Multi-target ship scheduling method based on GA-SA algorithm

Abstract

The invention discloses a multi-target ship scheduling method based on a GA-SA algorithm. The method comprises the following steps: obtaining ship information; comprehensively evaluating the throughput rate of the ship lock, the area utilization rate of a lock chamber of the ship lock, the average lock waiting time and the lockage fairness by using a fitness function; the selection state of the ship to be locked is coded into a binary character string, an initial population is generated, and individuals with high fitness are selected through selection operation to enter the next generation; performing crossover and mutation operation on the selected individuals to obtain a new population; in the main cycle of the genetic algorithm, part of individuals with high fitness are selected for each generation to perform simulated annealing optimization, disturbance operation is performed on the individuals with high fitness to obtain a new solution, whether the new solution is accepted or not is selected according to the probability, and after the temperature is reduced to the specified temperature, an optimal solution obtained by simulated annealing is reserved; and judging whether the number of iterations is reached, and if so, outputting the scheduling scheme with the highest fitness. The problem that traditional ship lockage scheduling depends on manpower and is low in efficiency is solved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and ship scheduling, and in particular to a multi-objective ship scheduling method based on a GA-SA algorithm. Background Art

[0002] In modern water transport systems, ship locks, as important channels connecting different waters, play a vital role in the efficiency and safety of water transport. With the continuous growth of global trade volume and water transport demand, the traffic pressure of ship locks has increased year by year. In this case, it is particularly important to improve the throughput capacity and resource utilization of ship locks. However, ships may have a lot of waiting time while waiting to pass through the locks, which affects the traffic efficiency and increases the operating costs. In addition, the scheduling of ship locks must also take into account the diversity of ship types and the priority requirements of special situations. The traditional FIFO algorithm is simple and easy to implement, but it is difficult to balance and control multiple objectives, and performs poorly in large-scale and complex scenarios. The existing intelligent optimization algorithm is difficult to balance multiple objectives and global optimization. Therefore, it is necessary to design an algorithm to effectively implement multi-objective complex scenarios, such as the GA-SA algorithm (CN116451956A) applied to AGV collaborative scheduling. After initializing the parameters of the algorithm, the optimal solution is screened out by the genetic algorithm and then further optimized as the initial solution of the simulated annealing algorithm, and finally the scheduling scheme is obtained. Although this method can smoothly transition from genetic algorithm to simulated annealing algorithm, treating the two as independent optimization processes may lead to local optimality, resulting in low quality of the final solution. In addition, the constraints on generating new solutions and the dynamic adjustment of parameters are not considered during the algorithm iteration process, resulting in low flexibility. Summary of the invention

[0003] In order to overcome the problems of poor performance of traditional methods in complex scenarios and difficulty in balancing multi-objective balance and global optimization, the present invention provides a multi-objective ship scheduling method based on GA-SA algorithm, which realizes ship scheduling with balanced multi-objective in complex scenarios by combining genetic algorithm (GA) with simulated annealing algorithm (SA) and utilizing the global optimization ability of genetic algorithm and the local search ability of simulated annealing, thereby improving the efficiency and quality of lock passage while ensuring the fairness of ship passage.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] A multi-objective ship scheduling method based on GA-SA algorithm comprises the following steps:

[0006] S1. Obtain ship information, obtain ship type and number information through camera based on deep learning method, obtain ship size through radar 3D point cloud technology, and measure ship draft and weight through water level gauge;

[0007] S2. Use the fitness function to comprehensively evaluate the lock throughput rate, lock chamber area utilization, average waiting time and lock fairness;

[0008] S3, encode the selection status of the waiting ship into a binary string, randomly generate an initial population, and select individuals with high fitness to enter the next generation through selection operation;

[0009] S4, perform crossover and mutation operations on the selected individuals, and obtain a new population while ensuring that the constraints are met;

[0010] S5. In the main loop of the genetic algorithm, some individuals with high fitness are selected in each generation for simulated annealing optimization;

[0011] S6. Perform perturbation operations on individuals with high fitness to obtain new solutions, and choose whether to accept the new solutions according to probability until the temperature drops to the specified temperature, and retain the optimal solution obtained by simulated annealing;

[0012] S7. Determine whether the number of iterations has been reached. If so, output the scheduling plan with the highest fitness. Otherwise, return to step S4 to continue the next generation of operations.

[0013] Furthermore, the fitness function of step S2 is:

[0014]

[0015] Where f represents the fitness function of the comprehensive ship lock throughput rate, ship lock chamber area utilization rate, average waiting time and lock fairness. It represents the ratio of the x-th scheduling throughput to the maximum throughput, Q x Indicates the total weight of the ship that passed the xth dispatch, Indicates the maximum weight that the lock can bear during the xth dispatch; ∑A x represents the utilization rate of the lock chamber area during the x-th dispatch, represents the sum of the priorities of the ships scheduled for the xth time, It represents the ratio of the average waiting time to the maximum waiting time of the xth dispatch. represents the average waiting time of the ship selected for the xth dispatch, represents the maximum waiting time of the ship selected for the xth dispatch; μ1, μ2, μ3, and μ4 represent the weight of each item respectively;

[0016] The constraints include:

[0017] The ratio of the sum of the weights of the ships selected for this dispatch to the total weight that the lock chamber can withstand:

[0018]

[0019] Where k represents the kth ship dispatched for the xth time, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0020] Ratio of the sum of the selected ship areas to the total area of ​​the lock chamber:

[0021]

[0022] Where S i represents the overhead area of ​​the i-th ship, L and W represent the length and width of the lock chamber respectively;

[0023]

[0024] Where P c It means that when the ship is a special ship, the priority is 1. For non-special ships and within the maximum waiting time range, the priority increases with the waiting time. If the waiting time exceeds, it becomes a constant, where λ1 and λ2 are weight coefficients;

[0025] The ratio of the total waiting time of a ship to the maximum waiting time of a ship:

[0026]

[0027] where t i ,t imax represents the waiting time and maximum waiting time of the i-th ship;

[0028] The total weight of the ships carried does not exceed the weight that the lock can bear:

[0029]

[0030] Among them, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0031] The sum of the length and width of any row or column of ships plus the safety interval shall not exceed the length and width of the lock chamber:

[0032]

[0033] Where d represents the safe distance between the ship and the lock chamber, L i , W i represents the length and width of the i-th ship, L and W represent the length and width of the lock chamber;

[0034] The x+1th dispatch is the remaining ships of the xth dispatch and the new ships that arrive after the start of the xth dispatch and before the start of the x+1th dispatch:

[0035] N x+1 =N x -R x +C x ;

[0036] Where N x+1 N represents the set of ships waiting for the x+1th scheduling. x represents the set of ships waiting for the xth scheduling, R x represents the set of ships selected to pass through the lock in the xth scheduling, C x Represents the set of newly arrived ships during the xth scheduling period.

[0037] Furthermore, in step S3, the chromosome with the highest fitness is selected from the initial population and each subsequent generation of population, and the roulette wheel method is used to randomly select chromosomes with higher fitness from the remaining chromosomes to combine and generate the next generation of population.

[0038] Further, a series of parameters are initialized, including the crossover probability P c , minimum mutation probability P mlow , the highest mutation probability P mhigh Initial temperature T0, final temperature T end , chain length length, population size population_size and randomly generate the initial population.

[0039] Furthermore, in step S4, a partial matching crossover operation is performed on the two selected chromosomes, two crossover points are randomly selected in the chromosomes, and the gene segments between the two crossover points in the two chromosomes are exchanged, while ensuring that the generated new chromosome meets the constraint conditions.

[0040] Furthermore, in step S4, a local correction strategy based on constraint repair is introduced into the new population obtained by the crossover operation to remove chromosomes that do not meet the constraint conditions.

[0041] Furthermore, in step S4, an inversion mutation operation is performed on one of the selected chromosomes, a gene segment in the chromosome is selected, and the sequence of the gene is sequentially reversed, while ensuring that the generated new chromosome meets the constraint conditions.

[0042] Furthermore, a dynamic mutation probability strategy based on iterative algebra is introduced in the process of generating new populations. The mutation probability is dynamically adjusted according to the number of iterations to improve the global search capability of the algorithm.

[0043] Furthermore, in step S6, using the local search capability of the simulated annealing algorithm, in each new population generated, only the first five chromosomes with the highest fitness are subjected to simulated annealing optimization, and the steps are as follows:

[0044] S61, select five chromosomes from chromosome R1 to chromosome R5, exchange the positions of two genes in the chromosomes, and legalize the chromosomes to generate new chromosomes R1' to new chromosome R5';

[0045] S62, calculate the fitness difference ΔE=f(R′)-f(R) between the new chromosome R1′-new chromosome R5′ and the chromosome R1-chromosome R5, where f(R) and F(R′) represent the fitness function values ​​corresponding to chromosomes R and R′;

[0046] S63. If the fitness difference ΔE>0, it means that the fitness value of the new chromosome is greater, and the new chromosome is accepted; otherwise, according to the Meltrov criterion, the probability Accept new chromosomes, where T represents the current temperature value;

[0047] S64, lower the temperature, so that T j =αT j-1 ,α∈(0,1), until the termination temperature T is reached end , where T j represents the temperature value at the jth iteration, and the constant α represents the cooling rate;

[0048] S65. When the termination temperature is reached, the individual with the highest current fitness is output as the optimal solution.

[0049] Furthermore, in step S7, after completing the simulated annealing optimization, it is determined whether the number of iterations has been reached. If not, the next generation of population is generated. If reached, the optimal solution is output, the algorithm is exited, and the gene sequence of the optimal solution is parsed as the solution for this scheduling.

[0050] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0051] The present invention utilizes the GA-SA artificial intelligence algorithm to schedule the passage of ships through locks, which can achieve the scheduling of ships passing through locks that takes into account multiple objectives and is globally optimized, and can meet the requirements of river locks for the high efficiency and fairness of ships passing through locks in water transportation. This method is a typical artificial intelligence method, and the entire scheduling process does not require human arrangement, thereby improving the scheduling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of a multi-objective ship scheduling method based on the GA-SA algorithm according to an embodiment of the present invention;

[0053] Figure 2A schematic diagram of an on-site ship lock of a multi-objective ship scheduling method based on a GA-SA algorithm according to an embodiment of the present invention;

[0054] Figure 3 It is a GA-SA convergence curve diagram during the iterative process of fitness value change in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] This section will describe in detail the specific embodiments of the present invention. The preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the accompanying drawings is to supplement the description of the text part of the specification with graphics, so that people can understand the overall technical solution of the present invention intuitively and vividly, but it should not be understood as a limitation on the scope of protection of the present invention.

[0056] like Figure 1 As shown, a multi-objective ship scheduling method based on the GA-SA algorithm of this embodiment includes the following steps:

[0057] S1. Obtain ship information, obtain ship type and number information through camera based on deep learning method, obtain ship size through radar 3D point cloud technology, and measure ship draft and weight through water level gauge;

[0058] S2. Use the fitness function to comprehensively evaluate the lock throughput rate, lock chamber area utilization, average waiting time and lock fairness;

[0059] Ship scheduling is a multi-objective optimization problem, which requires maximizing throughput rate, maximizing lock chamber utilization, minimizing the average waiting time of ships, and maximizing the priority of ships passing through the lock.

[0060] Assumptions and provisions for the ship dispatch process:

[0061] a) The time it takes for the lock chamber to complete the passage of a certain number of ships is the same and known;

[0062] b) The weight and area of ​​the ship that the lock chamber can bear have a certain upper limit and are known;

[0063] c) Ensure that ships arriving at any time are within the lock operation time;

[0064] d) Ships must maintain a certain safety distance between each other and the lock chamber can accommodate a maximum of two ships sailing side by side;

[0065] e) After the selected ship has been determined in the current dispatch, the newly arrived ship will participate in the next dispatch;

[0066] The fitness function is:

[0067]

[0068] Among them, f represents the fitness function of the comprehensive ship lock throughput rate, ship lock chamber area utilization rate, average waiting time and lock fairness, It represents the ratio of the x-th scheduling throughput to the maximum throughput, Q x Indicates the total weight of the ship that passed the xth dispatch, Indicates the maximum weight that the lock can bear during the xth dispatch; ∑A x represents the utilization rate of the lock chamber area during the x-th dispatch, represents the sum of the priorities of the ships scheduled for the xth time, It represents the ratio of the average waiting time to the maximum waiting time of the xth dispatch. represents the average waiting time of the ship selected for the xth dispatch, represents the maximum waiting time of the ship selected for the xth dispatch; μ1, μ2, μ3, and μ4 represent the weight of each item respectively;

[0069] Constraints:

[0070] The ratio of the sum of the weights of the ships selected for this dispatch to the total weight that the lock chamber can withstand:

[0071]

[0072] Where k represents the kth ship dispatched for the xth time, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0073] Ratio of the sum of the selected ship areas to the total area of ​​the lock chamber:

[0074]

[0075] Where S i represents the overhead area of ​​the i-th ship, L and W represent the length and width of the lock chamber respectively;

[0076]

[0077] It means that when the ship is a special ship, the priority is 1. For non-special ships and within the maximum waiting time range, the priority increases with the waiting time. If the waiting time exceeds, it becomes a constant, where λ1 and λ2 are weight coefficients;

[0078] The ratio of the total waiting time of a ship to the maximum waiting time of a ship:

[0079]

[0080] where t i ,t imax represents the waiting time and maximum waiting time of the i-th ship;

[0081] The total weight of the ships carried does not exceed the weight that the lock can bear:

[0082]

[0083] Among them, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0084] The sum of the length and width of any row or column of ships plus the safety interval shall not exceed the length and width of the lock chamber:

[0085]

[0086] Where d represents the safe distance between the ship and the lock chamber, L i , W i represents the length and width of the i-th ship, L and W represent the length and width of the lock chamber;

[0087] The x+1th dispatch is the remaining ships of the xth dispatch and the new ships that arrive after the start of the xth dispatch and before the start of the x+1th dispatch:

[0088] N x+1 =N x -R x +C x ;

[0089] Where N x+1 N represents the set of ships waiting for the x+1th scheduling. x represents the set of ships waiting for the xth scheduling, R x represents the set of ships selected to pass through the lock in the xth scheduling, C x represents the set of newly arrived ships during the xth scheduling period;

[0090] S3, encode the selection status of the waiting ship into a binary string, 1 indicates that the ship is selected, and 0 indicates that the ship is not selected, randomly generate an initial population, and select individuals with high fitness to enter the next generation through selection operation;

[0091] When designing the GA-SA algorithm, we first need to initialize a series of parameters including the crossover probability P c , minimum mutation probability P mlow , the highest mutation probability P mhigh , initial temperature T0, final temperature T end , chain length length, population size population_size and randomly generate the initial population.

[0092] The chromosome with the highest fitness is selected from the initial population and each subsequent generation of the population, and the roulette wheel method is used to randomly select chromosomes with higher fitness from the remaining chromosomes to combine and generate the next generation of the population.

[0093] S4, perform crossover and mutation operations on the selected individuals, and obtain a new population while ensuring that the constraints are met;

[0094] A partial matching crossover operation is used for the two selected chromosomes. Two crossover points are selected in the chromosomes, and the gene segment between the two crossover points in the two chromosomes is exchanged, while ensuring that the generated new chromosome meets the constraint conditions.

[0095] A reversal mutation operation is performed on one of the selected chromosomes. A gene in the chromosome is selected and the sequence of the gene is reversed sequentially, while ensuring that the generated new chromosome meets the constraints.

[0096] S5. In the main loop of the genetic algorithm, some individuals with high fitness are selected in each generation for simulated annealing optimization;

[0097] S6, perform perturbation operations on individuals with high fitness to obtain new solutions, and choose whether to accept the new solutions according to a certain probability until the temperature drops to the specified temperature, and retain the optimal solution obtained by simulated annealing;

[0098] By utilizing the local search capability of the simulated annealing algorithm without increasing the computational burden, in each new population generated, only the first five chromosomes with the highest fitness are optimized by simulated annealing, which are determined by the following steps:

[0099] S61, select five chromosomes R1-R5, exchange the positions of two genes in the chromosomes, and legalize the chromosomes to generate new chromosomes R1'-R5';

[0100] S62, calculating the fitness difference ΔE=f(R′)-f(R) of the new chromosome R1′-R5′, where f(R) represents the fitness function value corresponding to chromosome R;

[0101] S63. If the fitness difference ΔE>0, it means that the fitness value of the new chromosome is greater, and the new chromosome is accepted; otherwise, according to the Metropolis criterion, the probability Accept new chromosome, where T represents the current temperature value;

[0102] S64, lower the temperature, so that T j =αT j-1 ,α∈(0,1), until the termination temperature T is reached end , where T jrepresents the temperature value at the jth iteration, and the constant α represents the cooling rate; S65, when the termination temperature is reached, output the individual with the highest current fitness as the optimal solution.

[0103] S7: Determine whether the number of iterations has been reached. If it has been reached, output the scheduling plan with the highest fitness. Otherwise, return to S4 to continue the next generation operation.

[0104] In one embodiment of the present invention, Figure 2 As shown, a multi-objective ship scheduling method based on GA-SA algorithm includes the following steps:

[0105] S1. Radar, camera and water level meter are set outside the downstream gate to measure and obtain the size, mass and type information of the ship. This embodiment does not involve specific measurement technology, but only obtains ship information and sends this information to the server, which dispatches the ship in real time.

[0106] S2. In the process of scheduling ships to pass through the lock, the benefits and fairness of passing through the lock should be taken into account. That is, not only the number and quality of ships passing through the lock should be maximized, but also on this basis, it should be ensured that the waiting time of ships is not too long or special ships are not given priority to pass through the lock. Therefore, the fitness function is defined by comprehensively evaluating the four indicators of lock throughput rate, lock chamber area utilization rate, average waiting time, and lock fairness;

[0107] S3. Encode the selection status of the waiting ship into a binary string, where 1 indicates that the ship is selected and 0 indicates that the ship is not selected. For example, (1,0,1,0,0,0,1,0,0,0,0,0,0,1,0,0,0,0,1,0) means that ships 1, 3, 7, 14 and 19 are selected to pass through the lock. Generate an initial population randomly, and select individuals with high fitness to enter the next generation through selection operation;

[0108] S4, perform crossover and mutation operations on the selected individuals, and obtain a new population while ensuring that the constraints are met;

[0109] S5. In the main loop of the genetic algorithm, some individuals with high fitness are selected in each generation for simulated annealing optimization;

[0110] S6, perform perturbation operations on individuals with high fitness to obtain new solutions, and choose whether to accept the new solutions according to a certain probability until the temperature drops to the specified temperature, and retain the optimal solution obtained by simulated annealing;

[0111] S7, determine whether the number of iterations has been reached, if so, output the scheduling plan with the highest fitness, otherwise return to S4 to continue the next generation operation;

[0112] like Figure 2As shown, in this scenario, a certain number of cameras, water level gauges and radars are required to obtain the type, quality and size information of the ship, and then send the information to the server in the computer room.

[0113] As a preferred embodiment, this embodiment adopts a fitness function that comprehensively evaluates the lock throughput rate, lock chamber area utilization rate, average waiting time, and lock fairness.

[0114] Some assumptions and provisions are made for the model building process:

[0115] a) The time it takes for a lock chamber to complete the passage of a certain number of ships is the same and known, assuming it is 1 hour;

[0116] b) The weight and area of ​​the lock chamber that can bear ships have a certain upper limit and are known. Assume that the maximum bearing weight is 10,000 tons and the lock chamber area is 300m×30m;

[0117] c) Ensure that ships arriving at any time are within the lock operation time;

[0118] d) Ships must maintain a certain safety distance of 3m between each other, and the lock chamber can accommodate a maximum of two ships sailing side by side;

[0119] e) After the selected ship has been determined in the current dispatch, the newly arrived ship will participate in the next dispatch;

[0120] Fitness function definition:

[0121]

[0122] Where f represents the fitness function of the comprehensive ship lock throughput rate, ship lock chamber area utilization rate, average waiting time and lock fairness. It represents the ratio of the x-th scheduling throughput to the maximum throughput, Q x Indicates the total weight of the ship that passed the xth dispatch, Indicates the maximum weight that the lock can bear during the xth dispatch; ∑A x represents the utilization rate of the lock chamber area during the x-th dispatch, represents the sum of the priorities of the ships scheduled for the xth time, It represents the ratio of the average waiting time to the maximum waiting time of the xth dispatch. represents the average waiting time of the ship selected for the xth dispatch, represents the maximum waiting time of the ship selected for the xth scheduling; μ1, μ2, μ3, and μ4 represent the weight of each item respectively; in this embodiment, they are set to 0.25, 0.25, 0.25, and 0.25.

[0123] Constraints:

[0124] The ratio of the sum of the weights of the ships selected for this dispatch to the total weight that the lock chamber can withstand:

[0125]

[0126] Where k represents the kth ship dispatched for the xth time, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0127] Ratio of the sum of the selected ship areas to the total area of ​​the lock chamber:

[0128]

[0129] Where S i represents the overhead area of ​​the i-th ship, L and W represent the length and width of the lock chamber respectively;

[0130]

[0131] It means that when the ship is a special ship, the priority is 1. For non-special ships and within the maximum waiting time range, the priority increases with the waiting time. If the waiting time exceeds, it becomes a constant, where λ1 and λ2 are weight coefficients;

[0132] The ratio of the total waiting time of a ship to the maximum waiting time of a ship:

[0133]

[0134] where t i ,t imax represents the waiting time and maximum waiting time of the i-th ship;

[0135] The total weight of the ships carried does not exceed the weight that the lock can bear:

[0136]

[0137] Among them, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear;

[0138] The sum of the length and width of any row or column of ships plus the safety interval shall not exceed the length and width of the lock chamber:

[0139]

[0140] Where d represents the safe distance between the ship and the lock chamber, L i , W i represents the length and width of the i-th ship, L and W represent the length and width of the lock chamber;

[0141] The x+1th dispatch is the remaining ships of the xth dispatch and the new ships that arrive after the start of the xth dispatch and before the start of the x+1th dispatch:

[0142] N x+1 =N x -R x +C x ;

[0143] Where N x+1 N represents the set of ships waiting for the x+1th scheduling. x represents the set of ships waiting for the xth scheduling, R x represents the set of ships selected to pass through the lock in the xth scheduling, C x represents the set of newly arrived ships during the xth scheduling period;

[0144] When designing the GA-SA algorithm, the parameters are initialized first, as shown in Table 1.

[0145] Table 1 GA-SA algorithm parameter initialization

[0146] parameter value Population size (ns) 40 <![CDATA[Crossing probability (P c )]]> 0.8 <![CDATA[Minimum mutation probability (P mlow )]]> 0.2 <![CDATA[Highest mutation probability (P mhigh )]]> 0.8 Genetic iteration number (t) 100 <![CDATA[Initial temperature (T0)]]> 500 <![CDATA[Termination temperature (T end )]]> 0.1 Temperature drop coefficient (α) 0.95 Chain length 50

[0147] In this embodiment, a queue of 20 ships waiting for the lock is selected, and various information of the ships is obtained through the method in step S1 as the genetic encoding method of the scheduling algorithm and the calculation index of the fitness function, as shown in Table 2.

[0148] Table 2 Ship information data

[0149]

[0150]

[0151] After setting the algorithm parameters and obtaining various data of the waiting ship, a population of 40 individuals is randomly generated according to the binary vector form in step S3, and each individual represents a scheduling method. Excellent individuals are selected to enter the next generation through the selection operation. As a preferred embodiment, the selection operation adopts a combination of elite strategy and roulette, and the top 10%, i.e., the top 4 individuals in fitness value, are selected to enter the next generation directly, and 12 individuals are randomly selected from the remaining individuals to enter the next generation, and a total of 40% of the individuals of this generation are selected to enter the next generation.

[0152] In order to ensure that the total number of individuals in the next generation reaches 40, the remaining individuals in the parent generation except those selected by the elite strategy to enter the next generation directly are subjected to crossover and mutation operations. The crossover operation adopts partial matching crossover, that is, two gene points are randomly determined in the individuals selected for the crossover operation, and the gene sequences of the two individuals between the two gene points are exchanged to obtain two new individuals. After introducing the local correction strategy (AConstraint-Handling Technique for Genetic Algorithms using a Violation Factor), the generated new solution is checked. For chromosomes that do not meet the constraints, the ships with a selected state of 1 are randomly eliminated and recalculated until the constraints are met.

[0153] The mutation operation adopts inversion mutation, that is, two gene points are selected for the individual selected for mutation operation, and the gene sequence between the two points is inverted to obtain a new individual. After introducing the dynamic mutation probability strategy (ARank based Adaptive Mutation in Genetic Algorithm, Avijit Basak, IBM India Pvt. Ltd. Kolkata, India), the mutation operation of this embodiment is performed with the following probability:

[0154]

[0155] Where P m represents the probability of contemporary population mutation, t n represents the current iteration number, P mlow is the minimum mutation probability, P mhigh is the highest mutation probability.

[0156] As a preferred embodiment, in step S6, in the new population generated in each generation, the top five individuals with the highest fitness values ​​are selected for simulated annealing optimization. The specific steps are as follows:

[0157] S61, for the selected five chromosomes R1-R5, the positions of two genes in the chromosomes are exchanged, and the chromosomes are legalized to generate new chromosomes R1'-R5';

[0158] S62, calculate the fitness difference ΔE=f(R′)-f(R);

[0159] S63. If ΔE>0, it means that the new chromosome has a larger fitness value, so the new chromosome is accepted; otherwise, according to the Metropolis criterion, the probability Acceptance of new chromosomes;

[0160] S64, lower the temperature, so that T j =αT j-1 ,(α∈(0,1)), until the termination temperature T end ;

[0161] S65. When the termination temperature is reached, the individual with the highest current fitness is output as the optimal solution.

[0162] The overall process of the algorithm is as follows Figure 1 As shown, the algorithm is iterated continuously until the maximum number of iterations is reached, and the individual with the highest fitness value is output as the final scheduling solution. The fitness value changes during the iteration process are shown in Figure 3 GA-SA convergence curve diagram shown.

[0163] The present invention utilizes the GA-SA artificial intelligence algorithm to schedule the passage of ships through locks, which can achieve the scheduling of ships passing through locks that takes into account multiple objectives and is globally optimized, and can meet the requirements of river locks for the high efficiency and fairness of ships passing through locks in water transportation. This method is a typical artificial intelligence method, and the entire scheduling process does not require human arrangement, thereby improving the scheduling efficiency.

[0164] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well.

Claims

1. A multi-objective ship scheduling method based on GA-SA algorithm, characterized in that: The following steps are involved: S1. Obtain ship information, obtain ship type and number information through camera based on deep learning method, obtain ship size through radar 3D point cloud technology, and measure ship draft and weight through water level gauge; S2. Use the fitness function to comprehensively evaluate the lock throughput rate, lock chamber area utilization, average waiting time and lock fairness; S3, encode the selection status of the waiting ship into a binary string, randomly generate an initial population, and select individuals with high fitness to enter the next generation through selection operation; S4, perform crossover and mutation operations on the selected individuals, and obtain a new population while ensuring that the constraints are met; S5. In the main loop of the genetic algorithm, some individuals with high fitness are selected in each generation for simulated annealing optimization; S6. Perform perturbation operations on individuals with high fitness to obtain new solutions, and choose whether to accept the new solutions according to probability until the temperature drops to the specified temperature, and retain the optimal solution obtained by simulated annealing; S7. Determine whether the number of iterations has been reached. If so, output the scheduling plan with the highest fitness. Otherwise, return to step S4 to continue the next generation of operations.

2. A multi-objective ship scheduling method based on GA-SA algorithm according to claim 1, characterized in that: The fitness function of step S2 is: Where f represents the fitness function of the comprehensive ship lock throughput rate, ship lock chamber area utilization rate, average waiting time and lock fairness. It represents the ratio of the x-th scheduling throughput to the maximum throughput, Q x Indicates the total weight of the ship that passed the xth dispatch, Indicates the maximum weight that the lock can bear during the xth dispatch; ∑A x represents the utilization rate of the lock chamber area during the x-th dispatch, represents the sum of the priorities of the ships scheduled for the xth time, It represents the ratio of the average waiting time to the maximum waiting time of the xth dispatch. represents the average waiting time of the ship selected for the xth dispatch, represents the maximum waiting time of the ship selected for the xth dispatch; μ1, μ2, μ3, and μ4 represent the weight of each item respectively; The constraints include: The ratio of the sum of the weights of the ships selected for this dispatch to the total weight that the lock chamber can withstand: Where k represents the kth ship dispatched for the xth time, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear; Ratio of the sum of the selected ship areas to the total area of ​​the lock chamber: Where S i represents the overhead area of ​​the i-th ship, L and W represent the length and width of the lock chamber respectively; Where P c It means that when the ship is a special ship, the priority is 1. For non-special ships and within the maximum waiting time range, the priority increases with the waiting time. If the waiting time exceeds, it becomes a constant, where λ1 and λ2 are weight coefficients; The ratio of the total waiting time of a ship to the maximum waiting time of a ship: where t i ,t imax It represents the waiting time and maximum waiting time of the i-th ship; The total weight of the ships carried does not exceed the weight that the lock can bear: Among them, M i represents the mass of the i-th ship, M max It indicates the maximum mass that the lock chamber can bear; The sum of the length and width of any row or column of ships plus the safety interval shall not exceed the length and width of the lock chamber: Where d represents the safe distance between the ship and the lock chamber, L i , W i represents the length and width of the i-th ship, L and W represent the length and width of the lock chamber; The x+1th dispatch is the remaining ships of the xth dispatch and the new ships that arrive after the start of the xth dispatch and before the start of the x+1th dispatch: N x+1 =N x -R x +C x ; Where N x+1 N represents the set of ships waiting for the x+1th scheduling. x represents the set of ships waiting for the xth dispatch, R x represents the set of ships selected to pass through the lock in the xth scheduling, C x Represents the set of newly arrived ships during the xth scheduling period.

3. The multi-objective ship scheduling method based on the GA-SA algorithm according to claim 1 is characterized in that: In step S3, the chromosome with the highest fitness is selected from the initial population and each subsequent generation of population, and the roulette wheel method is used to randomly select chromosomes with higher fitness from the remaining chromosomes and combine them to generate the next generation population.

4. The multi-objective ship scheduling method based on the GA-SA algorithm according to claim 3 is characterized in that: Initialize a series of parameters, including the crossover probability P c , minimum mutation probability P mlow , the highest mutation probability P mhigh Initial temperature T0, final temperature T end , chain length length, population size population_size and randomly generate the initial population.

5. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In step S4, a partial matching crossover operation is performed on the two selected chromosomes, two crossover points are randomly selected in the chromosomes, and the gene segments between the two crossover points in the two chromosomes are exchanged, while ensuring that the generated new chromosome meets the constraint conditions.

6. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In step S4, a local correction strategy based on constraint repair is introduced into the new population obtained by the crossover operation to remove chromosomes that do not meet the constraint conditions.

7. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In step S4, an inversion mutation operation is performed on one of the selected chromosomes, a gene segment in the chromosome is selected, and the sequence of the gene is sequentially reversed, while ensuring that the generated new chromosome meets the constraint conditions.

8. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In the process of generating new populations, a dynamic mutation probability strategy based on iterative algebra is introduced, and the mutation probability is dynamically adjusted according to the number of iterations to improve the global search capability of the algorithm.

9. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In step S6, using the local search capability of the simulated annealing algorithm, in each new population generated, only the first five chromosomes with the highest fitness are optimized by simulated annealing. The steps are as follows: S61, select five chromosomes from chromosome R1 to chromosome R5, exchange the positions of two genes in the chromosomes, and legalize the chromosomes to generate new chromosomes R1' to new chromosome R5'; S62, calculate the fitness difference ΔE=f(R′)-f(R) between the new chromosome R1′-new chromosome R5′ and the chromosome R1-chromosome R5, where f(R) and f(R′) represent the fitness function values ​​corresponding to chromosomes R and R′; S63. If the fitness difference ΔE>0, it means that the fitness value of the new chromosome is greater, and the new chromosome is accepted; otherwise, according to the Meltrov criterion, the probability Accept new chromosomes, where T represents the current temperature value; S64, lower the temperature, so that T j =αT j-1 ,α∈(0,1), until the termination temperature T is reached end , where T j represents the temperature value at the jth iteration, and the constant α represents the cooling rate; S65. When the termination temperature is reached, the individual with the highest current fitness is output as the optimal solution.

10. The multi-objective ship scheduling method based on GA-SA algorithm according to claim 1 is characterized in that: In step S7, after completing the simulated annealing optimization, it is determined whether the number of iterations has been reached. If not, the next generation of population is generated. If reached, the optimal solution is output, the algorithm is exited, and the gene sequence of the optimal solution is parsed as the solution for this scheduling.

Images