Aperiodic ship scheduling method and system considering uncertain harbor time, and medium

Abstract

The invention discloses an irregular ship scheduling method and system considering uncertain harbor time, and a medium, and belongs to the technical field of shipping management. The method comprises the following steps: collecting ship, cargo, port and interference event data; generating an initial scheduling scheme by adopting a greedy strategy; generating multiple groups of random interference scenes based on Monte Carlo simulation; a two-stage stochastic programming model is constructed, in the first stage, income maximization serves as a target, and an initial plan is generated through damage and repair operator iterative optimization by means of a self-adaptive large neighborhood search algorithm; in the second stage, aiming at each interference scene, a recovery network is constructed for the influenced goods, and an optimal recovery strategy and adjustment cost are solved by utilizing a shortest path algorithm; and finally, outputting a scheduling scheme including a ship path, a navigational speed and a multi-scene recovery strategy through iterative optimization. According to the method, the earnings and the operation efficiency of the shipping companies are improved, the adjustment cost caused by uncertainty is reduced to the greatest extent, the benefits of the shipping companies and cargo owners can be balanced in a complex scheduling problem, and the service quality is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of shipping management, and particularly relates to a tramp ship scheduling method and system considering uncertain port time and a medium. BACKGROUND

[0002] Tramp shipping, as a flexible transportation mode, plays an important role in the global shipping market. Its main features include flexibility of routes, diversity of cargo types, and rapid response capability to dynamic market demand. Shipping companies can selectively accept new cargo transportation plans according to the changes in the spot market and flexibly adjust the routes to achieve greater benefits in the changing market environment. However, this flexibility also makes the tramp ship scheduling problem extremely complex.

[0003] In actual operation, shipping companies need to handle mixed transportation demands of contract cargo and spot cargo. Contract cargo usually has a mandatory requirement to be transported in full, while spot cargo can be selectively accepted according to the benefits and transportation capacity of the ship. Under the premise of meeting time window constraints and load limits, shipping companies need to reasonably select cargo and plan routes to maximize benefits.

[0004] In addition, tramp ship scheduling also faces challenges from various uncertainty factors. Common interference events include adverse weather, cargo owner default, and port congestion, etc. These events can cause delays in loading and unloading tasks, prolong the port time of the ship, and thus affect the execution of subsequent transportation tasks. The initial scheduling plan often fails to achieve good results in actual operation due to these uncertainty factors, resulting in increased adjustment costs and affecting the operational efficiency and benefits of shipping companies.

[0005] Although shipping companies can obtain the probability of cargo owner default, weather change trend, and the possibility of port congestion in advance through historical data of cargo owners and meteorological prediction data, existing scheduling methods fail to fully consider the impact of these uncertainty factors, lack of quantitative assessment and avoidance strategies for potential risks. Existing technologies fail to effectively combine the multi-dimensional decisions of ship-cargo matching, route planning, and speed adjustment in the optimization process, making it difficult to maximize the benefits of shipping companies and improve operational efficiency in uncertain environments.

[0006] Therefore, the existing technology has the following deficiencies in dealing with the tramp ship scheduling problem: it fails to fully consider the impact of uncertainty factors (such as weather, cargo owner default, and port congestion) on the scheduling plan; it lacks quantitative assessment and optimization strategies for risk losses caused by uncertain port time; it does not combine multi-dimensional joint optimization methods of ship-cargo matching, route planning, and speed adjustment.

[0007] To solve the above problems, the application provides an irregular ship scheduling optimization method based on a two-stage stochastic programming model, which aims to quantify the risk loss of uncertain port time, optimize the initial scheduling plan, improve the revenue and operation efficiency of shipping companies, and minimize the adjustment cost caused by uncertainty. SUMMARY

[0008] The application aims to provide an irregular ship scheduling method, system and medium considering uncertain port time, a ship-cargo matching, route and speed optimization method based on a two-stage stochastic programming model, which is used to maximize the revenue of shipping companies and reduce risk loss under uncertain port time.

[0009] The application achieves the above-mentioned purpose by the following technical solutions.

[0010] An irregular ship scheduling method considering uncertain port time, specifically comprising:

[0011] Step 1: Collect ship data, cargo data, port data and possible interference event data, and perform preprocessing;

[0012] Step 2: Based on the limited scenario support set of interference events mastered by the shipping company, generate random interference scenarios in the planning period by using Monte Carlo simulation, the more scenarios generated, the more comprehensive the decision-making, and the longer the solution time;

[0013] Step 3: Adopt a greedy strategy, preferentially allocate contract cargo, then allocate spot cargo, and meet the load limit, time window constraint and speed limit, generate an initial scheduling scheme MF, and then generate an adjustment scheme MFR(k) under different scenarios k by using a path recovery algorithm;

[0014] Step 4: One-stage solution generation: maximize the revenue of the shipping company as the target, and iteratively optimize the current solution by using a destruction operator and a repair operator to generate an initial scheduling plan;

[0015] Step 5: Two-stage solution generation: for each interference scenario, construct a recovery network, calculate the adjustment cost and generate a recovery scheme by using a shortest path algorithm;

[0016] Step 6: Repeat steps 4 and 5, continuously optimize the scheduling scheme when the maximum number of iterations is not reached, evaluate the temporary solution generated in each iteration, if the temporary solution is better than the current solution, update the current solution, otherwise, accept the poor solution according to the set probability to avoid falling into local optimum, and output the final scheduling scheme containing the transportation path, speed and recovery strategy when the maximum number of iterations is reached.

[0017] Further, the data in step 1 includes: the maximum load limit of each ship, the port where the ship is located at the beginning of the planning period, available time, light weight, economic speed, maximum speed, cargo type, the weight of the cargo, the income obtained by transporting the cargo, the loading port and unloading port of the cargo, the loading and unloading time window and the loading and unloading time requirement, the distance between ports, event type, occurrence probability, impact degree, compensation information, fuel price.

[0018] Further, step 2 generates scenarios according to the limited scenario support set of interference events mastered by the shipping company through Monte Carlo simulation, and each scenario contains the following information: the port where the interference event occurs, the time period when the interference event occurs, the impact degree of the interference event on the loading and unloading task, i.e. the delay days, and the output scenario set is used for solving the model.

[0019] Further, the greedy strategy used in step 3 specifically includes:

[0020] Prioritize the allocation of contract cargo, arrange the cargo according to the time window, and allocate the cargo in turn to ensure that the ship allocated to each cargo does not exceed its maximum load limit and meets the cargo loading and unloading time window constraints and speed restrictions;

[0021] Each piece of cargo is allocated to the ship that consumes the least fuel cost from the start of the available port to the loading port of the cargo;

[0022] The start available time and port of the ship allocated to the cargo are updated to the unloading time window and unloading port of the last item of cargo in the cargo sequence;

[0023] After the allocation of contract cargo is completed, the spot cargo is allocated according to the freight from high to low, and each contract cargo is allocated to the cargo sequence and the corresponding position that can obtain the highest profit difference while meeting the load limit, time window requirement, and speed limit;

[0024] Cargo that cannot find a suitable allocation position will be cancelled;

[0025] After obtaining MF, an adjustment scheme MFR(k) under different scenarios k is generated through a path recovery algorithm.

[0026] Further, the destruction operator in step 4 includes a random removal operator, a shaw removal operator, a worst removal operator, and a highest risk removal operator; and the repair operator includes a greedy insertion operator, a minimum opportunity insertion operator, and a risk minimum insertion operator.

[0027] Further, the construction of the recovery network specifically includes:

[0028] (1) For each affected cargo f in each scenario k, a recovery network is constructed The node N includes a loading node, an unloading node, and a start node of the next task or a node at the end of the planning period; the arc A represents the sailing between ports or the loading and unloading task performed in the port; the weight C is the weight of each arc, representing the cost of performing the task, including the additional cost of time saving and delay;

[0029] (2) Loading and unloading node processing: loading node: the node actually starting loading after being disturbed and the node set completing the loading task; unloading node: the node set starting unloading and completing the unloading task, including the node arriving at the planned time and the delay time;

[0030] (4) Path weight calculation: the weight of the arc is dynamically adjusted according to the length of the time saved or the delay time, including the additional fuel cost, the berthing cost and the delay cost;

[0031] (5) Shortest path finding: the Dijkstra algorithm is used to find the shortest path from the start node to the target node in the recovery network, which is the best recovery scheme.

[0032] Further, the generating a recovery scheme specifically comprises:

[0033] (1) Inputting a planned scheduling scheme MF and a set of possible scenarios K in the planning period;

[0034] (2) For each scenario k∈K, performing iteration of steps (3) to (6), and executing step (7) after traversing all scenarios;

[0035] (3) Initializing the adjustment cost AC of the scenario k k = 0, for each cargo f performed by the planned scheduling scheme MF, judging whether the related task is affected by the scenario k, if affected, performing iteration of steps (4) to (6), and if not affected, skipping;

[0036] (4) For each cargo f affected by the scenario k, constructing a recovery network

[0037] (5) Finding the shortest path by the Dijkstra algorithm, and recording the shortest path weight as

[0038] (6) The compensation obtained for the cargo f disturbed in the scenario k;

[0039] (7) Stopping the algorithm, and outputting the expected total adjustment cost AC = ∑ k∈K AC k / |K|, |K| is the number of generated scenarios.

[0040] Further, the objective function in step 6 is:

[0041]

[0042] wherein the first term of the objective function is the freight revenue, the second and third terms are the fuel cost and the port cost, and the fourth term is the expected adjustment cost, i.e., the objective function of the second stage model;

[0043] is the additional cost caused by the adjustment of the scheme in all scenarios k after the determination of the one-stage decision variable, p is the probability of the occurrence of each scenario, m represents the ship of the shipping company, f represents the cargo, k represents the set of scenarios generated by the Monte Carlo simulation, and p is the probability of the occurrence of each scenario, which is the reciprocal of the number of scenarios; is 1 when the ship m transports the cargo f through the arc (s, t), and 0 otherwise, (s, t) ∈ A(f); is 1 when the ship m is empty from the s node to the t node, and 0 otherwise; is 1 when the ship m performs the transportation demand of the cargo f, and 0 otherwise; is 1 when the ship m is at anchor from the s node to the t node, and 0 otherwise;d f is 1 when the ship m performs the transportation demand of the cargo f, and 0 otherwise; is the fuel cost consumed by the ship m from the s node to the t node, and γ is the risk preference, which is a value between 0 and 1, and the greater the value, the higher the degree of attention to risk in the decision-making process;

[0044]

[0045] wherein the first term is the economic compensation obtained by the shipping company in the scenario k, the second term is the cost paid for shortening the port time, and the third term is the cost paid for delaying the delivery of the cargo and adjusting the speed; is a 0-1 variable, which is 1 when the work related to the cargo f through the arc (s, t) in the first stage is adjusted to (i, j) in the scenario k, and 0 otherwise, (s, t) ∈ A1(f) ∪ A'1(f), (i, j) ∈ A3(s, t), and the work includes two voyage segments of transporting the cargo f from the loading port to the unloading port and leaving the unloading port; is a continuous variable, which represents the shortened loading time of the cargo f at the loading port; is a continuous variable, which represents the shortened unloading time of the cargo f at the unloading port; represents the additional fuel cost paid for completing the original planned arc (s, t) work through the arc (i, j) in the scenario k; represents the additional delay cost paid for completing the original planned arc (s, t) work through the arc (i, j) in the scenario k; denotes the time shortened compared to the original plan when the cargo f is completed by the arc (i, j) in the scene k; ct denotes the cost paid for each unit of shortened loading and unloading time.

[0046] A computer device / apparatus / system comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to implement the steps of the tramp ship scheduling method considering uncertain port time.

[0047] A computer readable storage medium having stored thereon a computer program / instructions which, when executed by a processor, implement the steps of the tramp ship scheduling method considering uncertain port time.

[0048] The beneficial effects of the present application are:

[0049] The present application provides a tramp ship scheduling optimization method based on a two-stage stochastic programming model, which can comprehensively quantify the risk loss of uncertain port time, incorporate interference factors such as weather changes, cargo owner defaults and port congestion into the scheduling plan, and significantly improve the robustness and adaptability of the scheduling scheme. By simultaneously optimizing ship-cargo matching, route planning and speed adjustment, the present application maximizes the shipping company's revenue under the premise of meeting time window constraints and load restrictions. The heuristic algorithm based on ALNS combined with the path recovery algorithm can efficiently solve large-scale scheduling problems, significantly shorten the calculation time, and be suitable for complex scenarios in actual operation. The present application not only effectively reduces the adjustment cost caused by uncertain factors, including delay cost, fuel cost and cargo owner compensation fees, but also improves the satisfaction of cargo owners and enhances the market competitiveness and service quality of shipping companies by reducing cargo delays and cancellations. In addition, the present application is suitable for tramp ships and cargo transportation demands of various scales, and can flexibly cope with different market environments and uncertain factors, having wide applicability and promotional value. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 is a flowchart of the heuristic algorithm designed by the present application based on the ALNS algorithm;

[0051] Figure 2 is a recovery network design diagram based on the initial scheduling plan and the interference scenario of the method of the present application. DETAILED DESCRIPTION

[0052] The present application will be further described below in conjunction with the accompanying drawings.

[0053] The purpose of the present application is to fully consider the interests of shipping company benefits and cargo transport efficiency, make appropriate trade-off and balance, and use the structural characteristics of ship operation as much as possible to improve the scheduling quality and efficiency in the process of scheduling according to the actual data of shipping company operation and the market demand for cargo transportation.

[0054] The scheduling problem of tramp ship is a very complex problem in actual operation, which needs to comprehensively consider various factors such as the characteristics of the ship (such as maximum load, starting position, available time, etc.), cargo information (such as weight, loading and unloading port, benefit, time window, etc.), fuel price, weather forecast data and history record of cargo owner default, etc. For the shipping company, the scheduling scheme of the ship has a direct impact on its operating cost and benefit. Therefore, a good scheduling scheme needs to consider the above factors comprehensively.

[0055] Generally, the optimization model problem of tramp ship scheduling is described as follows: a shipping company has several heterogeneous ships in operation, which need to complete a series of cargo transportation tasks within the planning period. In the process of operation, the ship may not be able to perform the task as planned due to various interference events (such as bad weather, cargo owner default, port congestion, etc.), resulting in the need to adjust the original scheduling plan to restore normal operation as soon as possible, so as to minimize the loss caused by uncertainty.

[0056] In the corresponding design scenario of the present application, if the ship is delayed or the task is cancelled, the cargo transportation will face the risk of delay and cancellation; the shipping company will pay the corresponding economic cost due to the delay of the ship and the cancellation of the task, and the cargo owner will suffer economic loss due to the delay and cancellation of the cargo transportation. When the interference event occurs, the shipping company needs to design a new ship scheduling scheme to maintain the normal execution of the transportation task as much as possible, which is called ship recovery scheme in the present application. According to the theory of the present application, the recovery scheme of the ship is the corresponding mode of the ship and the cargo transportation task, that is, the shipping company needs to decide the transportation path and speed of each ship.

[0057] From the technical scheme, due to the large number of constraints involved, the computer algorithm is generally used to generate the scheduling scheme. The most core problem in the scheduling algorithm is to solve the continuity constraint of the ship, that is, the unloading port of the ship completing the last transportation task must be the loading port of the next transportation task. In order to solve this problem, from the perspective of neighborhood search, several basic situations of exchanging transportation tasks between ships are considered to design neighborhood operators, which avoids violating the continuity constraint from the structure, thereby improving the operation efficiency of the algorithm from the perspective of structural design.

[0058] In the actual operation process, the specific implementation steps are as follows:

[0059] Step 1, data acquisition and pretreatment:

[0060] The relevant data required for shipping company operation is collected and processed and stored in the database, which is updated with the daily operation plan of the shipping company. The collected data includes:

[0061] Ship data: describes the characteristics and initial state of each ship in the fleet, including: maximum load limit, starting position, available time, weight of each cargo, loading and unloading port, income, loading and unloading time window, light weight, economic speed, maximum speed;

[0062] Cargo data: describes the characteristics and requirements of the cargo to be transported, including: cargo type (divided into contract cargo and spot cargo), weight of cargo (unit: ton), income obtained by transporting the cargo (unit: yuan), loading and unloading port of the cargo, loading and unloading time window (the earliest start time and the latest start time of cargo loading and unloading operation) and loading and unloading time requirement (the time required to complete the loading and unloading operation (unit: day)) and other requirements;

[0063] Port data: mainly the distance between ports (unit: nautical miles);

[0064] Interference event data: interference event data describes the uncertainty factors that may affect ship scheduling, including: event type (such as bad weather, cargo owner breach of contract, port congestion, etc.), occurrence probability, impact degree, compensation information (delay caused by interference events, economic compensation that the shipping company may obtain), fuel price, etc;

[0065] Weather forecast data: weather forecast (weather conditions and probability of each port);

[0066] Cargo owner credit data: cargo owner breach of contract history (probability of breach of contract and compensation amount at each port).

[0067] Step 2, initial solution generation:

[0068] The MF of the initial solution is generated with a greedy strategy, which prioritizes allocating contract cargo, arranges the cargo in time windows, and sequentially allocates the cargo to ensure that the allocated ship for each cargo does not exceed its maximum load limit and meets the cargo loading and unloading time window constraints and speed restrictions. The greedy strategy requires that the ship allocated to each piece of cargo has the lowest fuel cost from the start of the available port to the loading port of the cargo. The start of the allocated cargo ship and the port are updated to the unloading time window and unloading port of the last item in the cargo sequence. After the contract cargo is allocated, the spot cargo is allocated, i.e. the spot cargo is arranged in descending order of freight, each contract cargo is allocated to the cargo sequence and the corresponding position that can obtain the highest profit difference while meeting the load limit, time window requirement, and speed limit. If a suitable allocation position cannot be found, the cargo will be cancelled. After obtaining the MF, the path recovery algorithm is used to generate the adjustment scheme MFR(k) under different scenarios k.

[0069] Step three, random scene generation:

[0070] Based on the limited scene support set of interference events mastered by shipping companies, Monte Carlo simulation technology is used to generate random interference scenarios that may occur during the planning period.

[0071] The specific steps are as follows: data input: input the probability of possible interference events at each port, including weather changes (such as heavy rain, heavy snow, etc.), cargo default, port congestion, etc. These data come from historical records and weather prediction platforms. Scene simulation: a large number of possible future scenarios are generated through Monte Carlo simulation. Each scenario contains the following information: the port where the interference event occurs: determine which ports may be affected. Time period of interference event occurrence: determine the specific time of interference event occurrence. Influence degree of interference event on loading and unloading task: for example, delay days. According to historical data and prediction information, assign a corresponding delay day to the severity of each interference event (such as snow will cause the port operation of the day to be delayed for one day, rain will cause a delay of 0.5 days, and port congestion will affect the ship to dock, which will be delayed until the congestion ends and the operation begins). Compensation information: according to the type and severity of the interference event, determine the economic compensation that the shipping company may obtain. Scene output: the generated scene set is used for the solution of the subsequent two-stage model to ensure that the scheduling plan can cope with future uncertainties.

[0072] Step four, one-stage solution generation:

[0073] In the first stage, the goal is to maximize the revenue of the shipping company, and a new scheduling scheme is generated by designing a destruction and repair operator to develop an initial scheduling plan. The specific steps are as follows:

[0074] The constructed space-time network is described as a directed network G, G = (N, A). Wherein, N is the node set, the node includes different ports required by the ship to complete the cargo transportation task in different time periods, that is, the actual node, and the virtual node added to meet the modeling and solving requirements, all ships are driven from the virtual starting node to the starting available node of each ship at the beginning of the planning period, all ships will reach the virtual termination node from the current node at the end of the planning period, the time dimension in the space-time network is divided into "days" from the beginning of the planning period to the end of the planning period, and the space dimension is numbered according to the ports involved in the transportation process, so each actual node includes time and space two dimensions, which are represented by an ordered pair. A is an arc set, which represents the movement between nodes. Each network arc a = (s, t) has its own starting point s and ending point t. The movement in different space-time nodes is represented as sailing arc A1, that is, the ship sails from one port to another port. Since the ship is in the process of sailing, it has two states of loading and emptying, in order to distinguish whether the ship carries goods during the movement between nodes, the sailing arc A1 is divided into carrying sailing arc A1(f) and empty sailing arc A'1, and the port arc A2 represents the movement in the same space node, that is, the process of the ship staying in the port. Therefore, the scheduling scheme of the ship m in the planning period can be represented as the transportation process from the virtual starting node s0 to the virtual termination node t0 of the ship, which is represented by a series of arcs.

[0075] Objective function: maximize the total revenue of the shipping company, the total revenue is the income of transporting goods minus the fuel cost, considering the weight limit, time window constraint and speed limit.

[0076] The objective function can be expressed as:

[0077]

[0078] Wherein: indicates that the ship m transports goods f through arc (s, t) is 1, otherwise 0, (s, t) ∈ A(f); indicates that the ship m is empty from s node to t node is 1, otherwise 0; indicates that the ship m executes the transportation demand of goods f is 1, otherwise 0; indicates that the ship m is parked from s node to t node is 1, otherwise 0;d f indicates that the transportation demand of goods f is accepted is 1, otherwise 0. is the fuel cost consumed by the ship m from node s to node t.

[0079] The constraint conditions mainly include the ship weight limit, the cargo loading and unloading time window constraint, the ship speed limit, the ship flow constraint, which are specifically:

[0080]

[0081] where q m is the total number of shipping company vessels, l m is the maximum load limit of ship m, w f is the weight of cargo f. Constraint 2 indicates that after the shipping demand of cargo f is assigned to ship m, the ship needs to perform the corresponding sailing arc; constraint 3 indicates that the shipping task needs to be performed by the corresponding ship; constraint 4 indicates that the contract cargo must be shipped by a ship; constraint 5 indicates that the spot cargo can be optionally shipped; constraint 6 indicates that the shipping of the cargo by the ship needs to meet the load limit; constraint 7 indicates that when the shipping demand of cargo f is accepted, the cargo can be shipped by at most one sailing arc, and the cargo that is not accepted will not be shipped by any sailing arc. Constraints 8 and 9 indicate that after ship m selects to ship cargo f, the corresponding loading and unloading tasks need to be performed, and constraint 10 indicates that the flow conservation constraints are imposed at the virtual departure node, the virtual termination node and other nodes to ensure that each ship can form a complete shipping path within the planning period.

[0082] The solution of the first stage is generated based on a heuristic algorithm of ALNS, which is a meta-heuristic algorithm for solving combinatorial optimization problems, proposed by Ropke et al. The algorithm is developed from the large neighborhood search (LNS) in discrete optimization and can effectively solve various optimization problems. The core idea is to perform destruction and repair operations and an adaptive mechanism.

[0083] The adaptive mechanism of the ALNS algorithm is reflected in the dynamic adjustment of weights and the selection of operators. Initially, all operators have the same weight and score. In each iteration process, the removal operator and the repair operator used in this round of iteration are determined by using the roulette method, that is, the greater the weight of an operator, the greater the probability of being selected. The score of an operator is given according to the effect of the obtained solution, and a higher score indicates that the operator performs better. The weight of an operator is updated according to the score of the operator after a certain number of iterations. If an operator can bring greater effect improvement, the weight of the operator will increase, so that the operator has a greater probability of being used in subsequent iterations. In addition, the adaptive mechanism of the ALNS algorithm is also reflected in the acceptance criterion for the current solution. Common acceptance criteria include the improvement criterion and the simulated annealing criterion. The improvement criterion requires that only when the new solution is better than the current solution can it be accepted, while the simulated annealing criterion allows a poor solution to be accepted with a certain probability in each iteration to avoid falling into a local optimum.

[0084] The steps of the algorithm are: (1) initialization: generate an initial solution; (2) destruction: select a removal operator to destroy the current solution; (3) repair: select a repair operator to repair the destroyed solution and generate a temporary solution; (4) evaluation: evaluate the temporary solution and compare it with the current solution; (5) update: if the temporary solution is better than the current solution, then the new solution is taken as the current solution, otherwise, accept the worse solution with a certain probability. Repeat the above steps until the termination condition is reached.

[0085] In order to apply the ALNS algorithm to solve the tramp ship scheduling problem considering risk loss, the algorithm is extended, and the problems discussed in the present application include the initial planning and scheduling stage and the rescheduling stage facing the interference scenario. The solution of the problem consists of two parts: the planning and scheduling scheme and the adjustment scheme under different scenarios k. The planning and scheduling scheme MF is the transportation path of each ship, which is determined by the cargo sequence transported by each ship. In each iteration, it is generated by removal operators and repair operators. After the planning and scheduling scheme and scenario k are determined, the adjustment scheme MFR(k) is generated by a path recovery algorithm designed based on Dijkstra algorithm, which solves the rescheduling (adjustment) problem required by the planning and scheduling scheme when facing the interference of scenario k.

[0086] Among them, the destruction operator includes a random removal operator, a shaw removal operator, a worst removal operator, and a highest risk removal operator. These operators are used to remove part of the cargo from the current solution to generate a new solution space. The repair operator includes a greedy insertion operator, a regret insertion operator, and a risk minimum insertion operator. These operators are used to insert the removed cargo into the scheduling scheme to generate a new feasible solution.

[0087] The steps of the random removal operator are:

[0088] (1) randomly select a cargo, and the contract cargo removal probability is p1, that is, the spot cargo removal probability is p2 (p1 < p2).

[0089] (2) generate a random number, if it is less than the removal probability, then remove the cargo.

[0090] (3) repeat until q cargos are removed.

[0091] (4) return the current solution.

[0092] The steps of the shaw removal operator are:

[0093] (1) randomly select a cargo removal.

[0094] (2) calculate the similarity of the remaining cargos and the removed cargo, and sort them in descending order of similarity.

[0095] (3) remove the (q-1) cargos with the highest similarity.

[0096] (4) Return the current solution.

[0097] The worst removal operator step:

[0098] (1) Calculate the freight unit price, and sort from low to high.

[0099] (2) Remove the lowest freight unit price, and remove the contract freight with a probability p, or directly remove the future freight.

[0100] (3) Return the current solution.

[0101] The highest risk removal operator step:

[0102] (1) Calculate the delay expectation of the freight, and sort from high to low.

[0103] (2) Remove the q freight with the highest delay expectation.

[0104] (3) Return the current solution.

[0105] The greedy insertion operator step:

[0106] (1) Sort the unexecuted freight by freight from high to low.

[0107] (2) Select the freight in turn and insert it into the position that maximizes the change in revenue.

[0108] The minimum opportunity insertion operator step:

[0109] (1) Calculate the number of applicable ships and time windows for the freight, and sort by fewer applicable ships and shorter time windows.

[0110] (2) Select the freight with fewer allocation opportunities in turn and insert it into the position with the smallest regret value.

[0111] (3) Repeat until all freight is arranged.

[0112] The risk minimum insertion operator step:

[0113] (1) Calculate the delay expectation of the freight, and sort from low to high.

[0114] (2) Select the low-risk freight in turn and insert it.

[0115] (3) Repeat until all freight is arranged.

[0116] Step five, two-stage solution generation:

[0117] According to the one-stage solution, a recovery scheme is designed and the adjustment cost is calculated for the possible interference scenarios. For each scenario, a recovery network is constructed and the Dijkstra algorithm is used to find the shortest path to obtain the optimal recovery scheme.

[0118] The recovery network construction method is as follows:

[0119] For each affected cargo f in scenario k, a recovery network is constructed Wherein, the node N includes the loading node, the unloading node and the start node of the next task (or the end node of the planning period). The arc A represents the sailing between ports or the loading and unloading task performed in the port. The weight C of each arc represents the cost of performing the task, including the additional cost of shortening the time and delay.

[0120] Loading and unloading node processing: loading node: the node actually starting loading after interference and the node set completing the loading task. Unloading node: the node set starting unloading and completing the unloading task, including the node arriving at the planned time and the delay time.

[0121] Path weight calculation: the weight of the arc is dynamically adjusted according to the length of the shortened port time or the delay time, including the additional fuel cost and the delay cost.

[0122] Shortest path finding: the Dijkstra algorithm is used to find the shortest path from the start node to the target node in the recovery network, which is the optimal recovery scheme.

[0123] The specific steps of generating the optimal recovery scheme are as follows:

[0124] (1) input the planned scheduling scheme MF and the set of possible scenarios K in the planning period;

[0125] (2) for each scenario k ∈ K, perform steps (3) to (6) iteration, and execute step (7) after completing all scenario traversal;

[0126] (3) initialize the adjustment cost AC of scenario k k = 0, for each cargo f performed in the planned scheduling scheme MF, judge whether its related task is affected by scenario k, if affected, perform steps (4) to (6) iteration, if not affected, skip;

[0127] (4) for each affected cargo f in scenario k, construct a recovery network

[0128] (5) find the shortest path by the Dijkstra algorithm, and the shortest path weight and are recorded as

[0129] (6) the compensation obtained for the disruption of the cargo f in the scenario k;

[0130] (7) stop the algorithm and output the expected AC = ∑ k∈K AC k |K|, |K| is the number of generated scenarios.

[0131] According to Figure 2 For a case of recovery network: a small example of the disruption of the cargo f in the scenario k causing the loading task to be delayed by 1 day. The ship needs to complete the loading of the cargo at the port SHA, then transport it to the port TYO and complete the unloading, and finally sail to the port PUS to start the next cargo task according to the original schedule. The recovery network includes 3 selectable complete loading nodes s2_0, s2_1 and s2_2, 3 start unloading nodes t1_0, t1_1 and t1_2, and 3 complete unloading nodes t2_0, t2_1 and t2_2. When reaching the s2_2 node, it means that the loading time has been shortened by 2*0.5 days by selecting the arc (s1, s2_2), which eliminates the disruption caused by the delayed start of the loading by 1 day. The weight of the arc (s1, s2_2) is the cost paid for shortening the loading time by 2*0.5 days, but the subsequent unloading tasks can be performed according to the original schedule, corresponding to the arcs (s2_2, t1_0), (s1_0, t2_0) and (t2_0, n0), whose weights are all 0. Similarly, the arc (s1, s2_0) represents loading at the original speed, and the weight of this arc is 0. The weight of the arc (s1_0, t1_0) is the additional fuel cost caused by acceleration, and the weight of the arc (s2_0, t1_1) is the additional fuel cost paid for acceleration and the delay cost paid for the delay of 1*0.5 days.

[0132] Step 6, iterative optimization:

[0133] (1) initialization: generate an initial solution;

[0134] (2) destruction: select a removal operator to destroy the current solution;

[0135] (3) repair: select a repair operator to repair the destroyed solution to generate a temporary solution;

[0136] (4) evaluation: evaluate the temporary solution obtained and compare it with the current solution;

[0137] (5) update: if the temporary solution is better than the current solution, then the new solution is taken as the current solution, otherwise the worse solution is accepted with a certain probability. Repeat the above steps until the termination condition is reached.

[0138] The iterative process is referred to Figure 1The algorithm flow shown.

[0139] Step 7, report the output scheduling scheme to the shipping company operation management personnel, and select the final scheduling scheme by the shipping company. For the reference scheduling scheme given by the program, the operation management personnel can select and adjust according to the actual situation of the company operation (such as the state of the ship, the priority of the cargo, the port restriction, etc.), and usually it is suggested to select the scheduling scheme with high ranking.

[0140] In particular, in some preferred embodiments of the present application, a computer device is also provided, which comprises a memory and a processor, and a computer program stored in the memory, and the processor implements the steps of the tramp ship scheduling method considering uncertain port time in any of the above embodiments when executing the computer program.

[0141] In other preferred embodiments of the present application, a computer readable storage medium is also provided, which stores a computer program / instruction, and the computer program is executed by a processor to implement the steps of the tramp ship scheduling method considering uncertain port time in any of the above embodiments.

[0142] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium, and the computer program can include the processes of the above-mentioned tramp ship scheduling method considering uncertain port time when executed, which will not be repeated here.

[0143] The computer readable storage medium covers various types, including persistent and non-persistent, portable and fixed. These media realize information storage through different technical means, and the content can be machine instructions, data structures, program modules or other types of data. Some typical computer storage medium examples include: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), various types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory and other storage technologies, compact discs (CD-ROM), digital video discs (DVD) and other optical storage media, magnetic tapes, magnetic disks and other magnetic storage devices, and other non-transmission media for storing information accessible to computing devices. It should be noted that the computer readable medium described herein does not include temporary storage media such as modulated data signals and carriers.

[0144] Those skilled in the art should further appreciate that the module's operation is able to be implemented by relying on the prior art protocol or program itself, and does not depend on a new computer program. The units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0145] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be implemented in hardware, software executed by a processor, or a combination of both. The software module can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0146] The above description is only the preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for scheduling of tramp ships considering uncertainty in port times, characterized in that, Specifically comprising: Step 1: Collecting ship data, cargo data, port data and possible interference event data, and preprocessing; Step 2: Based on the limited scenario support set of interference events mastered by the shipping company, Monte Carlo simulation is used to generate random interference scenarios within the planning period. The more scenarios generated, the more comprehensive the decision-making, and the longer the solution time will be; Step 3: Adopting a greedy strategy, preferentially allocating contract cargo, then allocating spot cargo, and meeting load limit, time window constraint and speed limit, generating an initial scheduling scheme MF, and then generating adjustment scheme MFR(k) under different scenarios k through path recovery algorithm; Step 4: One-stage solution generation: maximizing the shipping company's revenue as the goal, iteratively optimizing the current solution through destruction and repair operators to generate an initial scheduling plan; Step 5: Two-stage solution generation: for each interference scenario, a recovery network is constructed, and the shortest path algorithm is used to calculate the adjustment cost and generate a recovery scheme; Step 6: Repeat steps 4 and 5, continuously optimize the scheduling scheme when the maximum number of iterations is not reached, evaluate the temporary solution generated for each iteration, if the temporary solution is better than the current solution, update the current solution; otherwise, accept the poor solution according to the set probability to avoid falling into local optimum; when the maximum number of iterations is reached, output the final scheduling scheme containing transportation path, speed and recovery strategy.

2. The tramp ship scheduling method of claim 1, wherein, The data in step 1 includes: the maximum load limit of each ship, the port where the ship is located at the beginning of the planning period, the available time, the light weight, the economic speed, the maximum speed, the type of cargo, the weight of the cargo, the revenue obtained by transporting the cargo, the loading port and unloading port of the cargo, the loading and unloading time window and time requirement, the distance between ports, the type of event, the probability of occurrence, the impact degree, the compensation information, and the fuel price.

3. The tramp ship scheduling method of claim 1, wherein, Step 2 generates scenarios through Monte Carlo simulation based on the limited scenario support set of interference events mastered by the shipping company. Each scenario contains the following information: the port where the interference event occurs, the time period when the interference event occurs, the impact of the interference event on loading and unloading tasks, i.e. the number of days delayed, and the output scenario set is used for model solution.

4. The tramp ship scheduling method of claim 1, wherein, The greedy strategy used in step 3 specifically includes: Prioritize contract cargo allocation, arrange by time window, and allocate cargo in sequence to ensure that the maximum load limit of each allocated ship is not exceeded, and the loading and unloading time window constraints and speed limits are met; Each piece of cargo is allocated to the ship that spends the least fuel cost from the start of the available port to the loading port of the cargo; The start available time and port of the allocated ship are updated to the unloading time window and unloading port of the last item of cargo in the cargo sequence; After the contract cargo is allocated, the spot cargo is allocated from high to low freight, and each piece of contract cargo is allocated to the cargo sequence and corresponding position that can obtain the highest profit difference while meeting the load limit, time window requirement and speed limit; Cargo that cannot find a suitable allocation position will be canceled; After obtaining MF, generate adjustment scheme MFR(k) under different scenarios k through path recovery algorithm.

5. The tramp ship scheduling method of claim 1, wherein, The destruction operator in step 4 includes a random removal operator, a shaw removal operator, a worst removal operator, and a highest risk removal operator; and the repair operator includes a greedy insertion operator, a minimum opportunity insertion operator, and a minimum risk insertion operator.

6. The tramp ship scheduling method of claim 1, wherein, The construction of the recovery network specifically comprises: (1) For each affected cargo f in scenario k, construct a recovery network The nodes N include loading nodes, unloading nodes, and start nodes of next tasks or end nodes of planning periods; the arcs A represent sailing between ports or loading and unloading tasks performed at ports; the weights C are weights of each arc representing the cost of performing the task, including the cost of shortening time and additional cost of delay; (2) Loading and unloading node processing: loading nodes: a node set actually starting loading after being interfered and completing the loading task; unloading nodes: a node set starting unloading and completing the unloading task, including nodes arriving at the planned time and the delayed time; (4) Path weight calculation: the weight of an arc is dynamically adjusted according to the length of the shortened port time or the delayed time, including an additional fuel cost, a port cost and a delay cost; (5) Shortest path finding: a Dijkstra algorithm is used to find a shortest path from a starting node to a target node in the recovery network, and the path is the best recovery scheme.

7. The tramp ship scheduling method of claim 6, wherein, The generation of the recovery scheme specifically comprises: (1) inputting a planned scheduling scheme MF and a scenario set K possibly appearing in a planning period; (2) for each scenario k in K, iteration of steps (3) to (6) is performed, and step (7) is executed after all scenarios are traversed; (3) initialize the adjustment cost AC under scenario k k = 0, for each freight f executed by the planned scheduling scheme MF, judge whether its related task is affected by scenario k, if affected, execute the iteration of steps (4) to (6), if not affected, skip; (4) For each scenario k, construct a recovery network for the affected goods f (5) Find the shortest path by Dijkstra algorithm, the shortest path weight and record as (6) the compensation obtained for the goods f under the scenario k disturbed; (7) Stop the algorithm, output the expected AC = ∑ k∈K AC k |K|, |K| is the number of generated scenarios.

8. The tramp ship scheduling method of claim 1, wherein, The target function in step 6 is: Wherein, the first item of the target function is freight income, the second item and the third item are fuel cost and port cost, and the fourth item is an expected adjustment cost, that is, a target function of the second stage model; is the additional cost of adjusting the solution after the one-stage decision variable is determined under all scenarios k, m represents the shipping company's ships, f represents the cargo, K represents the set of scenarios generated by Monte Carlo simulation, and the probability p of each scenario appearing is the reciprocal of the number of scenarios; is 1 when ship m transports cargo f through arc (s, t), and 0 otherwise, (s, t) e A(f); is 1 when ship m is empty from node s to node t, and 0 otherwise; is 1 when ship m performs the transportation demand of cargo f, and 0 otherwise; is 1 when ship m is at anchor from node s to node t, and 0 otherwise;d f is 1 when the transportation demand of cargo f is accepted, and 0 otherwise; is the fuel cost consumed by ship m from node s to node t, and γ is the risk preference, which is 0-1, and the greater the value indicates the higher the degree of attention to risk in the decision-making process. where the first term is the economic compensation obtained by shipping company in scenario k, the second term is the cost paid for shortening the port time, and the third term is the cost paid for delaying the delivery of goods and adjusting the speed; is a 0-1 variable, and is 1 if the shipping work of goods f through arc (s, t) is adjusted to (i, j) in scenario k, otherwise 0, (s, t) e A1(f) U A'1(f), (i, j) e A3(s, t), the work includes two shipping segments of transporting goods f from the loading port to the unloading port and leaving the unloading port; is a continuous variable, and represents the shortened loading time of goods f at the loading port; is a continuous variable, and represents the shortened unloading time of goods f at the unloading port; represents the additional fuel cost paid for completing the original planned work of arc (s, t) through arc (i, j) in scenario k; represents the additional delay cost paid for completing the original planned work of arc (s, t) through arc (i, j) in scenario k; represents the shortened time of completing the shipping work of goods f through arc (i, j) in scenario k compared to the original plan, and ct represents the cost paid for shortening each unit of loading and unloading time.

9. A computer apparatus / device / system comprising a memory, a processor, and a computer program stored on the memory, characterized in that: The processor executes the computer program to implement the steps of the method in any one of claims 1 to 8.

10. A computer readable storage medium having stored thereon computer programs / instructions, characterized in that: The computer program / instruction is executed by the processor to implement the steps of the method in any one of claims 1 to 8.

Images