Container liner transportation scheduling control method considering environmental pollution and transportation efficiency

Abstract

The invention discloses a container liner transportation scheduling control method considering environmental pollution and transportation efficiency, and relates to the technical field of ship deployment control in shipping. The method includes the following steps that the number of containers entering a port, the number of containers needing to leave the port, the number of ships of each type, and ship loss and pollution emission generated during operation of the ships of each type on each ship line of ship company service are collected; the method comprises the following steps: constructing a container liner transportation deployment and scheduling control model 1 VDS1 of homogeneous ships based on the number of routes served by a ship company and a port needing to be served by each route; according to the transportation scheduling control method, a ship company can reasonably allocate each ship, the ship capacity and port resources are utilized to the maximum extent, efficient ship scheduling can be achieved among different ports, the shipping efficiency is remarkably improved, the shipping time is shortened, and the discharge amount of environmental pollutants is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of ship deployment control in shipping, and in particular to a container liner transportation dispatching control method taking environmental pollution and transportation efficiency into consideration. Background Art

[0002] In recent years, the development of international trade has promoted the rapid growth of the container shipping industry. According to statistics, container shipping has become one of the most important modes of transportation in international trade, and its proportion continues to grow. At the same time, with the deepening of globalization, people increasingly need to obtain a wider range of goods and services through international trade. Therefore, the container shipping industry must continuously improve the operational efficiency of container ships to meet the growing demand for freight. With the increase in the number of container ships and the expansion of capacity, the number of container shipping companies has continued to increase. Therefore, the competition among them has become increasingly fierce. In order to remain competitive, container shipping companies need to continuously control their operating technology and improve efficiency to meet customer needs. Therefore, research on the deployment and scheduling of container liners has become increasingly important. In order to achieve the effectiveness of cargo transportation, shipping companies must reasonably allocate each ship and maximize the use of ship capacity and port resources to ensure efficient operation and customer satisfaction.

[0003] At present, domestic and foreign researchers have studied the deployment and scheduling control of container liner shipping. Wang et al. (2012) studied the optimal sailing speed of container ships in each route and section of the liner shipping network. Branchini et al. (2015) established a mixed integer linear programming model by representing contract voyages and spot voyages as nodes of a directed graph, and finally obtained the optimal solution through the CPLEX solver. Wang et al. (2021) considered the selection of the most suitable ship to participate in the operation among candidate ships with different capacity, operating technology and fuel consumption, and adjusted their order, schedule and sailing speed in the route to improve transportation efficiency. The research results of the above scholars have made corresponding contributions to the deployment and scheduling of container liner shipping, but the factors considered are not comprehensive enough and the exact algorithm cannot be used to solve the model. Based on the above research background and practical problems, in order to comprehensively consider various factors, improve transportation efficiency, reduce pollutant emissions and control ship operation, it is urgent to solve the problem of ship deployment and scheduling control in container liner shipping. Summary of the invention

[0004] The object of the present invention is to provide a container liner transportation scheduling control method taking environmental pollution and transportation efficiency into consideration, so as to solve the problems existing in the prior art mentioned in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A container liner transportation scheduling control method considering environmental pollution and transportation efficiency comprises the following steps:

[0007] S1: Collect the number of containers entering the port, the number of containers that need to leave the port, the number of ships of each type, the ship losses and pollution emissions of each type of ship during operation on each route served by the shipping company;

[0008] S2: Based on the number of routes served by the shipping company and the ports that each route needs to serve, a container liner transportation deployment and scheduling control model VDS1 for homogeneous ships is constructed. The type of ships, number of ships and speed plan for each section that the shipping company needs to invest in each route are calculated based on the collected data;

[0009] S3: Add constraints on different types of ships on the basis of the control model VDS1. Different types of ships are uniformly constrained in their speed ranges according to their specifications. On the basis of the control model VDS1, a container liner transportation deployment and scheduling control model VDS2 considering ships of different specifications is constructed;

[0010] S4: According to the established container liner shipping deployment and scheduling control model, the seven designed algorithms are combined for solution, and the optimal solution is obtained after comparative analysis, thus obtaining the optimal ship deployment and scheduling solution. Among them, the seven algorithms are genetic algorithm, simulated annealing algorithm, branch and bound algorithm, genetic algorithm combined with branch and bound algorithm, simulated annealing algorithm combined with branch and bound algorithm, Gurobi solver, and Gurobi solver combined with branch and bound algorithm.

[0011] Preferably, in S2, by formula P k =M k -D k +E r -F r -B r -G r Obtain the type and number of ships that the shipping company needs to invest in each route, as well as the speed plan for each section; where M k The value of allocating a ship to other shipping companies because the ship's capacity is too small or too large to operate on the shipping company's route is generated. k It represents the loss of the shipping company incurred by renting ships from other companies because the existing number of ships does not meet the shipping capacity. r F represents the value brought by the efficient transportation of containers by shipping companies. r Indicates the loss of the ship during operation, B r Represents the amount of pollution emissions, G rrepresents the penalty incurred when the shipping company cannot transport the container itself due to some reasons, k represents the type of ship, and r represents the route served by the ship.

[0012] Preferably, in S3, according to the capacity data of the ship, the various types of ships can be divided into three specifications: if the capacity of the ship is less than M, then If the ship's capacity is greater than M and less than L, then If the ship's capacity is greater than L, then Among them, M is the capacity of the smallest medium-sized ship, and L is the capacity of the smallest large ship. A binary decision variable indicating that the ship specification is a small ship, A binary decision variable indicating that the ship specification is a medium-sized ship, is a binary decision variable indicating that the ship specification is a large ship, and k represents the ship type.

[0013] Preferably, in S2, the number of type k ships owned by the shipping company itself is l k 、The number of k-type ships that the shipping company needs to rent k , the number of ships that can be chartered out q k , the number of type k ships owned by the shipping company deployed on route r rk , the number of k-type ships that the shipping company needs to rent on route r rk and the sum of the number of type k ships operating on the route b rk ;

[0014] First, for the number j of type k ships that the shipping company needs to rent k The following constraints are set: k ≤L k This is to achieve the effectiveness of cargo transportation. Shipping companies must reasonably allocate each ship and are not allowed to rent ships beyond the limit. k It represents the maximum number of k-type ships allowed to be rented. It can further calculate the loss D of the shipping company when it rents ships from other companies because the existing number of ships does not meet the shipping capacity. k , according to the formula Calculate, where It represents the value that chartering a K-type ship brings to a third-party shipping company;

[0015] Then, step 2 involves the number of ships leased out, q k , according to the formula q k = l k -∑ r x rk +j kCalculation, further can be calculated to calculate the value M generated by allocating this type of ship to other shipping companies k , according to the formula Calculate, where represents the value that leasing a type k ship can bring to the shipping company; secondly, the number of type k ships owned by the shipping company deployed on route r is x rk The following constraints are set to limit the number of type K ships owned by the shipping company deployed on route r to not exceed the total number of type K ships owned by the shipping company: In addition, the number of k-type ships that the shipping company needs to rent on route r is z. rk The following constraints are set: The number of Type K vessels chartered on route R must not exceed the total number of Type K vessels chartered by the shipping company; in addition, the number of vessels operating on the route B is also involved. rk , the constraint on it is: 1≤∑ r b rk ≤l k +j k Further, the ship loss and total pollution emissions generated during the operation of the ship can be calculated. r , according to the formula Calculate, where K represents the set of all available ship types, It represents the ship loss and pollution emission generated by a K-type ship when operating on route R.

[0016] Preferably, S2 involves a penalty G caused by the shipping company being unable to transport the container itself due to some reasons. r This part is mainly generated by purchasing container slots from ships not operated by the company. The number of container slots purchased from ships not operated by the company is n od According to the formula n od =ξ od -g od , with the following constraints: Further calculation of G r , according to the formula Calculate, where ξ od represents the number of containers that need to be transported from port o to port d on route r, θ od represents the slot purchase volume per unit container from port o to port d, g od represents the number of containers transported from port o to port d by ship type k on route r, where o and d represent the ports on route r, and V represents the set of all ports;

[0017] S2 also involves the value E brought by the efficient transportation of containers by the shipping company. r , according to the formula Calculate, where represents the number of containers actually transported from port o to destination port d along route r, It represents the value brought to the shipping company by transporting one unit of container from the departure port o to the destination port d on route r;

[0018] Finally, it also involves pollution emissions, using the linear formula The time required for a ship to complete segment i can be calculated using the formula Calculate, so there are the following constraints: Furthermore, the pollution emissions of ships during operation are r The formula Calculate, where represents the slope of the tangent line on segment i, τ i represents the amount of pollution discharged by the ship after completing the voyage segment i, represents the intersection of the tangent line and the y-axis, Indicated by τ i The amount of fuel required to complete segment i in time, i represents the segment in the route, p represents the secant of the fuel amount, represents the departure time of type k ship from port i on route r, It indicates the time when the type k ship on route r arrives at port i-1. represents the stay time of type k ship on route r at port i, η r,i is a binary variable, which is 1 if port i is the starting port of route r; otherwise, it is 0. bunk Indicates the amount of pollutant emissions produced per unit amount of fuel.

[0019] Preferably, in the container liner transportation deployment and scheduling control model VDS1 of homogeneous ships, since the types and number of ships available in busy container liner transportation are very limited, the types of ships deployed on each route must be constrained first, and the constraint conditions are: ∑ k f rk =1 and x rk ≤Mf rk , where f rk is a binary variable, which is 1 if the k-type ship is deployed on route r, otherwise it is 0. R represents the set of routes, and M is a very large number;

[0020] Secondly, the number of ships deployed on each route must be constrained so that they cannot be greater than the sum of the number of ships of the same type owned by the shipping company and the number of ships that can be rented. At the same time, it must be ensured that the number of ships of each type used for rent cannot be greater than the number of ships rented from other shipping companies. While constraining the number of ships, it is also necessary to ensure that the service frequency of the shipping company meets the service frequency required by the ports on the route, ensuring that the goods can be transported efficiently and delivered to customers in a timely manner. That is, the following constraints need to be met:

[0021]

[0022] Among them, u ri represents the distance of the i-th segment on route r, v ri represents the speed of the i-th segment on route r, δ r,i,k represents the time that a type k ship spends at port i on route r, b rk represents the number of type k ships deployed on route r, but this constraint is not a linear constraint, so a decision variable O is introduced ri =1 / v ri , then the above constraints will become the following formula:

[0023]

[0024] Among them, O ri represents the inverse of the speed of the i-th segment on route r;

[0025] In terms of penalties, the number of purchased slots needs to be limited so that the number of slots is greater than 0 and less than the number of containers that need to leave the port. At the same time, it is ensured that after entering the port, the ship must leave the port at the same time, and the inflow and outflow of containers must be balanced. Therefore, the flow conservation constraints on containers and service variables are set;

[0026] The flow conservation constraints of the container are as follows: as well as in, represents the number of containers stored on the ship of type k departing from port o on the i-1 leg of route r, represents the number of containers loaded by a vessel of type k starting from port o and at the jth port of call on route r, I represents the number of containers unloaded by a ship of type k starting from port o at the jth port of call on route r. rd represents the set of port indices pointing to a specific port d in route r;

[0027] The flow conservation constraints of service variables are subject to: as well as k∈K, where is a binary variable, which is 1 if the k-type ship sails from port o to port d; otherwise, it is 0. In addition, it links the time when the ship arrives at port i to the time when the ship arrives at port i+1. The specific constraints are: in, It indicates the time when the k-type ship on route r arrives at port j. represents the stay time of type k ship at port j on route r, It represents the sailing time of type k ship on route r to port j;

[0028] In terms of speed, there are the following constraints: in, represents the fastest sailing speed of the ship on the i-th segment of route r, represents the slowest sailing speed of the ship on the i-th segment of route r, v ri represents the speed of the ship on the i-th segment of route r, I r represents the set of all ports on route r;

[0029] Finally, the following non-negativity constraints are imposed on all other decision variables:

[0030]

[0031] Preferably, in S3, first, the constraints on speed in the control model VDS1 are rewritten as the following constraints:

[0032]

[0033] in, represents the speed of the k-type ship in the i-th segment of route r, Respectively represent the minimum speed of small ships, medium ships and large ships. Respectively represent the maximum speed of small ships, medium ships and large ships;

[0034] Then, the model sets the variables variable variable The non-negativity constraint of , and the constraint that each type of ship is only allowed to have one ship specification: as well as

[0035] Preferably, the control model 1 VDS1 and the control model 2 VDS2 are solved by genetic algorithm:

[0036] First, the genetic algorithm uses real number coding to randomly generate chromosomes. The coding method is as follows: a gene is used to represent a feasible ship deployment plan, where each sub-code is denoted as Y i , represents the i-th deployment scheme, each gene consists of three parts, among which, Y i1 ,Y i2, Y i3 is the first part of the gene, which controls the service route number, the number of ships deployed on the route, and the type of ships deployed on the route; Y i4 The second part of the gene is used to control the order of arrival at ports during a voyage served by the shipping company; the third part is Y i5 , used to describe the speed of the ship on the route, and then repeat these parts, only repeating the functions but not the values, until all route deployment plans are generated;

[0037] Secondly, in the process of calculating the fitness function, the randomly generated genes are first decoded to obtain the number of services provided by each ship type to the ith route and the speed of the ship in each section; once the number of ships deployed on the route is determined, the number of leased ships and whether there are leased ships can be determined; if there are leased ships, the value of the leased ships to the shipping company can be calculated using the formula in the model; the pollution emissions are calculated based on the known ship speed and the order of port visits using the calculation formula in the model, and at the same time, the value that can be brought to the shipping company through efficient transportation of containers can be calculated based on the number of containers transported by each port; finally, based on the cargo demand between given ports, it is calculated whether there is a penalty part for each port, so as to obtain the fitness value of the gene; in terms of selection operation, the roulette method is used to design the genetic algorithm selection operator;

[0038] Roulette is one of the commonly used selection operations in genetic algorithms. It normalizes the fitness value of individuals into selection probability and creates a roulette-like mechanism to select individuals. Individuals with higher fitness values ​​have a greater probability of being selected during the selection process, and thus have more chances to be retained to the next generation. Roulette is relatively simple and easy to implement. It can effectively retain excellent individuals and promote the evolution of the population. The process of roulette is as follows:

[0039] (1) Calculate the fitness value of each individual, because the fitness value reflects the quality of the individual;

[0040] (2) Normalize the fitness values ​​of all individuals so that their sum is 1. The purpose of normalization is to convert the fitness value into a selection probability. For example, Represents the jth chromosome individual, using To represent the fitness function value of the chromosome, the range of j should not exceed the maximum population size MAX size , that is, j∈{1,2,3,…,MAX size}, can be expressed by the following formula The probability of a chromosome being selected and inherited to the next generation:

[0041]

[0042] (3) Create a roulette wheel. The length of the roulette wheel is equal to the number of individuals in the population, and the size of the roulette wheel area occupied by each individual is proportional to its fitness value;

[0043] (4) performing several selection operations during the selection process, selecting one individual each time;

[0044] (5) Randomly generate a random number between 0 and 1, and then select individuals based on the roulette wheel area where the random number is located; the larger the roulette wheel area, the greater the probability that the individual will be selected;

[0045] (6) Repeat steps (4) and (5) until a sufficient number of individuals are selected;

[0046] Crossover operation takes into account that longer driving distances generate more pollutant emissions and ship rentals, so the specific crossover strategy is:

[0047] The present invention assumes that there are two genes A: (a1, a2, ..., a n ) and B: (b1,b2,…,b n ), where a1, a2, …, a n ∈I r and b1,b2,…,b n ∈I r , and a1≠a2≠…≠a n , b1≠b2≠…≠b n . If there are three ports x, y, and z, define a triangular distance function F(x, y, z) = d(x, y) + d(y, z) - d(x, z), where d(x, z) means the distance from port x to port z. The order of sequential crossover steps is as follows:

[0048] (1) Identify A as the basic gene and set B as the reference gene. It is necessary to find a reference port b in B. i (i=1,2,…,I r );

[0049] (2) Find port a in gene A j =b i ,a k =bi+1 , and then use the defined triangular distance function to calculate F(a j ,b i+1 ,a j+1 ) and F(a k-1 ,a k ,a k+1 ), and compare the two calculated results. If the calculated F(a j ,b i+1 ,a j+1 )≥F(a k-1 ,a k ,a k+1 ), then gene A remains unchanged, if F(a j ,b i+1 ,a j+1 ) <F(a k-1 ,a k ,a k+1 ), then port a is deleted in gene A k , then in a j with a j+1 Add port b i+1 Get new genes

[0050] (3) Take the reference port b in the reference gene in turn i+1 ,…,b n ,b1,…,b i-1 ,Repeat step 2 until the final gene A1 is generated;

[0051] (4) Gene B is used as the basic gene and gene A is used as the reference gene, and steps 1-3 are repeated;

[0052] (5) The genes that have been crossed are judged. If the port served by the gene after crossing is not a port on one of the shipping company's routes, that is, there are ports on two routes, the gene is deleted to ensure that the ship serves a complete route;

[0053] Finally, in the mutation operation, the present invention adopts the crossover mutation method on the port stop sequence, but only the genes of the same route are selected for mutation during the mutation process.

[0054] Preferably, the control model 1 VDS1 and the control model 2 VDS2 are solved by a simulated annealing algorithm:

[0055] The idea of ​​using the simulated annealing algorithm to solve the container liner shipping fleet deployment and scheduling problem is: first, use the designed encoding rules to generate an initial solution, then design the rules for generating new solutions, and then combine the receiving function, annealing strategy and ending strategy of the simulated annealing algorithm to finally obtain the final solution; the specific implementation method is as follows:

[0056] In the simulated annealing algorithm, the encoding and decoding process is the process of determining the order of ports to be visited and the number and type of ships serving the route. Assume that 12 ships are deployed on route 1 and type 2 ships are selected to serve the route. If the port numbers on the route are 2, 5, 11, 9, and 8, and the random speed values ​​of each section are initialized within a reasonable range, then the encoding is [1, 2, 12, 5, 9, 2, 8, 11, 22, 25, 20, 19].

[0057] Based on the above encoding and decoding rules, the initial solution can be generated, and then a new solution is generated. When designing the new solution generation rules, it is divided into two parts: the first part is the order of service ports. The strategy adopted in this part is to randomly select the order of several ports on the route and swap them; the second part is the generation of speed on each segment. The speed of some segments is randomly generated in a reasonable range based on the distance.

[0058] Finally, the acceptance function, annealing strategy, and ending strategy of the simulated annealing algorithm are presented. The acceptance function in the simulated annealing algorithm is used to determine whether to accept a new solution as the current solution. It allows accepting inferior solutions with a certain probability during the search process to avoid falling into the local optimum. The simulated annealing algorithm selects the Metropolis criterion as the acceptance function:

[0059]

[0060] Among them, E(n+1) represents the newly calculated allocation plan of the type and quantity of ships on each route of the shipping company, and E(i) represents the last calculated allocation plan of the type and quantity of ships on each route of the shipping company. When E(i)≤E(n+1), it means that the allocation of the type and quantity of ships in the current plan is more reasonable and the transportation efficiency of goods is faster, so the current solution is adopted. On the contrary, if E(i)>E(n+1), it means that the allocation of the type and quantity of ships in the current plan or the transportation efficiency of goods is not as good as the previous solution, then θ=∪(0,1). If Then adopt the current solution, that is, adopt the worse solution;

[0061] Exponential annealing refers to the temperature decreasing exponentially, for example, T j =T0·α k , where T j is the temperature after the jth iteration, T0 is the set initial temperature, α is the annealing factor and α<1; the simulated annealing algorithm uses this method to construct the annealing function, that is, T j+1 =T j s, where s∈[0.95, 0.99];

[0062] Finally, the ending strategy of the simulated annealing algorithm is to set the maximum number of iterations, that is, it ends when the termination temperature is reached.

[0063] Preferably, the control model 1 VDS1 and the control model 2 VDS2 are solved by a branch and bound algorithm:

[0064] The branch-and-bound algorithm can effectively solve the mixed integer linear programming (MILP) problem. In order to quickly obtain the global optimal solution, two auxiliary methods are embedded on the basis of the algorithm: the first is to use the Gurobi solver as a preprocessor, and the second is to combine it with a heuristic algorithm.

[0065] In the first method, the Gurobi solver is used as a preprocessing method to solve the relaxed model. This process is performed before the branch and bound algorithm is used in order to greatly reduce the time required for the branch and bound algorithm to search. The relaxation of the model is mainly focused on the selection of ship type, which is separated from the binary decision variable f rk Relax to real decision variables, so that non-integer values ​​may appear in the solution process, which is also one of the parts where the branch and bound algorithm will go to branch, where f rk It means that if the k-type ship service route r is used, it is 1, otherwise it is 0; in the ship type part, only these non-integer ship types need to be branched; in the second algorithm combining the branch and bound algorithm with the heuristic algorithm, in order to speed up the search efficiency, the heuristic method is used to obtain the relatively optimal solution as the initial solution of the branch and bound algorithm, and then the solution is obtained by the branch and bound algorithm;

[0066] In the branch and bound algorithm, the pruning process is necessary, so special pruning rules are formulated for the algorithm. In the branch and bound algorithm, the search strategy used is the depth-first search strategy. This strategy is used because the depth-first search strategy can definitely find the global optimal solution, and the depth-first search is more efficient than the breadth-first search in terms of memory usage, because it only needs to store the nodes on a branch path, rather than the nodes of the entire layer; then the part of the ship selection in the code will be judged by an integer. If it is a non-integer, the upper bound value will be updated, otherwise the lower bound will be updated; then the constraints in the model will be judged, such as the constraints on speed and service frequency. If the above conditions are not met, the algorithm will take pruning measures; after calculating the target value, the corresponding coding part will be judged by an integer. If they are all integers and the target value is less than the lower bound, it means that no better solution will appear if the branch is further branched, so it will be pruned to stop the branch.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] The transport scheduling control method of the present invention enables shipping companies to reasonably allocate each ship, maximize the use of ship capacity and port resources, achieve efficient ship scheduling between different ports, significantly improve shipping efficiency, reduce shipping time, and reduce emissions of environmental pollutants. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 The present invention is a flowchart of a method for ship deployment and scheduling control applied to container liner shipping operations in an embodiment of the present invention.

[0070] Figure 2 This is a chromosome coding design diagram of the genetic algorithm in an embodiment of the present invention.

[0071] Figure 3 This is an example diagram of encoding and decoding of the simulated annealing algorithm in an embodiment of the present invention.

[0072] Figure 4 Flowchart of the genetic algorithm in the embodiment of the present invention.

[0073] Figure 5 Flow chart of simulated annealing algorithm in an embodiment of the present invention.

[0074] Figure 6 Flow chart of the branch and bound algorithm in an embodiment of the present invention.

[0075] Figure 7 Table showing the efficiency and accuracy of the seven algorithms for solving the model VDS1.

[0076] Figure 8 Table showing the efficiency and accuracy of the seven algorithms for solving the model VDS2.

[0077] Fig. 9 Error data table for solving model VDS1 for seven algorithms.

[0078] Fig.10 Table of error data for solving model VDS2 for seven algorithms.

[0079] Fig.11 Efficiency comparison table of seven algorithms for solving model VDS1.

[0080] Fig.12 The efficiency comparison table of seven algorithms for solving model VDS2 is given.

[0081] Fig.13 It is a value data table corresponding to the speed ranges of different ship types and the corresponding allocation plans. DETAILED DESCRIPTION

[0082] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0083] See also Figure 1-13 , the present invention provides the following technical solutions:

[0084] In order to evaluate the effectiveness of the two proposed models and six algorithms, the experiment was conducted using the real route between Asia and Northern Europe as the first scenario. First, the number of containers entering the port and the number of containers that need to leave the port are collected in real time on each route served by the shipping company. Then the capacity and speed range of different types of ships owned by the shipping company are collected and used as input. Subsequently, the container liner transportation deployment and scheduling control model VDS1 for homogeneous ships and the container liner transportation deployment and scheduling control model VDS2 considering different ship specifications are established.

[0085] For further information, see Figures 4 to 6 , using genetic algorithm, simulated annealing algorithm and branch-and-bound combined algorithm to solve model VDS1 and model VDS2 respectively, and then analyzing and evaluating the solution accuracy and efficiency of all algorithms, the evaluation is carried out through 10 different scenarios. In the present invention, these examples are named "K / R / Imax", where K, R and Imax represent the number of ship types, the number of segments and the maximum number of ports on the route respectively. Figure 7 and Figure 8 The results obtained by the six algorithms for each scenario and the corresponding running time are shown. The "Obj" column represents the value of the optimal ship deployment type and number on each route and the optimal route speed plan for each leg calculated by the shipping company using the above six algorithms. The "Time" column represents the CPU running time (in seconds), measured from the start of the algorithm calculation.

[0086] For further information, see Fig. 9 and Fig.10 ,The present invention finds that in terms of solution accuracy, the heuristic algorithm often cannot reach the optimal solution, and there is a certain gap with the optimal solution. Specifically, compared with the optimal result, the gap of the genetic algorithm is 13% on average, while the gap of the simulated annealing algorithm is 8%. In fact, in all scenarios except scenario 2, the simulated annealing algorithm is better than the genetic algorithm in solution accuracy.

[0087] On the other hand, when solving model VDS1, the simulated annealing algorithm reduced the computation time by 55.5% on average compared to the genetic algorithm. In model VDS2, the simulated annealing algorithm was 53.6% faster than the genetic algorithm on average. However, this pattern does not exist in smaller cases. For example, in instance 2 based on model VDS1, the simulated annealing algorithm took 4% more execution time than the genetic algorithm. Similarly, when testing instance 2 of model VDS2, the genetic algorithm was 6% faster than the simulated annealing algorithm. However, as the model size increases, it is found that the simulated annealing algorithm gradually surpasses the genetic algorithm in solution speed, which is consistent with the average solution speed trend mentioned initially. This is because the genetic algorithm performs more operations (such as crossover / mutation) than the simulated annealing algorithm and therefore takes longer to run.

[0088] In general, the simulated annealing algorithm saved 44% to 76% of the time when solving model VDS1. For solving model VDS2, the simulated annealing algorithm saved 4% to 71% of the computing time. Compared with the optimal solution of model VDS2, the simulated annealing algorithm has only a 0.4% error in the best case. However, the error of the better solution calculated by the genetic algorithm is slightly higher (1%), which is less accurate than the simulated annealing algorithm.

[0089] See also Fig. 9 and Fig.10 , respectively, show the gaps of seven algorithms by solving two models VDS1 and VDS2. The gaps of heuristic algorithms (i.e., GA and SA) are quite large, ranging from 0.4% to 38%. In particular, B&B sometimes produces small gaps, such as 0.001% (K8 / R5 / Imax) and 0.002% (K5 / R3 / Imax), while the gaps of other algorithms are zero. Further, the solving efficiency of Gurobi solver and Gurobi+B&B is compared. In the case of small-scale instances, such as solving the first 5 scenarios of model VDS1 and model VDS2, the speed of Gurobi solver is always better than Gurobi+B&B. In the case of the largest gap, Gurobi solver is 1.84 seconds faster than Gurobi+B&B, saving 82.5% of the time. However, in extremely small-scale scenarios, it is found that the efficiency of Gurobi solver and Gurobi+B&B is similar. As the case size increases, it is observed that the solving efficiency of Gurobi+B&B gradually exceeds that of Gurobi. Specifically, when the model size is expanded to the last three instances of model VDS2, the solving efficiency of Gurobi+B&B is significantly improved. The sixth instance saves 0.14 seconds, accounting for 31.8% of the total time, the seventh instance saves 0.65 seconds, accounting for 51.6% of the total time, and the eighth instance saves 5.71 seconds, accounting for 87.7% of the total time.

[0090] Further, the solving effects of the other three algorithms are analyzed: B&B, GA+B&B and SA+B&B. In large-scale cases, namely the 6th to 10th instances of model VDS1 and the 5th to 8th scenarios of model VDS2, the B&B algorithm outperforms GA+B&B and SA+B&B, saving an average of 7.4% and 11.6% of the solving time, respectively. Specifically, the B&B algorithm shows advantages in solving some instances with smaller model sizes, such as solving the 3rd, 4th, and 5th instances of model VDS1 and the 4th instance of model VDS2. Compared with GA+B&B, an average of 51.7% of the time is saved, and compared with SA+B&B, an average of 54.3% of the time is saved. For the solution of model VDS1, compared with GA+B&B, 0.6% to 99.4% of the time can be saved, and compared with SA+B&B, 1.1% to 99.8% of the time can be saved. When solving model VDS2, the time savings relative to GA+B&B range from 0.8% to 92.1%, and the time savings relative to SA+B&B range from 2.9% to 98.8%.

[0091] For further information, see Fig.11 and Fig.12 , the present invention can also be used to derive and compare the computational efficiency of all algorithms. Based on the running time of GA, for small-scale transportation scenarios, the branch and bound algorithm outperforms other algorithms (0.004% for VDS1 and 0.052% for VDS2). The Gurobi solver significantly outperforms other algorithms in solving medium-scale cases (VDS1 from 0.006% to 0.05%; VDS2 from 0.023% to 0.214%). For large-scale transportation scenarios, Gurobi+B&B has the highest efficiency (from 0.053% to 0.065% in the VDS2 model). The second model VDS2 further estimates the adaptability of multiple types of ships characterized by changes in sailing speed. Specifically, small, medium and large ships perform [3-23] knots / nautical mile, [13-33] knots / nautical mile and [23-42] knots / nautical mile, respectively. The three speed intervals are divided into six levels to test the effectiveness. The inventor found that even if the speed continues to increase, the value rate corresponding to the allocation plan cannot maintain high growth. In other words, the value rate corresponding to the allocation plan drops from 41.9% (first speed level) to 4.9% (highest speed level).

[0092] Whether the sailing speed of the ship is fixed or different, two models VDS1 and VDS2 are formed respectively. Although the latter leads to a more complicated problem, the speed dynamics can make the ship deployment more advantageous. So further, the present invention can quantitatively compare the two models, and find out whether the speed dynamics can generate benefits according to the objective function. The data of the first 8 scenarios in the comparison model VDS1 and all scenarios in the model VDS2 are selected. Observing the comparison of the benefits calculated by the two models in the 8 scenarios, with the increase of routes and ship types, the proportion of additional value benefits is also rising rapidly, from 15% to 85%.

[0093] The present invention can also analyze the impact of navigation speed on the type and number of ships deployed on each route and the navigation speed of each section on the routes operated by the shipping company. The ship size is set from small to large, a speed gradient of ±2 knots per nautical mile is implemented, and six different shipping speed intervals are designed. The present invention finds that as the speed increases, the growth of the value corresponding to the type and number of ships on each route and the navigation speed allocation scheme for each section is not linear. In the initial stage, the growth rate (41.9%) is faster when the speed level is 2. As for the 3rd, 4th and 5th speed categories, the growth rate is slightly higher than 10%. Finally, if the actual speed increases to the maximum value, that is, the 6th speed category, the value percentage corresponding to each allocation scheme is the smallest, which is 4.9%. This finding shows that when the ship speed is slow, a slight speed increase can greatly save the fleet size. Therefore, the shipping company has achieved significant value in the allocation scheme corresponding to the 1st speed category. Obviously, the percentage increase in speed per knot is 8.3%. However, the high-speed solution shows that the number of ships cannot be saved again. Therefore, as the last speed category, the value rate of the allocation scheme corresponding to each additional knot of speed decreases to 0.9%.

[0094] Finally, the present invention can also analyze the gaps in the values ​​corresponding to the types and numbers of ships deployed on each route and the sailing speed allocation schemes for each segment calculated by all algorithms, and it is found that only the heuristic algorithms (i.e., GA and SA) have large gaps, ranging from 0.4% to 38%. B&B occasionally produces small gaps, such as 0.001% and 0.002%. Apart from this, the gaps of other algorithms are zero. The computational efficiency of all solutions is compared using the running time of GA as a benchmark. For small-scale cases, the B&B algorithm outperforms other algorithms (0.004% for VDS1 and 0.052% for VDS2). However, for medium-scale cases, the Gurobi solver significantly outperforms other algorithms (from 0.006% to 0.05% for VDS1 and from 0.023% to 0.214% for VDS2). Finally, for larger cases, Gurobi+B&B achieves the best efficiency (from 0.053% to 0.065% for VDS2). In summary, the present invention is applied to the control effect of ship deployment and scheduling in container liner shipping operations under the background that each ship must be reasonably allocated to maximize the use of ship capacity and port resources to ensure efficient operation and customer satisfaction. First, the number of containers entering the port and the number of containers that need to leave the port are collected in real time at each port on each route served by the shipping company, and then the capacity and speed range of different types of ships owned by the shipping company are collected. A container liner shipping deployment and scheduling control model is constructed according to the number of routes served by the shipping company and the ports that each route needs to serve. By controlling the number and speed range of ships of each specification serving each route, the algorithm designed in the present invention is combined to solve the problem, so as to obtain the best ship type used on each route on the shipping company's operating route, the minimum number of ships deployed, the best route speed for each section, and the ship deployment and scheduling plan with the least environmental pollution; compared with the prior art, the present invention comprehensively considers more control factors, can more accurately control the ship deployment and scheduling plan, improve transportation efficiency, and reduce the emission of environmental pollutants.

[0095] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A container liner transportation dispatching control method considering environmental pollution and transportation efficiency, characterized in that: The following steps are involved: S1: Collect the number of containers entering the port, the number of containers that need to leave the port, the number of ships of each type, the ship losses and pollution emissions of each type of ship during operation on each route served by the shipping company; S2: Based on the number of routes served by the shipping company and the ports that each route needs to serve, a container liner transportation deployment and scheduling control model VDS1 for homogeneous ships is constructed. The type of ships, number of ships and speed plan for each section that the shipping company needs to invest in each route are calculated based on the collected data; S3: Add constraints on different types of ships on the basis of the control model VDS1. Different types of ships are uniformly constrained in their speed ranges according to their specifications. On the basis of the control model VDS1, a container liner transportation deployment and scheduling control model VDS2 considering ships of different specifications is constructed; S4: Based on the established container liner shipping deployment and scheduling control model, the seven designed algorithms are combined to solve the problem, and the optimal solution is obtained after comparative analysis, thus obtaining the optimal ship deployment and scheduling solution.

2. A container liner transportation dispatching control method considering environmental pollution and transportation efficiency according to claim 1, characterized in that: In S2, the formula P k =M k -D k +E r -F r -B r -G r Obtain the type and number of ships that the shipping company needs to invest in each route, as well as the speed plan for each section; where M k The value of allocating a ship to other shipping companies because the ship's capacity is too small or too large to operate on the shipping company's route is generated. k It represents the loss of the shipping company incurred by renting ships from other companies because the existing number of ships does not meet the shipping capacity. r F represents the value brought by the efficient transportation of containers by shipping companies. r Indicates the loss of the ship during operation, B r Represents the amount of pollution emissions, G r represents the penalty incurred when the shipping company cannot transport the container itself due to some reasons, k represents the type of ship, and r represents the route served by the ship.

3. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 1 is characterized in that: In S3, according to the capacity data of the ship, various types of ships can be divided into three specifications: if the capacity of the ship is less than M, then If the ship's capacity is greater than M and less than L, then If the ship's capacity is greater than L, then Among them, M is the capacity of the smallest medium-sized ship, and L is the capacity of the smallest large ship. A binary decision variable indicating that the ship specification is a small ship, A binary decision variable indicating that the ship specification is a medium-sized ship, is a binary decision variable indicating that the ship specification is a large ship, and k represents the ship type.

4. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 2 is characterized in that: In S2, the number of type k ships owned by the shipping company itself is l k 、The number of k-type ships that the shipping company needs to rent k , the number of ships that can be chartered out q k , the number of type k ships owned by the shipping company deployed on route r rk , the number of k-type ships that the shipping company needs to rent on route r rk and the sum of the number of type k ships operating on the route b rk ; First, for the number j of type k ships that the shipping company needs to rent k The following constraints are set: This is to achieve the effectiveness of cargo transportation. Shipping companies must reasonably allocate each ship and are not allowed to rent ships beyond the limit. k Allows rental The maximum number of ships of this type can be further calculated to calculate the loss of the shipping company in renting ships from other companies because the existing number of ships does not meet the shipping capacity. k , according to the formula Calculate, where It represents the value that chartering a K-type ship brings to a third-party shipping company; Then, step 2 involves the number of ships leased out, q k , according to the formula q k = l k -∑ r x rk +j k Calculate, and further calculate the value M generated by allocating this type of ship to other shipping companies k , according to the formula Calculate, where represents the value that leasing a type k ship can bring to the shipping company; secondly, the number of type k ships owned by the shipping company deployed on route r is x rk The following constraints are set to limit the number of type K ships owned by the shipping company deployed on route r to not exceed the total number of type K ships owned by the shipping company: In addition, the number of k-type ships that the shipping company needs to rent on route r is z. rk The following constraints are set: The number of Type K vessels chartered on route R must not exceed the total number of Type K vessels chartered by the shipping company; in addition, the number of vessels operating on the route B is also involved. rk , and its constraints are: Further, the ship loss and total pollution emissions generated during the operation of the ship can be calculated. r , according to the formula Calculate, where K represents the set of all available ship types, It represents the ship loss and pollution emission generated by a K-type ship when operating on route R.

5. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 2 is characterized in that: S2 involves the penalty G caused by the shipping company being unable to transport the container itself due to some reasons. r This part is mainly generated by purchasing container slots from ships not operated by the company. The number of container slots purchased from ships not operated by the company is n od According to the formula n od =ξ od -g od , with the following constraints: Further, we can calculate G r , according to the formula Calculate, where ξ od represents the number of containers that need to be transported from port o to port d on route r, θ od represents the slot purchase volume per unit container from port o to port d, g od represents the number of containers transported from port o to port d by ship type k on route r, where o and d represent the ports on route r and V represents the set of all ports; S2 also involves the value E brought by the efficient transportation of containers by the shipping company. r , according to the formula Calculate, where represents the number of containers actually transported from port o to destination port d along route r, It represents the value brought to the shipping company by transporting one unit of container from the departure port o to the destination port d on route r; Finally, it also involves pollution emissions, using the linear formula The time required for a ship to complete segment i can be calculated using the formula Calculate, so there are the following constraints: Furthermore, the pollution emissions of ships during operation are r The formula Calculate, where represents the slope of the tangent line on segment i, τ i represents the amount of pollution discharged by the ship after completing the voyage segment i, represents the intersection of the tangent line and the y-axis, Indicated by τ i The amount of fuel required to complete segment i in time, i represents the segment in the route, p represents the secant of the fuel amount, represents the departure time of type k ship from port i on route r, It indicates the time when the type k ship on route r arrives at port i-1. represents the stay time of type k ship on route r at port i, η r,i is a binary variable, which is 1 if port i is the starting port of route r; otherwise, it is 0. bunk Indicates the amount of pollutant emissions produced per unit amount of fuel.

6. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 2 is characterized in that: In the container liner transportation deployment and scheduling control model VDS1 of homogeneous ships, since the types and number of ships available in busy container liner transportation are very limited, the types of ships deployed on each route must be constrained first. The constraints are: and x rk ≤Mf rk , where f rk is a binary variable, which is 1 if the k-type ship is deployed on route r, otherwise it is 0. R represents the set of routes, and M is a very large number; Secondly, the number of ships deployed on each route must be constrained so that they cannot be greater than the sum of the number of ships of the same type owned by the shipping company and the number of ships that can be rented. At the same time, it must be ensured that the number of ships of each type used for rent cannot be greater than the number of ships rented from other shipping companies. While constraining the number of ships, it is also necessary to ensure that the service frequency of the shipping company meets the service frequency required by the ports on the route, ensuring that the goods can be transported efficiently and delivered to customers in a timely manner. That is, the following constraints need to be met: Among them, u ri represents the distance of the i-th segment on route r, v ri represents the speed of the i-th segment on route r, δ r,i,k represents the time that a type k ship spends at port i on route r, b rk represents the number of type k ships deployed on route r, but this constraint is not a linear constraint, so a decision variable O is introduced ri =1 / v ri , then the above constraints will become the following formula: Among them, O ri represents the inverse of the speed of the i-th segment on route r; In terms of penalties, the number of purchased slots needs to be limited so that the number of slots is greater than 0 and less than the number of containers that need to leave the port. At the same time, it is ensured that after entering the port, the ship must leave the port at the same time, and the inflow and outflow of containers must be balanced. Therefore, the flow conservation constraints on containers and service variables are set; The flow conservation constraints of the container are as follows: as well as in, represents the number of containers stored on the ship of type k departing from port o on the i-1 leg of route r, represents the number of containers loaded by a vessel of type k starting from port o and at the jth port of call on route r, I represents the number of containers unloaded by a ship of type k starting from port o at the jth port of call on route r. rd represents the set of port indices pointing to a specific port d in route r; The flow conservation constraints of service variables are subject to: as well as in, is a binary variable, which is 1 if the k-type ship sails from port o to port d; otherwise, it is 0. In addition, it links the time when the ship arrives at port i to the time when the ship arrives at port i+1. The specific constraints are: in, It indicates the time when the k-type ship on route r arrives at port j. represents the stay time of type k ship at port j on route r, It represents the sailing time of type k ship on route r to port j; In terms of speed, there are the following constraints: in, represents the fastest sailing speed of the ship on the i-th segment of route r, represents the slowest sailing speed of the ship on the i-th segment of route r, v ri represents the speed of the ship on the i-th segment of route r, I r represents the set of all ports on route r; Finally, the following non-negativity constraints are imposed on all other decision variables:

7. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 3 is characterized in that: In S3, first, the constraints on speed in the control model VDS1 are rewritten as the following constraints: in, represents the speed of the k-type ship in the i-th segment of route r, Respectively represent the minimum speed of small ships, medium ships and large ships. Respectively represent the maximum speed of small ships, medium ships and large ships; Then, the model sets the variables variable variable The non-negativity constraint of , and the constraint that each type of ship is only allowed to have one ship specification: as well as 8. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 1 is characterized in that: The genetic algorithm is used to solve the control model 1 VDS1 and the control model 2 VDS2: First, the genetic algorithm uses real number coding to randomly generate chromosomes. The coding method is as follows: a gene is used to represent a feasible ship deployment plan, where each sub-code is denoted as Y i , represents the i-th deployment scheme, each gene consists of three parts, among which, Y i1 ,Y i2, Y i3 is the first part of the gene, which controls the service route number, the number of ships deployed on the route, and the type of ships deployed on the route; Y i4 The second part of the gene is used to control the order of arrival at ports during a voyage served by the shipping company; the third part is Y i5 , used to describe the speed of the ship on the route, and then repeat these parts, only repeating the functions but not the values, until all route deployment plans are generated; Secondly, in the process of calculating the fitness function, the randomly generated genes are first decoded to obtain the number of services provided by each ship type to the i-th route and the speed of the ship in each section; once the number of ships deployed on the route is determined, the number of leased ships and whether there are leased ships can be determined; if there are leased ships, the formula in the model can be used to calculate the value brought by the leased ships to the shipping company; using the calculation formula in the model, the pollution emissions are calculated according to the known ship speed and the order of port visits, and at the same time, the value that can be brought to the shipping company through efficient transportation of containers can be calculated according to the number of containers transported by each port; finally, according to the cargo demand between given ports, it is calculated whether there is a penalty part for each port, so as to obtain the fitness value of the gene; in terms of selection operation, the roulette method is used to design the genetic algorithm selection operator.

9. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 1, characterized in that: The simulated annealing algorithm is used to solve the control model 1 VDS1 and the control model 2 VDS2: The idea of ​​using the simulated annealing algorithm to solve the container liner shipping fleet deployment and scheduling problem is: first, use the designed encoding rules to generate an initial solution, then design the rules for generating new solutions, and then combine the receiving function, annealing strategy and ending strategy of the simulated annealing algorithm to finally obtain the final solution; the specific implementation method is as follows: In the simulated annealing algorithm, the encoding and decoding process is the process of determining the order of ports to be visited and the number and type of ships serving the route. Assume that 12 ships are deployed on route 1 and type 2 ships are selected to serve the route. If the port numbers on the route are 2, 5, 11, 9, and 8, and the random speed values ​​of each section are initialized within a reasonable range, then the encoding is [1, 2, 12, 5, 9, 2, 8, 11, 22, 25, 20, 19]. Based on the above encoding and decoding rules, the initial solution can be generated, and then a new solution is generated. When designing the new solution generation rules, it is divided into two parts: the first part is the order of service ports. The strategy adopted in this part is to randomly select the order of several ports on the route and swap them; the second part is the generation of speed on each segment. The speed of some segments is randomly generated in a reasonable range based on the distance. Finally, the acceptance function, annealing strategy, and ending strategy of the simulated annealing algorithm are presented. The acceptance function in the simulated annealing algorithm is used to determine whether to accept a new solution as the current solution. It allows accepting inferior solutions with a certain probability during the search process to avoid falling into the local optimum. The simulated annealing algorithm selects the Metropolis criterion as the acceptance function: Among them, E(n+1) represents the newly calculated allocation plan of the type and quantity of ships on each route of the shipping company, and E(i) represents the last calculated allocation plan of the type and quantity of ships on each route of the shipping company. When E(i)≤E(n+1), it means that the allocation of the type and quantity of ships in the current plan is more reasonable and the transportation efficiency of goods is faster, so the current solution is adopted. On the contrary, if E(i)>E(n+1), it means that the allocation of the type and quantity of ships in the current plan or the transportation efficiency of goods is not as good as the previous solution, then θ=∪(0,1). If Then adopt the current solution, that is, adopt the worse solution; Exponential annealing refers to the temperature decreasing exponentially, for example, T j =T0·α k , where T j is the temperature after the jth iteration, T0 is the set initial temperature, α is the annealing factor and α<1; the simulated annealing algorithm uses this method to construct the annealing function, that is, T j+1 =T j s, where s∈[0.95, 0.99]; Finally, the ending strategy of the simulated annealing algorithm is to set the maximum number of iterations, that is, it ends when the termination temperature is reached.

10. The method for dispatching and controlling container liner transportation considering environmental pollution and transportation efficiency according to claim 1, characterized in that: The control model 1 VDS1 and the control model 2 VDS2 are solved by the branch and bound algorithm: The branch-and-bound algorithm can effectively solve the mixed integer linear programming (MILP) problem. In order to quickly obtain the global optimal solution, two auxiliary methods are embedded on the basis of the algorithm: the first is to use the Gurobi solver as a preprocessor, and the second is to combine it with a heuristic algorithm. In the first method, the Gurobi solver is used as a preprocessing method to solve the relaxed model. This process is performed before the branch and bound algorithm is used in order to greatly reduce the time required for the branch and bound algorithm to search. The relaxation of the model is mainly focused on the selection of ship type, which is separated from the binary decision variable f rk Relax to real decision variables, so that non-integer values ​​may appear in the solution process, which is also one of the parts where the branch and bound algorithm will go to branch, where f rk It means that if the k-type ship service route r is used, it is 1, otherwise it is 0; in the ship type part, only these non-integer ship types need to be branched; in the second algorithm combining the branch and bound algorithm with the heuristic algorithm, in order to speed up the search efficiency, the heuristic method is used to obtain the relatively optimal solution as the initial solution of the branch and bound algorithm, and then the solution is obtained by the branch and bound algorithm; In the branch and bound algorithm, the pruning process is necessary, so special pruning rules are formulated for the algorithm. In the branch and bound algorithm, the search strategy used is the depth-first search strategy. This strategy is used because the depth-first search strategy can definitely find the global optimal solution, and the depth-first search is more efficient than the breadth-first search in terms of memory usage, because it only needs to store the nodes on a branch path, rather than the nodes of the entire layer; then the part of the ship selection in the code will be judged by an integer. If it is a non-integer, the upper bound value will be updated, otherwise the lower bound will be updated; then the constraints in the model will be judged, such as the constraints on speed and service frequency. If the above conditions are not met, the algorithm will take pruning measures; after calculating the target value, the corresponding coding part will be judged by an integer. If they are all integers and the target value is less than the lower bound, it means that no better solution will appear if the branch is further branched, so it will be pruned to stop the branch.

Images