marshaling order of the train is adjusted, the marshaling end time will not be later than this time), so the latest possible marshaling start time is also determined, that is, the marshaling end time of each train minus the corresponding marshaling operation time, that is Construct a 1-row and m-column matrix by taking the earliest possible train marshaling start time into account according to the train departure order. At the same time, construct an n-row and 1-column all-1 matrix, that is Using linear algebra and matrix operation rules, multiply the two matrices to obtain a matrix M with n rows and m columns. LXKSBZ =RM lxksbz .
[0172]
[0173] ⑥ Ideal connection time constraint matrix of arrival and departure flow
[0174] M LXKSBZ Matrix minus M LXEJT , get the ideal connection time matrix M LXJX =M LXKSBZ -M LXEJT Matrix M LXJX The value of row i and column j If its value is positive, it means that without considering equipment occupancy conflicts, the i-th arriving train can provide a connecting flow for the j-th departing train. If it is negative, it means that even under ideal conditions, the i-th arriving train cannot provide a connecting flow for the j-th departing train. To calculate the connecting flow, M LXJX Values greater than 0 are set to 1, and values less than 0 are set to 0.
[0175] ⑦ Ideal traffic flow matrix
[0176] If an arriving train can provide a connecting train for a departing train, then it must be satisfied that the arriving train's traffic composition contains traffic with the same marshaling and destination as the departing train, and at the same time, the arriving train and the departing train meet the time connection constraint. To this end, the maximum connecting traffic number matrix CAR_JX is used to connect the ideal arrival and departure traffic connection time constraint matrix M LXJX Multiplication, the nonlinear algebraic operation here refers to: multiplication of two matrices, using point-to-point multiplication, that is, the value of the i-th row and j-th column of the ideal traffic flow continuation matrix is equal to the value of the i-th row and j-th column of the CAR_JX matrix multiplied by M LXJX The result of multiplying the values in row i and column j. LX(ij) = M LXJX (ij)·CAR_JX(ij), where M LXJX The value is either 0 or 1, Car LX (ij) represents the connecting traffic value that the i-th arriving train can provide for the j-th departing train under the constraints of the connecting direction and traffic connecting time.
[0177] (2) Calculation of actual connecting traffic flow:
[0178] Compared with the ideal connecting traffic flow, the actual connecting traffic flow only changes the connection time constraint matrix of the arrival and departure traffic flow, without considering the time constraint. The matrix CAR_JX of the maximum number of connecting traffic flows that the arrival traffic flow can provide for each column of the departure flow does not change. Therefore, the key to calculating the actual connecting traffic flow lies in calculating the actual connection time matrix.
[0179] ①Construct the actual disassembly end time matrix of the arriving train
[0180] Considering equipment occupancy conflicts, the train arrival times are The train arrival technical inspection operation time is Then the corresponding train technical inspection operation end time is However, the actual dismantling time of the train is not the time when the train reaches the end of the technical operation inspection, but the actual dismantling start time calculated according to the current dismantling sequence and the train dismantling time constraint. Will Add the dismantling time of the corresponding train to obtain the actual earliest dismantling end time of each arriving train under the current dismantling order. Construct an n-row and 1-column matrix based on the actual dismantling end time according to the train arrival order. At the same time, construct a 1-row, m-column all-1 matrix, that is, L = [1, 1, ..., 1] m , using linear algebra and matrix operation rules, multiply the two matrices to obtain a matrix M with n rows and m columns SJEJT =M sjejt L.
[0181]
[0182] ②Construct the actual marshaling start time matrix of the departure train
[0183] The train departure times are The time for technical inspection of train departure is Then the latest departure technical inspection start time for the corresponding train is However, the actual latest marshaling end time of the train is not the latest departure technical operation inspection start time of the train, but the actual latest marshaling end time calculated according to the current marshaling sequence and train marshaling time constraints. Subtract the marshaling operation time of the train from the actual marshaling end time to obtain the actual latest marshaling start time of each departure train under the current marshaling sequence. Construct a 1-row and m-column matrix based on the actual latest marshaling start time of the train according to the train departure order. At the same time, construct an n-row and 1-column all-1 matrix, that is Using the linear algebra matrix operation rules, multiply the two matrices to obtain a matrix M with n rows and m columns. SJKSBZ =RM sjksbz .
[0184]
[0185] ③ Actual arrival and departure flow connection time constraint matrix
[0186] Using linear algebra matrix operation rules, M SJKSBZ Matrix minus M SJEJT , we get M SJJX =M SJKSBZ -M SJEJT At this time, the matrix M SJJX The value of the i-th row and j-th column is, If its value is positive, it means that considering the equipment occupation conflict and the current disassembly sequence, the i-th arriving train can provide a connecting flow for the j-th departing train. If it is negative, it means that under the current disassembly sequence, the i-th arriving train cannot provide a connecting flow for the j-th departing train. To calculate the connecting flow, M SJJX Values greater than 0 are set to 1, and values less than 0 are set to 0.
[0187] ④Actual traffic flow matrix
[0188] Similar to the method of calculating the ideal vehicle flow connection matrix, the maximum vehicle flow connection matrix CAR_JX is calculated by using nonlinear algebraic operation rules and the actual arrival and departure vehicle flow connection time constraint matrix M SJJX Multiplication, the nonlinear algebraic operation here refers to: multiplication of two matrices, using point-to-point multiplication, that is, the value of the i-th row and j-th column of the ideal traffic flow continuation matrix is equal to the value of the i-th row and j-th column of the CAR_JX matrix multiplied by M SJJX The result of multiplying the values in the i-th row and j-th column.
[0189] Car SJ (ij) = M SJJX (ij)·CAR_JX(ij), where MSJJX The value is either 0 or 1, Car SJ (ij) represents the connecting traffic value that the i-th arriving train can provide for the j-th departing train under the constraints of the current disassembly sequence, connecting direction and traffic connection time.
[0190] (3) Poor connection between arrival and departure traffic:
[0191] Use the ideal continuous traffic matrix Car LX (ij) minus the actual connecting vehicle flow matrix Car SJ (ij), we can get that the arriving train cannot provide the departing train with the following traffic due to the equipment occupancy and the constraints of the unmarshaling order when the unmarshaling order is not considered and the current unmarshaling order is considered. If the unmarshaling order can be optimized, that is, the traffic that does not meet the connection constraint can be made to meet the constraint by adjusting the unmarshaling order, then this part of the traffic must be included in the arrival and departure traffic difference matrix. Therefore, we only need to find the traffic position greater than 0 in the arrival and departure traffic difference matrix and make targeted adjustments to it to optimize the unmarshaling order.
[0192] Calculation of train unmarshaling sequence adjustment:
[0193] After calculating the difference in the flow of arrival and departure, all possible optimized adjustment points of the train disassembly sequence are found. According to the characteristics of train operation, if there is a difference in the flow of arrival and departure between the arriving train i and the departing train j under the current disassembly sequence, there are three ways to transform this part of the traffic from not satisfying the flow connection constraint to satisfying the flow connection constraint: advance the disassembly sequence of the arriving train i; move the assembly sequence of the departing train j backward; or do both at the same time. At the same time, there may be three situations in the adjustment of the train disassembly sequence: the adjustment amount of the disassembly sequence is not enough, such as the train disassembly position needs to be adjusted forward by 2, but it is actually only adjusted by 1, which does not achieve the purpose of adjustment; the adjustment amount of the disassembly sequence is just right; the adjustment amount of the disassembly sequence is too large, such as the train disassembly sequence only needs to be adjusted forward by 1, but it is actually adjusted by 2, which causes redundant adjustment. To this end, it is necessary to find a suitable disassembly adjustment amount. Before calculating the appropriate value of the train disassembly adjustment amount, it is necessary to first calculate the train disassembly waiting time, the disassembly time advance amount corresponding to the train disassembly sequence adjustment, the train formation advance time, and the formation time delay amount corresponding to the train formation sequence adjustment.
[0194] (1) Train dismantling waiting time:
[0195] In order to calculate the redundant waiting time for train dismantling, the train arrival situation is discussed in four cases. It is assumed that the order of train dismantling is the same as the order of train arrival, (n-1)≥i≥2. At the same time, according to existing research, the operation time of technical inspection of train arrival generally does not fluctuate much, so it is regarded as a constant.
[0196] ① The arrival interval between two adjacent trains is greater than the disassembly time of the preceding train, and the preceding train has no redundant disassembly time:
[0197] like Figure 2 As shown in (a), the arrival interval between the i+1th train and the ith train at technical station C is long enough, and its value is greater than the dismantling operation time of the ith train. The i-th arriving train does not have any redundant time for disassembly and waiting. In this case, the redundant time for disassembly and waiting of the i+1-th arriving train is:
[0198]
[0199] In formula (28), is the total dismantling waiting time of the i+1th train, is the waiting time for the i+1th train to disassemble relative to the ith train. Since this value is less than 0, the waiting redundant time for the i+1th arriving train to disassemble is 0.
[0200] ② The interval between two adjacent arriving trains is large enough, and the preceding train has redundant time to disassemble and wait:
[0201] like Figure 2 As shown in (b), the arrival interval between the i+1th train and the ith train at technical station C is long enough, and its value is greater than the dismantling operation time of the ith train. If the redundant waiting time for the disassembly of the i-th train is not considered, the redundant waiting time for the disassembly of the i+1-th train is the same as (a). However, since the i-th arriving train has redundant waiting time for disassembly, the redundant waiting time for the disassembly of the i+1-th arriving train is:
[0202]
[0203]
[0204] ③ The arrival interval between two adjacent trains is less than the disassembly time of the preceding train, and the preceding train has no redundant disassembly time:
[0205] like Figure 2 As shown in (c) in the figure, the arrival interval between the i+1th train and the ith train at technical station C is shorter, and its value is less than the dismantling operation time of the ith train. And the preceding train i has no disassembly redundant time. In this case, the disassembly waiting redundant time of the i+1th arriving train is:
[0206]
[0207] ④ The arrival interval between two adjacent trains is less than the disassembly time of the preceding train, and the preceding train has redundant time for disassembly and waiting:
[0208] like Figure 2 As shown in (d) in the figure, the arrival interval between the i+1th train and the ith train at technical station C is shorter, and its value is less than the dismantling operation time of the ith train. At the same time, there is redundant time for the disassembly and waiting of the i-th arriving train. At this time, the redundant time for the disassembly and waiting of the i+1-th arriving train is:
[0209]
[0210] Summarizing the four train arrival situations, we can find that in any case, the total waiting time for the i+1th train to arrive can be calculated using the following formula
[0211]
[0212] It is not difficult to understand that in the first two cases, since there is no redundant waiting time for the i-th arriving train to disassemble, The value of is 0, so the formula can be transformed in these two cases At the same time, if the final value of the formula is less than 0, it means that there is no disassembly redundant time, that is, the disassembly waiting redundant time is 0.
[0213] In the above formula, assuming i≥2, that is, i-1≥0, the disassembly waiting redundant time of the third arriving train, in order to calculate the waiting redundant time of all trains, it is necessary to discuss the waiting redundant time of the first arriving train and the second arriving train. Obviously, the disassembly waiting redundant time of the first arriving train must be 0, and the disassembly waiting redundant time of the second arriving disassembly train can be regarded as a special case of the two situations ①③ for calculation. So far, the disassembly waiting redundant time of all arriving trains can be calculated. At the same time, according to the formula, it can be seen that the disassembly waiting redundant time of the i+1 train is obtained by the disassembly time of the i train, the technical inspection end time of the i train, and the technical inspection end time of the i+1 train, and i and i+1 are determined by the disassembly order, not the arrival order. That is, the above formula is valid under any disassembly order.
[0214] (2) The corresponding dismantling time advance of the train dismantling sequence adjustment:
[0215] For ease of understanding, it is assumed that the train disassembly order is the same as the arrival order. According to the train disassembly waiting redundant time calculation formula, the disassembly waiting redundant time of the i+3th arrival disassembly train is At this time, the disassembly order is [i,i+1,i+2,i+3]. Assume that the value of the i+3th row and jth column of the arrival and departure flow connection difference matrix is greater than 0, that is, under the current disassembly order, the i+3th arriving train cannot provide a connecting flow for the jth departure train. In order to satisfy the flow connection constraint, a possible adjustment method is to advance the disassembly order of the i+3th arriving train. Assuming that the adjusted disassembly order is [i,i+1,i+3,i+2], then the adjusted disassembly waiting redundant time of the i+3th arriving disassembly train is After this adjustment, the disassembly end time of arriving train i+3 is advanced by
[0216] In the above formula, and can be obtained by calculation, so It is also known that the advance value of the dismantling time of the i+3 arriving train when the dismantling sequence is advanced by 1 can be calculated based on this. Similarly, all the advance values of the dismantling time of the i+3 arriving train from the current dismantling sequence until it is advanced by 1 can be obtained. The calculated results are sequentially placed in a column vector of n rows and 1 column. The value of each row represents the change in the dismantling time of the i+3 arriving train when it moves from the current dismantling sequence to the dismantling sequence represented by the row. Obviously, in general, the values of the first few rows of the column vector are the same, because in most cases, after the dismantling sequence is adjusted by a certain amount, the train waiting redundant time is already 0. At this time, if it is adjusted forward, the change in the train dismantling time will not change. The values of the rows after the dismantling sequence of the i+3 arriving train and the subsequent rows are all 0, because the dismantling time will not be advanced if the dismantling sequence moves backward.
[0217] (3) Train marshaling advance time:
[0218] In order to calculate the train marshaling advance time, the train departure situation is also divided into four cases for discussion. It is assumed that the train marshaling order is the same as the train departure order, (m-1)≥j≥2. At the same time, according to existing research, the technical inspection operation time of train departure generally does not fluctuate much, and it is regarded as a constant.
[0219] ① The departure interval between two adjacent departure trains is greater than the marshaling time of the rear train, and the rear train has no marshaling advance time:
[0220] like Figure 3 As shown in (e), the departure interval between the j-1th train and the jth train at technical station C is long enough, and its value is greater than the marshaling operation time of the jth train. And the following train j has no marshaling advance time. In this case, the marshaling advance time of the j-1th departure train is:
[0221]
[0222] In formula (34), The advance time of the total marshaling of the J-1 departure train, It is the marshaling advance time of the j-1th departing train relative to the jth departing train.
[0223] Since this value is less than 0, the marshaling advance time of the j-1th departure train is 0.
[0224] ② The interval between two adjacent departure trains is large enough, and the rear train has marshaling advance time:
[0225] like Figure 3 As shown in (f) in the figure, the departure interval between the j-1th train and the jth train at technical station C is long enough, and its value is greater than the marshaling operation time of the jth train. If the marshaling advance time of the jth train is not considered, the marshaling advance time of the j-1th train is the same as (e). However, since the jth departure train has a marshaling advance time, the marshaling advance time of the j-1th train is:
[0226]
[0227]
[0228] ③ The departure interval between two adjacent departure trains is less than the marshaling time of the rear train, and the rear train has no marshaling advance time:
[0229] like Figure 3 As shown in (g), the departure interval between the j-1th train and the jth train at technical station C is shorter, and its value is less than the marshaling operation time of the jth train. And the following train j has no marshaling advance time. In this case, the marshaling advance time of the j-1th departure train is:
[0230]
[0231] ④ The departure interval between two adjacent departure trains is less than the marshaling time of the rear train, and the rear train has marshaling advance time:
[0232] like Figure 3 As shown in (h), the departure interval between the j-1th train and the jth train at technical station C is shorter, and its value is less than the marshaling operation time of the jth train. At the same time, there is a marshaling advance time for the j-th train to depart. At this time, the marshaling advance time for the j-1-th train to depart is:
[0233]
[0234] Summarizing the four train arrival situations, we can find that in any case, the total waiting time for the j-1th train can be calculated using the following formula:
[0235]
[0236] It is not difficult to understand that in the first two cases, since there is no marshaling advance time for the jth departure train, The value of is 0, so the formula can be transformed in these two cases At the same time, if the final value of the formula is less than 0, it means that there is no marshaling advance time, that is, the marshaling advance time is 0.
[0237] In the above formula, assuming that (m-1)≥j≥2, that is, m≥j+1, the marshaling advance time of the third-to-last departure train is calculated. In order to calculate the marshaling advance time of all trains, it is necessary to discuss the marshaling advance time of the last departure train and the second-to-last departure train. Obviously, the marshaling advance time of the last departure train must be 0, and the marshaling advance time of the second-to-last departure train can be regarded as a special case of the two situations ①③ for calculation. So far, the marshaling advance time of all departure trains can be calculated. At the same time, according to the formula, it can be seen that the marshaling advance time of the j-1 departure train is obtained by the j train marshaling time, the j train departure technical inspection operation start time and the j-1 train departure technical inspection operation start time, and j and j-1 are determined by the marshaling order, not the arrival order. That is, the above formula is valid under any marshaling order.
[0238] (4) The corresponding delay in train marshaling time due to train marshaling sequence adjustment:
[0239] For ease of understanding, assume that the train marshaling order is the same as the train departure order. According to the train marshaling advance time calculation formula, the marshaling advance time of the j-3rd train departure is At this time, the grouping order is [j-3,j-2,j-1,j].
[0240] Assume that the value of the i-th row and j-3-th column of the arrival and departure flow connection difference matrix is greater than 0, that is, under the current marshaling order, the i-th arriving train cannot provide a connecting flow for the j-3-th departing train. In order to satisfy the flow connection constraint, a possible adjustment method is to postpone the marshaling order of the j-3-th departing train. Assuming that the adjusted marshaling order is [j-2, j-3, j-1, j], then the marshaling advance time of the j-3-th departing train after adjustment is After this adjustment, the marshaling start time of the departure train j-3 is postponed to
[0241] In the above formula, and can be obtained by calculation, so It is also known that the delayed marshaling time value of the j-3 departure train marshaling sequence delayed by 1 can be calculated based on this. Similarly, the delayed amount of all marshaling times of the j-3 departure train from the current marshaling sequence until it is postponed to m can be obtained. The calculated results are sequentially placed in a row vector of 1 row and m columns. The value of each column represents the change in marshaling time of the j-3 departure train from the current marshaling sequence to the marshaling sequence represented by the column. Obviously, in general, the values of the last few columns of the row vector are the same, because in most cases, after the marshaling sequence is adjusted by a certain amount, the train marshaling advance time is already 0. At this time, if it is adjusted backward, the change in train marshaling time will not change. The values of the j-3 arrival train marshaling sequence and the previous columns are all 0, because the marshaling sequence moves forward and the marshaling time will not be delayed.
[0242] (5) Determination of the adjustment amount of train disassembly sequence:
[0243] According to the arrival and departure traffic flow connection difference matrix, all traffic flows that may change from not meeting traffic flow connection to meeting traffic flow connection by adjusting the disassembly sequence can be found. At this time, select the point positions greater than 0 in the matrix in turn, calculate the train connection time difference between the disassembly train and the marshaling train corresponding to the point (the value is negative), and calculate the corresponding time change value matrix of the disassembly and marshaling train disassembly sequence adjustment corresponding to the adjustment point. At this time, as long as the sum of the disassembly change value and the marshaling change value is found from the time change matrix, it is just greater than or equal to the absolute value of the train connection time difference at this point, and the appropriate train disassembly sequence adjustment amount is found. Exactly greater than or equal means that, assuming that there are 8 arrival trains and departure trains, the disassembly sequence is the same as the arrival and departure sequence, and the connection time difference between the 5th arrival train and the 4th departure train is negative, its value is -38, and the disassembly time change matrix corresponding to the disassembly sequence adjustment of the 5th arrival train is [35 35 35 20 0 0 0 0] T , the marshaling time change matrix corresponding to the marshaling sequence adjustment of the 4th departure train is [0 0 0 0 18 27 27 27]. Then the exact point is that the dismantling sequence is adjusted forward by 2, and the marshaling sequence is adjusted backward by 1; or the dismantling sequence is adjusted forward by 1, and the marshaling sequence is adjusted backward by 2. The rest of the adjustments do not meet the meaning of exact.
[0244] Design of neighborhood search algorithm based on the poor connection between arrival and departure traffic flow:
[0245] This embodiment is a collaborative flow distribution of two technical stations under the consideration of adjustable train departure time, but the collaborative flow distribution between technical stations is based on the flow distribution results of a single technical station, while taking into account the collaborative flow distribution adjustment constraints. The core of the flow distribution problem lies in the determination of the disassembly order. For this reason, this embodiment designs a method to adjust the disassembly order based on the difference in the connection between the arrival and departure flows to construct a neighborhood, and designs a corresponding local neighborhood search algorithm to solve the problem of collaborative flow distribution between technical stations considering adjustable departure times. The algorithm is described as follows:
[0246] Step 1: Adjust the departure train value based on the train arrival and departure times (Initially, the departure times of the trains at the two technical stations are regarded as fixed values), and the ideal arrival and departure traffic matrices of the two technical stations are calculated based on the time of various technical operations.
[0247] Step 2: Select the initial solution, use the current train arrival and departure sequence of each technical station as the train disassembly sequence, and perform flow distribution to obtain the initial solution Z. Let the initial Z best =Z,Z now =Z, the train arrival and departure order is taken as the initial optimal disassembly order and the current optimal disassembly order.
[0248] Step 3: Calculate the actual arrival and departure traffic flow matrix and arrival and departure traffic flow difference matrix under the optimal disassembly sequence, and find all positions greater than 0 in the arrival and departure traffic flow difference matrix to generate a candidate set of disassembly adjustment points U. At the same time, u1 is the first adjustment point in the candidate set, and the total number of adjustment points is N. TZ .
[0249] Step 4: Neighborhood construction: select adjustment points u from the unmarshaled adjustment candidate set U in sequence i , each time only one of the points is constructed for the neighborhood (initial i = 1), and the selected disassembly adjustment point is calculated to obtain the disassembly time advance amount corresponding to the disassembly order adjustment of the arriving train at this point under the optimal disassembly order (the result is a column vector with n rows and 1 column), and the marshaling time delay amount corresponding to the marshaling order adjustment of the departing train at this point (the result is a row vector with 1 row and m columns); at the same time, the connection time difference (the value is negative) CZ between the arriving train and the departing train corresponding to the disassembly adjustment point is calculated, and all suitable train disassembly order adjustment amounts that satisfy the change value that is exactly greater than the absolute value of the train connection time difference at this point are found from the time change matrix, thereby generating the neighbors of the current solution.
[0250] Step 5: Update the current solution. Compare the flow distribution results of all neighboring solutions with the current solution in turn. If the neighboring solution result is better than the current solution, assign the neighboring solution result to the current solution Z. now , the corresponding solution order is used as the current solution order. If no neighborhood solution is better than the current solution, the current result remains unchanged.
[0251] Step 6: If i<N TZ , let i=i+1, and return to step 4. If i=N TZ , judge Z now and the previous optimal solution Z best Size, if Z now <Z best , let Z best =Z now , the optimal solution sequence is updated to Z now Corresponding to the decompilation order, return to step 3. Otherwise, output Z best , Z best Corresponding to the train disassembly sequence, the technical station flow distribution results are obtained.
[0252] Step 7: Based on the flow distribution results of a single technical station, optimize the trains that failed to be dispatched from station D, consider the coordinated flow distribution constraints and the train departure time adjustment constraints, and determine whether the departure time and marshaling content of the trains departing from station C can be adjusted to achieve coordinated flow distribution between the two technical stations. If not, output the separate flow distribution results of the two technical stations and the algorithm ends; if so, calculate the corresponding train departure adjustment time. The departure train formation is adjusted and the algorithm returns to the first step to recalculate and obtain the coordinated flow distribution result.
[0253] Case analysis:
[0254] according to Figure 1 For the railway network shown in the figure, it is assumed that the marshaling operation time of the two technical stations obeys the normal distribution, the marshaling parameters of technical stations C and D are shown in Tables 2 and 3, the train weights corresponding to different marshaling destinations are shown in Table 4, the number of full axles of the departure train is 45-55, the distance between the two technical stations is 240km, the freight train runs at a speed of 80km / h, and the confidence levels α and β are both 95%.
[0255] Table 2 Various operation times of technical station C
[0256]
[0257] Table 3 Various operation times of technical station B
[0258]
[0259]
[0260] Table 4 Corresponding train weights for different marshaling and destination groups
[0261]
[0262] Example results and analysis:
[0263] The neighborhood search algorithm based on the difference in the connection between the arrival and departure flow was programmed using MATLAB. The computing environment was a personal computer with an Inter(R)Core(TM)i7-10875H 2.30GHz CPU and 16GB of DDR4 memory. The train unmarshaling operation schemes before the adjustment of technical stations C and D were obtained, as shown in Tables 5 and 7; the train flow distribution schemes of technical stations C and D are shown in Tables 6 and 8.
[0264] According to the flow distribution results of a single technical station, technical station D has lost trains 40009 and 40004, technical station C has lost trains 20003 and 20007, and the lost train 20003 at station C is a train destined for technical station D. At the same time, there is a flow of vehicles with the same marshaling and destination F as 40009 in the station storage at technical station C. For this reason, considering the coordinated flow distribution, the marshaling and destination of train 20003 is increased to F. However, after algorithm testing, train 20003 still cannot be dispatched just by adjusting the marshaling and destination. This is because the connecting time between the arriving train 10007 at station C and the departing train 20003 is insufficient. Therefore, an attempt is made to delay the departure time of the departing train 20003. According to calculations, a delay of 2 minutes for train 20003 can enable the arriving train 10007 to provide it with a connecting flow, and the delay time does not exceed the 5-minute limit, so it is adjusted. The adjusted Substituted into the algorithm, the adjusted train disassembly operation plan of technical stations C and D is obtained, see Table 9 and Table 11; the train flow distribution plan of technical stations C and D is shown in Table 10 and Table 12, among which 50001 among the trains arriving at station D is the departure train 20003 from technical station C to station D.
[0265] The flow distribution process before and after the adjustment of technical stations C and D is analyzed, as shown in Table 13. From Table 13 and the flow distribution schemes before and after the adjustment of technical stations C and D, it can be seen that before the coordinated flow distribution and the adjustment of train departure time, the two technical stations sent out a total of 20 trains. After considering the coordinated flow distribution and the adjustment of departure time, the two technical stations sent out one more departure train, and the stock of cars at the two technical stations was reduced, which improved the efficiency of freight car transfer. At the same time, after the adjustment, the trains with the destinations of D, E, and F provided follow-up traffic for the 20003 train, reducing the stay time of the vehicles at the technical station C, but the 20003 train was sent to the technical station D, which increased the stay time of the vehicles at the D station. Therefore, the stay time of the vehicles at the technical station D after the adjustment was increased compared with that before the adjustment. However, in general, the overall stay time of the vehicles at the two technical stations was still reduced by 86.96h, indicating that after considering the coordinated flow distribution and the adjustment of departure time, the two stations can benefit as a whole.
[0266] Table 5 Unpacking operation plan before adjustment of technical station C
[0267]
[0268] Table 6 Flow distribution plan before adjustment of technical station C
[0269]
[0270]
[0271] Table 7 Unpacking operation plan before adjustment of technical station D
[0272]
[0273] Table 8 Flow distribution plan before adjustment of technical station D
[0274]
[0275] Table 9 Adjusted de-marshalling operation plan of Technical Station C
[0276]
[0277]
[0278] Table 10 Adjusted flow distribution plan of technical station C
[0279]
[0280] Table 11 Adjusted de-marshalling operation plan of Technical Station D
[0281]
[0282]
[0283] Table 12 Adjusted flow distribution plan of technical station D
[0284]
[0285] Table 13 Comparative analysis of flow distribution results
[0286]
[0287]
[0288] Algorithm solution effect and reliability analysis:
[0289] By programming the algorithm designed in this embodiment with Matlab, the vehicle stay time and Z of each vehicle adjusted based on the difference in the connection between the arrival and departure traffic flows can be obtained before and after the adjustment of C and D. now , Z best As the number of adjustments changes, Figure 4-Figure 6 They are the vehicle stay time at technical station D before adjustment, now , Z bestThe diagram shows the changes with the number of adjustments. Table 14 shows the iterative data table of technical stations C and D before and after the adjustments.
[0290] According to the algorithm idea and Table 14, the number of algorithm iterations before and after the adjustment of each technical station is different, and the number of iterations is obtained based on the arrival and departure flow connection difference matrix, rather than being set in advance, which avoids the problem of insufficient search range due to artificial setting of the number of iterations in the existing intelligent optimization algorithm. At the same time, compared with the traditional intelligent optimization algorithm, the neighborhood search algorithm based on the arrival and departure flow connection difference can avoid invalid search (invalid search refers to the adjustment of the disassembly sequence only making some vehicles change from satisfying the flow time connection to not satisfying, or adjusting the disassembly sequence, but the arrival and departure connection relationship between trains has not changed), thereby greatly narrowing the scope of the solution space and improving the algorithm convergence efficiency. After testing, considering the examples in this article, the neighborhood search algorithm based on the arrival and departure flow connection difference has a calculation time of no more than 20s for the flow distribution results of a single technical station, and a calculation time of no more than 1min for the collaborative flow distribution results of two technical stations, which can meet the needs of on-site operations.
[0291] Since the train unmarshaling time is a random variable and obeys the normal distribution, Matlab is used to randomly generate the unmarshaling operation time for simulation testing. The experiment was carried out 10,000 times, and the feasible number of flow distribution schemes was 9823 times, with a reliability of 98.23%.
[0292] Table 14 Iterative data table of vehicle dwell time at the station
[0293] Technology Station Category Before adjustment of Technical Station C After adjustment of Technical Station C Technical Station D before adjustment After adjustment of Technical Station D Total number of iterations 20 26 40 60 Now solve the number of changes 6 5 5 4 Best solution change times 4 4 4 4
[0294] in conclusion:
[0295] This example introduces stochastic programming theory to study the problem of coordinated flow distribution between adjacent technical stations considering the randomness of train disassembly time and flexible constraints of departure time, constructs a random chance constrained programming model for this problem, calculates the difference between arrival and departure flow connection based on the characteristics of the flow distribution problem, and designs an efficient neighborhood search algorithm based on the difference between arrival and departure flow connection to solve it. Through the analysis of the example, the following conclusions are obtained:
[0296] (1) In the process of formulating the coordinated flow distribution plan between technical stations, if the flexible constraints of train departure time and the coordinated optimization between stations are considered at the same time, more trains can be dispatched from the technical station. Fine-tuning some trains within the allowable range of train departure time deviation will cause these trains to depart on time, but it can benefit the two stations as a whole. The example results show that compared with the flow distribution of a single technical station, considering the flexible constraints of train departure time and the coordinated optimization between stations, the departure train 20003 from Station C is adjusted, the total stay time of vehicles at the two technical stations is reduced by 86.96h, the number of successful departures is increased by 2, and the efficiency of truck transfer is effectively improved.
[0297] (2) The neighborhood search algorithm based on the difference in the connection between the arrival and departure flows designed in this embodiment is extremely applicable to the flow distribution problem of technical stations. Compared with the existing intelligent optimization algorithm, it can avoid invalid searches, narrow the search range of the solution space, and obtain the optimal solution for flow distribution in a short time. At the same time, the chance-constrained programming can obtain a more reliable solution, which can reduce the impact of uncertain factors in considerable cases.
[0298] The above-mentioned embodiments only express the specific implementation of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.
Claims
1. A method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method of the difference between arrival and departure flow, characterized in that: include: Model construction: Set the objective function and construct a random chance constrained programming model between adjacent technical stations that takes into account the randomness of train disassembly time and flexible constraints on departure times; Model solution: transform the opportunity constraints into deterministic equivalent classes, and then realize the coordinated flow distribution of adjacent technical stations based on the neighborhood search algorithm of the connection difference of the arrival and departure flow; The constraints are as follows: (A) Train dismantling time constraints: For the disintegration order of Y h The earliest dismantling time of the arriving train is the end time of the train inspection, and the actual dismantling time of the train shall not be earlier than the end time of the train inspection. Since the train dismantling time is a random variable subject to normal distribution, the chance constraint (6) is established with a certain probability α. At the same time, the actual dismantling time of the train must also be earlier than the end time of the dismantling of the previous dismantling train: (B) Train marshaling time constraints: For a departing train with a marshaling order of Zl, its latest marshaling end time is the departure time of the train minus the departure technical operation inspection time, and the actual marshaling end time of the train shall not be later than its latest marshaling end time; since the train marshaling time is a random variable that obeys a normal distribution, the chance constraint (10) is established with a certain probability β. At the same time, the actual marshaling end time of the train must also be earlier than the start marshaling time of the next train: (C) Traffic flow constraints: ① In a single technical station, if the arriving train i can provide a connecting train flow for the departing train j, the marshaling start time of the departing train j must be greater than the disassembly end time of the arriving train i: Where M is a large constant; ② The number of trains with the vehicle group number k provided by the arriving train i to the departing train cannot exceed the number of corresponding vehicle groups contained in the arriving train: ③ The traffic provided by the arriving train i to the departing train j must be the train set included in the j train: In this formula, M can be taken as the number of cars in the departure train j; (D) Train full axle constraint: For the departure train j to be successfully dispatched, at least one of the full-axle constraints, full-weight or full-length, must be satisfied; (E) Coordinated flow constraints: If the departure train j of the previous technical station corresponds to the arrival train i'' of the next technical station and can provide a connecting train flow for the departure train j' of the next technical station, then the following conditions shall be met: (F) Constraints on adjustment of departure train composition: If the departure train j' of the next technical station does not meet the full axle constraint of train departure, and there is a lost train j at the previous technical station, which corresponds to the arrival train i' of the next technical station, and there is a train flow with the same marshaling content as the departure train j' of the next technical station in the unsuccessful traffic flow of the previous technical station, then consider adjusting the marshaling content of the lost train at the previous technical station, and the adjustment should meet the following requirements: Before the content of the train j departing from the previous technical station is adjusted: After the content of the j formation of the train departing from the previous technical station is adjusted: Wherein, O represents a technical station, where O=1 is the previous technical station, O=2 is the next technical station, i is the arriving train, i=1,2,…,n; j is the departing train, j=1,2,…,m; is the arrival time of train i at technical station O; is the scheduled departure time of the departure train j; The end time of the stage of technical station O; is the arrival technical operation time of technical station O; The departure technical operation time of technical station O; is the dismantling operation time of the arriving train i at the technical station O; is the marshaling operation time of the departure train j from the technical station O; Y1, Y2, Y h , …, Y n is the train disassembly sequence, corresponding to the train arrival sequence i = 1, 2, ..., n; Z1, Z2, Z l , …, Z m is the train marshaling sequence, corresponding to a permutation of the train departure sequence j = 1, 2, …, m; is the deviation of the departure time of train j from technical station O from the on-time schedule. A positive value indicates a late departure, while a negative value indicates an early departure of the train. is the earliest disassembly start time of the arriving train i at the technical station O; is the actual disassembly start time of the arriving train i at the technical station O; is the disassembly end time of the arriving train i at the technical station O; is the marshaling start time of the departure train j from the technical station O; is the latest marshaling end time of the departure train j from the technical station O; is the actual marshaling end time of the departure train j at the technical station O; is a 0-1 variable. When the arriving train i in the technical station O can provide a connecting train flow for the departing train j, its value is 1, otherwise it is 0; k represents the train flow group number, and K is the set of all train flow group numbers. represents the number of vehicles with the vehicle flow group number k provided by the arriving train i at the technical station O to the departing train j; represents the number of vehicles with vehicle group number k among the arriving train i at technical station O; It is a 0-1 variable, indicating whether the departure train j contains the destination of the k-th train. If so, it is 1, otherwise it is 0; It is a 0-1 variable, indicating whether the departure train j contains the destination of k' after adjusting the marshaling content. If it does, it is 1, otherwise it is 0; and The maximum and minimum number of vehicles corresponding to the departure train j of technical station O; and is the maximum and minimum weight change corresponding to the departure train j of technical station O; w k Indicates the corresponding weight change of the vehicle group with vehicle group number k; It is a 0-1 variable, indicating whether the train j departing from technical station O meets the minimum number of trains required; is a 0-1 variable, indicating whether train j departing from technical station O meets the minimum weight change requirement; is a 0-1 variable, indicating whether the train j from technical station O is successfully dispatched, and is 1; L CD is the distance between adjacent technical stations; v is the running speed of the freight train.
2. The method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method for the difference between arrival and departure train flows according to claim 1 is characterized in that: The objective function is as follows: minG=G1+G2 (1) 3. The method for optimizing the coordinated flow distribution of freight trains between technical stations based on the neighborhood search method for the difference between arrival and departure flow according to claim 1 is characterized in that: The neighborhood search algorithm based on the difference in the connection between the arrival and departure traffic flow specifically includes the following steps: Step 1: According to the train arrival and departure times, the departure train adjusts the various technical operation times, and calculates the ideal arrival and departure connection flow matrix of two adjacent technical stations; Step 2: Select the initial solution, use the current train arrival and departure sequence of each technical station as the train disassembly sequence, and perform flow distribution to obtain the initial solution Z. Let the initial Z best =Z,Z now = Z, the train arrival and departure sequence is taken as the initial optimal disassembly sequence, the current optimal disassembly sequence; Step 3: Calculate the actual arrival and departure traffic flow matrix and arrival and departure traffic flow difference matrix under the optimal disassembly sequence, and find all positions greater than 0 in the arrival and departure traffic flow difference matrix to generate a candidate set of disassembly adjustment points U. At the same time, u1 is the first adjustment point in the candidate set, and the total number of adjustment points is N. TZ ; Step 4: Neighborhood construction: select adjustment points u from the unmarshaled adjustment candidate set U in sequence i , each time only one of the points is constructed for the neighborhood, initially i = 1, the selected disassembly adjustment point is calculated, and the disassembly order adjustment corresponding to the arrival train at this point under the optimal disassembly order is obtained, and the result is a column vector with n rows and 1 column, and the marshaling time delay corresponding to the marshaling order adjustment corresponding to the departure train at this point is calculated, and the result is a row vector with 1 row and m columns; at the same time, the connection time difference CZ between the arrival train and the departure train corresponding to the disassembly adjustment point is calculated, and the value is negative, and all suitable train disassembly order adjustment values that satisfy the change value just greater than the absolute value of the train connection time difference at this point are found from the time change matrix, thereby generating the domain of the current solution; Step 5: Update the current solution. Compare the flow distribution results of all neighboring solutions with the current solution in turn. If the neighboring solution result is better than the current solution, assign the neighboring solution result to the current solution Z. now , the corresponding solution order is used as the current solution order. If no neighbor solution result is better than the current solution, the current result remains unchanged; Step 6: If i<N TZ , let i=i+1, return to step 4, if i=N TZ , judge Z now and the previous optimal solution Z best Size, if Z now <Z best , let Z best =Z now , the optimal solution sequence is updated to Z now Corresponding to the decompilation order, return to step 3; otherwise, output Z best , Z best Corresponding to the train disassembly sequence, the technical station flow distribution results are obtained; Step 7: Based on the flow distribution results of a single technical station, optimize the trains that failed to be dispatched from the next station, consider the coordinated flow distribution constraints and the train departure time adjustment constraints, and determine whether the departure time and marshaling content of the train departing from the previous station can be adjusted to achieve coordinated flow distribution between the two technical stations. If not, output the separate flow distribution results of the two technical stations and the algorithm ends; if so, calculate the corresponding train departure adjustment time. The departure train formation is adjusted, and the algorithm returns to step 1 to recalculate and obtain the coordinated flow distribution result.
