CN101782987A - Port tug dynamic scheduling method - Google Patents

Abstract

The invention relates to a port tug dynamic scheduling method. Along with the development of the shipping industry, arriving ships at a port increase progressively, and ship forms become larger progressively, so that the successful proceeding of port entry and exit tasks is difficult to be ensured according to a scheduling scheme of manual experience. How to determine a reasonable tug scheduling scheme specific for complicated and changeable arriving ship conditions becomes a problem to be solved in port service, and developing a research on the aspect is imperative and has great meanings. The method of the invention comprises the following steps of: analyzing a traditional tug dynamic scheduling problem, establishing a tug dynamic scheduling model, encoding the problem in a proper form on the basis of actual conditions of operation, carrying out fitness function design and scheduling scheme solution on the problem by utilizing a particle swarm optimization algorithm, introducing the concepts of an elite set and an entropy function and improving the algorithm; finally comparing and analyzing the scheme after the algorithmis improved with the scheme before the algorithm is improved through simulation; and mastering the application of a particle group algorithm to the port tug dynamic scheduling.

Description

Port tug dynamic scheduling method

Technical field

The present invention relates to the port dispatching field, particularly the tug dynamic scheduling field.

Background technology

The smooth degree of boats and ships entering and leaving port is an important symbol of port service level.Tugboat is an operation ship kind indispensable in the entering and leaving port process of boats and ships.

For guaranteeing that the boats and ships entering and leaving port can put, in time obtain the service that tugboat provides in place, the harbour must be equipped with the tugboat of sufficient amount and power, and should tugboat stop base be set according to the harbour concrete condition, makes tugboat spend the minimum time to arrive the operation berth.On the one hand, if tugboat quantity is very few, then can't be for providing necessity to serve timely to the port boats and ships, entering and leaving port boats and ships wait tugboat queuing time is long, and boats and ships are increased at harbour dues, with make the harbour with the competition at peripheral harbour in be on a sticky wicket; Otherwise tugboat too much will make the tugboat utilization factor low even tugboat to occur idle, then waste the harbour resource and increase managerial cost.How the reasonable disposition tugboat just can make the harbour when satisfying present needs, adapts to the need of harbour future development, and this seems very important concerning each harbour operator.How to dispatch tugboat by vessel leaving oceangoing ship situation with to the demand reasonable disposition of tugboat according to each harbour in the harbor district, make the port tug reasonable quantity, total arrangement is reasonable, reducing tugboat purchases and operation costs, save the preceding invalid hours underway of tugboat operation, be directly connected to boats and ships and lean on from pool safety and harbour economic benefit.

Along with the development of shipping business, increasing to the port boats and ships, the ship type maximizes day by day, has been difficult to guarantee carrying out smoothly of boats and ships entering and leaving port task according to the scheduling scheme of artificial experience.

Some harbour of China is to adopting dynamic entering and leaving port system to the departure from port boats and ships.Under this system, the time period of approaching that is concentrated in to the port of boats and ships carries out, and departure from port is concentrated in the Departure airport section and carries out.Alternately occurring of these two time periods, make corresponding tugboat operation also present dynamic characteristics; Every less important work of tugboat, for example auxiliary approaching, auxiliary leaving the port, escort and auxiliary shifting berth these, seem concentrated relatively in time.Research of the present invention only is to carry out at this dynamic operating type.

The harbour is owing to be subjected to the restriction of channel span, implementation be the navigation of single navigation channel, concentrate and approach, concentrate and leave the port.It is 7:00 (approaching) 9:00 (leaving the port) that the dynamic time took place in one day, approaches every two hours dynamically, leaves the port dynamically every two hours.

When the dynamic entry plan of N need be escorted tugboat, just send tugboat to go to the anchorage to escort N-1 dynamic (entering and leaving port is dynamic), make when N dynamically begins that all boats and ships in the dynamic entry plan of N can both arrive the navigation channel mouth.After this boats and ships approach with certain formation in the dynamic entry plan of N, and simultaneously tugboat company accesses tugboat from tugboat with assembling and opens target berth to each incoming vessel.When boats and ships arrived, tugboat maked a return voyage after assisting boats and ships to come to the bank, berth.In leaving the port dynamically, tugboat company accesses tugboat to calculated each berth, bar ship place of leaving the port, and after tugboat is moved to boats and ships in the navigation channel, if this ship needs to escort, just stays in the operation tugboat one and escorts boats and ships and leave the port, and all the other tugboats make a return voyage; If this ship does not need to escort, all tugboat makes a return voyage.More than be exactly dynamic tugboat work flow roughly, the scheduling situation of shifting berth similarly.

Boats and ships are circulated a notice of ship information to harbour affair office about first three day to the port, the harbour information that comprises ship basic information (captain, drinking water, ship's registry, institute's cargo or the like) and estimate to stop.Each approaches dynamically, leaves the port and shifting berth plan (may stop more than one berth by a ship, therefore need shifting berth) according to these information preparations in port office, and will dynamically plan to be handed down to tugboat company before dynamically beginning.Tugboat company goes operation according to dynamic plan tugboat, hits the target, and has also just finished the entering and leaving port task of boats and ships.

General every group of ships entered harbor 6-7 or more, queuing is approached successively.Tugboat generally correspondingly distributes the Mali and the quantity of tugboat by the tonnage size of ships entered harbor, equipping rules is as follows:

1, below 100 meter: 2600PS*1;

2,100 meters to 200 meters: 2600PS*2 or 3200PS*2 or 4000PS*2;

3,200 meters to 250 meters: 3200PS*2 or 4000PS*2;

4,250 meters to 300 meters: 3200PS*3 or 4000PS*2;

5, more than 300 meters: 4000PS*2 and 5000PS*1;

Sometimes under the much the same situation of tonnage, the general many arrangements of high-power tugboat are used for foreign steamers.

Most of incoming vessel is when the F.F. basin, is followed by the tugboat that waits there and enters basin, and at this moment tugboat need be with cable to be connected with ships entered harbor, assists its alongside harbour by tugboat.When first finish after, tugboat goes to carry out the alongside operation of next ship again, treats that incoming vessel all enters the berth by after good, tugboat all comes back to the base and awaits orders, and waits for to the boats and ships that leave the port and helps the pool operation.Tugboat is received dispatch command, and the position that arrives the port ship of waiting to leave the port is dragged to the navigation channel with boats and ships, and the boats and ships that leave the port utilize self power to leave the port, for needs convoy boats and ships, tugboat makes a return voyage then from escort the anchorage in the basin always, and tugboat escorts the convoy ship entry of need approaching again in the way of making a return voyage sometimes.The problem that exists in the tugboat allotment:

1, it is dynamic that manual dispatching is difficult to tackle the ship entry that quantity increases

Continuous aggravation along with trade contacts, container is imported and exported the increasing of quantity, the boats and ships quantity at contact harbour rapidly increases, widening of harbour makes the dynamic operation quantity of once receivable ship entry be on the increase, in the face of the increasing boats and ships of each dynamic entering and leaving port operation quantity, manual dispatching may can't be tackled the demand of a large amount of boats and ships dynamic dispatching of complexity in the past, may cause utilization factor not enough to sending the tugboat that carries out operation, it is improper to send, time delay on perhaps dispatching, though can finish the work smoothly, from economy, angle might not reach the optimum effect of expection efficiently.

2, owing to causing the unreasonable of allotment to the assurance of activity duration of tugboat is inaccurate

Outstanding big ship captain and pilot must fully understand the activity duration of the tugboat of each class hierarchy, and in time communicate with each other.Inconsiderate situation about arriving also can appear sometimes, tugboat such as No. 1 and No. 14 berth is all 2 grades of tugboats, this moment, 2 grades of tugboat of boats and ships wait in No. 15 berths assisted to approach operation, and 2 grades of tugboats in No. 1 berth have fulfiled assignment, therefore the pilot goes operation before just rule of thumb sending 2 grades of tugboats in No. 1 berth this moment, but he does not get hold of the deadline of 2 grades of tugboat operations in No. 14 berths, in the time of may just having sailed to No. 5 berths at 2 grades of tugboats in No. 1 berth, 2 grades of tugboats in No. 14 berths just can fulfil assignment, and so just will cause the unreasonable situation to the tugboat allotment to occur.Though operation smoothly may cause the waste of fuel cost.

3, the choice of utilization factor and stand-by period is held less than the best

Sometimes contradiction can appear in the pilot, the promptly existing boats and ships that need 2 grades of tugboat to assist and approach are waited for the tugboat operation, but this moment, 2 grades of all tugboats all carried out operation, and have 4 grades of tugboats to carry out operation in wait command, be to allow boats and ships wait for that obtained efficient more increases at this moment, still remove the efficiency of operation height before the tugboat of the big grade of group, manual dispatching can't provide definite answer, because have only, know that just accepting or rejecting which aspect is only more efficient and economical by after the computer Simulation calculation.

4, manual dispatching is unable to estimate optimum allotment scheme

Manual dispatching is the experience and knowledge of pilot according to oneself, it is dynamic to face each ship entry, providing the scheme of tugboat allotment comparatively smoothly, if can finish in the dynamic time of regulation, is a relatively reasonable allotment scheme just be regarded as this so; But, whether should more consider the efficient and the economy that adapt to market economy of today pursuing reasonably simultaneously.The ability of manual dispatching is limited after all, and the optimal case that can consider also is limited, and can not simulate up to a hundred thousands of scheduling scheme as computing machine reference is provided, and therefore the estimation for optimum allotment scheme is not enough.

Summary of the invention

The present invention is by research, can't finish increasingly sophisticated tug dynamic scheduling situation efficiently based on considering manual dispatching, therefore wish to pass through computer platform, utilize the utilization of algorithm in the scheduling of constraint condition, find more excellent scheduling scheme, make the boats and ships entering and leaving port can reach operation situation unimpeded, efficient, least cost; And select for use particle cluster algorithm to solve the port tug dynamic scheduling problem; Consider that this algorithm itself exists, be easy to be absorbed in locally optimal solution, at this defective, the present invention also proposes corresponding means, the thought that is entropy function and elite collection is incorporated in the algorithm, original algorithm is carried out to a certain degree improvement, provided at last at the optimal scheduling scheme of dynamic port tug dynamic scheduling of finding the solution arbitrarily.

It dynamically is a kind of regulation that the harbour is used for dividing operation.Under dynamic boats and ships entering and leaving port system: concentrate on to approach to the port boats and ships and dynamically approach; The departure from port boats and ships concentrate on dynamically departure from port of departure from port.Suppose to approach all is T with the dynamic duration of leaving the port dynamically _D, sea port berths distribute along the navigation channel, and its set is B=[B ₁, B ₂..., B _m].Tug=[t ₁, t ₂..., t _n] be the set of port tug, Level=[1,2 ..., L] be the set of tugboat grade, and there are the many-one correspondence relation in the element among the Tug and the element among the Level.

Approach dynamically begin before, boats and ships wait in the anchorage, wherein have some boats and ships need send tugboat to go to escort in advance, suppose that its set is HH=[Ship ₁, Ship ₂..., Ship _H] (the boats and ships sum of H for needing to escort); Approaching the dynamic zero hour, incoming vessel enters harbor district from the navigation channel mouth, and its set is IN=[Ship ₁, Ship ₂..., Ship _I] (I is an incoming vessel quantity); Leaving the port the dynamic zero hour, the ship that leaves the port rests in berth separately, and its set is OUT=[Ship ₁, Ship ₂..., Ship _O] (O is the ship quantity of leaving the port).So the dynamic set of tasks in entering and leaving port can be expressed as follows:

D＝HH?Y?OUT?Y?IN (3-1)

The target of scheduling guarantees the suitable tugboat of scheduling go to finish the work all tasks in the set exactly, and requires these tasks to be done as early as possible.

The foundation of tug dynamic scheduling model, model is set up based on following hypothesis:

1, ships entered harbor always can arrive the limit, berth smoothly, can not stop because of the obstruction of preceding ship;

2, before this dynamic escort mission began scheduling, Mission Operations was before all finished.

From the tugboat schedule job data statistics of reality, suppose that 2 realistic probability are higher than 90%.Based on this hypothesis, can infer that the optimal scheduling scheme of being obtained still should be optimum on whole time span.

The tugboat scheduling is that minimum work unit carries out with single ship always.Therefore dynamic task set DT can be expressed as follows in more detail:

DT＝{Ship _i|1≤i≤K，i∈N ^*} (3-2)

Objective function one: boats and ships are waited for tugboat activity duration sum minimum.

${\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="H2009101988898E00012.png,H2009101988898E00013.png"\; state="\{\{state\}\}">f</figure-callout>}}_1=\mathrm{Min}\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}{Y}_i---(3-3)$

Objective function two: send the horsepower of tugboat to overflow the sum minimum.

${\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="H2009101988898E00013.png,H2009101988898E00014.png"\; state="\{\{state\}\}">f</figure-callout>}}_2=\mathrm{Min}\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{{\mathrm{num}}_i}{\mathrm{\&Sigma;}}}(t\_l_{i,j}-r\_l_i)---(3-4)$

Break the dimension of f1 and t2, the objective function f of model can be expressed as:

f＝Min{w ₁×f ₁，w ₂×f ₂} (3-5)

Constraint condition is as follows:

1, send the horsepower of tugboat necessary enough, this is embodied on the tugboat grade:

t_l _i，j≥r_l _i，j，1≤i≤K，1≤j≤≤num _i，j，k?∈N ^* (3-6)

2, the numbering of many tugboats that same task is sent can not repeat:

t_n _i，j-t_n _i，k?≠0，1≤i≤K，1≤j，k≤num _i，j≠k，i，j，k∈N ^* (3-7)

3, the activity duration can not conflict before and after the tugboat:

$\&ForAll;t\_n_{i,j}=t\_n_{k,l},$ T_start _i，j≥T_end _k，l，1≤k≤i≤K，1≤j≤num _i，1≤l≤num _k，i，j，k，l∈N ^* (3-8)

It is as follows to relate to parameter declaration in the model:

Need the boats and ships quantity of dispatching, K=H+I+O+C among the K:DT.

Ship _i: the i bar ship among the DT.

Yi:Ship _iWait for the time of tugboat.

Num _i: to sHIP _iCarry out the tugboat quantity that operation needs, be also referred to as tugboat wheel number of times.

R_l _i: Ship _iThe minimum level of the tugboat that needs.

T_n _{I, j}: for Ship _iThe numbering of the tugboat that j round transferred.(j∈[1，num _i])

T_l _{I, j}: for Ship _iThe grade of the tugboat that j round transferred is with t_n _{I, j}There is corresponding relation.(j∈[1，num _i])

T_start _{I, j}: for Ship _iThe beginning preparation work of the tugboat that j round transferred constantly.(j∈[1，num _i])

T_end _{I, j}: for Ship _iThe end of job of the tugboat that j round transferred constantly.(j∈[1，num _i])

T: institute's accent tugboat grade of regulation is greater than the maximal value of required tugboat grade.If institute's accent tugboat grade greater than this maximal value, will cause the waste of wheel power greater than the quantity of required tugboat grade so, be called tugboat horsepower and overflow.

w ₁, w ₂: f ₁Weight coefficient with f.

After DT carried out suitable coding, just can seek optimum solution by the algorithm operation.

Each task (Ship _i) can come coded representation with a character string.Divide three classes with task, the task of approaching is class SA, and escort mission is class SB, and the task of leaving the port is class SC, and the coding of each task can be expressed as form respectively so:

The task that approaches S _{A, i}=(" A ", P _{A, i}, D _i, l _{1, i}, l _{2, i}..., l _{N, i}) (3-9)

Escort mission S _{B, i}=(" B ", P _{B, i}, A _i, l _i) (3-10)

The task of leaving the port S _{C, i}=(" C ", P _{C, i}, H _i, l _{1, i}, l _{2, i}..., l _{N, i}) (3-11)

Shifting berth task S _{D, i}=(" D ", P _{F, i}, P _{T, i}, l _{1, i}, l _{2, i}..., l _{N, i}) (3-12)

In the above-mentioned coding, each task is all with a character beginning; P _{A, i}, P _{B, i}, P _{C, i}At S _{A, i}, S _{B, i}And S _{C, i}In the numbering in the corresponding operation berth of expression respectively; P _{F, i}Expression shifting berth ship shift out berth, P _{T, i}The immigration berth of expression shifting berth ship; D _iFor the formation of approaching apart from correction; A _iNumbering for the anchorage; H _iBe the convoy mark; l _{1, i}, l _{2, i}..., l _{N, i}It is a tugboat rate sequence.The tugboat number that n needs for this shipping work.Though the dissimilar tugboat differences that need with the boats and ships of different tonnages, but under set port tug allocation scheme, can determine needed minimum tugboat quantity and tugboat grade for every ship, therefore can directly ship information and tugboat conditions of demand be encoded.With certain task of approaching is example, provides tugboat operation synoptic diagram (Fig. 1),

The particle cluster algorithm of tug dynamic scheduling problem is used

Suppose certain dynamic task set D by p the task that approaches, q leave the port task and r escort mission composition, i.e. D=[S _{B, 1}, S _{B, 2}..., S _{B, r}, S _{C, r+1}, S _{C, r+2}..., S _{C, r+q}, T _{A, r+q+1}, T _{A, r+q+2}..., T _{A, r+q+p}], the tugboat round that task i wherein needs is num _i, candidate solution dimension d can be expressed as so:

$d=\underset{i=1}{\overset{H+C+O+I}{\mathrm{\&Sigma;}}}{\mathrm{num}}_i---(3-13)$

Population scale N (being the quantity of candidate solution in the population) and d have confidential relation.If craft tug adds up to M, then the capacity of solution space should be M ^dTherefore, it is enough big to keep N on the one hand, is absorbed in the possibility of locally optimal solution to reduce algorithm; Must limit N on the other hand again in certain scope, with the working time of control algolithm ^[18]The full mold variable string that it is d that population can be expressed as N length, the span of each variable all is (0.6 in the string, M+0.4), and the scheme string that adopts the method round up that the candidate of full mold is unstringed and projected into integer, each integer in the scheme string is the numbering of corresponding operation tugboat; The full mold variable string that it is d that the kind group velocity also can be expressed as N length accordingly.Such coded system can make the hunting zone of algorithm cover all feasible solutions, but also can make simultaneously and some infeasible solutions occur in the population, and these are separated and can be eliminated by penalty function and iterative process.

The design of fitness function f:

The fitness function f of tug dynamic scheduling problem is generated by four partial weightings.Suppose that certain candidate solution in the population is:

A _i＝[x ₁，x ₂，...，x _d]，x _j∈[0.6，(M+0.4)]，x _j∈R，i∈[1，N]，i∈N ^* (3-14)

M is the port tug total quantity in the following formula.So corresponding, the fitness function of candidate solution can be write as f (D, Ai).

(1) level adaptation degree function L (D, Ai)

Reference model is introduced penalty function, and the level adaptation value that defines each tugboat round is as follows:

In the following formula, send for institute in the scheme that the not enough and grade of tugboat grade is excessive to be provided with different penalty factors.The level adaptation degree function L of Ai (D Ai) defines according to following formula:

$L(D,A_i)=\underset{j=1}{\overset{d}{\mathrm{\&Sigma;}}}l\left(j\right)---(3-16)$

(2) postpone fitness function Y (D, Ai) and conflict fitness function C (D, Ai)

The deciphering procedure relation of the computation process of these two functions and coding is tight.With the example that is read as, provide coding and separate read procedure as shown in Figure 2 the task string that approaches.

Consider the front and back operation correlativity of tugboat operation, calculating certain tugboat subjob can be with reference to the time and the position of this tugboat end of job last time during the start time, and finishes the back and noted down in time and position that this tugboat finishes operation calculating.

Delay fitness function Y (D is the summation that the entering and leaving port boats and ships are waited for the activity duration Ai), and its expression formula is as follows:

$Y(D,A_i)=\underset{j=r+1}{\overset{r+q+p}{\mathrm{\&Sigma;}}}{Y}_j---(3-17)$

The unit of following formula is second.

(D is for fear of the duplicate allocation phenomenon Ai) to conflict fitness function C, promptly sends same tugboat to go a plurality of tugboat rounds of certain boats and ships and the penalty function that designs.Separate run through the task string after, for each bar tugboat tj (1≤j≤M, M are the total number of tugboat), the set of an activity duration segment record can be arranged all:

${\mathrm{Period}}_j=\left(\right[T_{\mathrm{start},1},T_{\mathrm{end},1}],[T_{\mathrm{start},2},T_{\mathrm{end},2}],...,[T_{\mathrm{start},n_j},T_{\mathrm{end},n_j}\left]\right)---(3-18)$

In the following formula, nj is the task quantity of tugboat j.If defined function Length ([A, B])=| B-A|, so C (D Ai) just can be expressed as:

$C(D,A_i)=\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{k\&le;n_j,l\&le;n_j,k\&NotEqual;l}{\mathrm{\&Sigma;}}\mathrm{Length}\left(\right[T_{\mathrm{start},k},T_{\mathrm{end},k}\left]I\right[T_{\mathrm{start},l},T_{\mathrm{end},l}\left]\right)---(3-19)$

Unit is second in the following formula, and M is a total number of tugboat.Among the Ai if call the situation that same tugboat goes the different tugboats wheel subjob of certain boats and ships, just have C (D, Ai)＞0.

(3) utilization factor fitness function R (D, Ai)

This function calculation is according to the tugboat average utilization of option A i scheduling.Function defines according to following formula:

$R(D,A_i)=\frac{\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n_i}{\mathrm{\&Sigma;}}}\mathrm{Length}\left(\right[T_{\mathrm{start},j},T_{\mathrm{end},j}\left]\right)}{M\&times;T_D}---(3-20)$

Unit is second in the following formula, and TD is dynamic T.T..

(4) total fitness function f (D, Ai) differentiation of He Xieing

Total fitness function f (D Ai) according to the situation of actual job, breaks the dimension of above four fitness functions and suitably weighting generation, and is as follows:

$f(D,A_i)=L(D,A_i)+\frac{1}{600}\&times;Y(D,A_i)+100\&times;C(D,A_i)+R(D,A_i)---(3-21)$

It is fixed that coefficient in the following formula is got with reference to the demand of harbour actual schedule operation situation and algorithm.Preferential with the smooth degree in entering and leaving port, taking into account simultaneously and saving tugboat horsepower is starting point, thus L (D Ai) keeps original numerical value constant, and (D, power Ai) is made as 1/600 to Y.(D Ai)＞the 0th, allows the situation of appearance to C anything but in the actual schedule, therefore give acting temporarily as to penalizing of its maximum.Therefore (D Ai) is a more less important reference index to R, keeps original value constant here.(D, Ai)＜1000 o'clock, scheme is feasible as f.

Population evolutionary process and end condition:

After initialization had generated population and planted group velocity, the evolution of population was carried out according to standard particle group algorithm substantially, carried out the renewal of population iteration according to the population optimum solution and the particle optimum solution in each generation.Uniquely different be the aceleration pulse c among the present invention ₁, c ₂Be not constant, its value is stipulated according to following formula:

$c_i=c_{i,0}\&times;(1-\frac{\mathrm{gene}}{\mathrm{most}}),i=\mathrm{1,2}---(3-22)$

C in the formula _{I, 0}Be the aceleration pulse initial value of determining, most is total algebraically that calculates, and gene is current algebraically.This way is more common, can improve near the convergence situation of algorithm extreme point.

If the population optimum solution adaptive value in certain generation is less than 1, then algorithm finds satisfactory solution, calculates to stop; If the optimum solution of population within 1 to 1000, and 600 generations more excellent separate or total evolutionary generation has reached most do not appear, then algorithm finds feasible solution, calculates to stop; If during greater than most, population optimum solution adaptive value is still greater than 1000 up to total evolutionary generation, then algorithm can not find feasible solution, calculates to stop.

According to the number of particle and the initial velocity of particle in the length decision population of task string.Each particle, or be called candidate solution, representing one of tug dynamic scheduling to find the solution scheme.The dimension of candidate solution is the round sum that the tugboat of being sent is dispatched.

Set up the fitness function f of this tug dynamic scheduling model in the model, therefore found the solution the adaptive value (fitness value) of each particle, shown as the delay of promptly calculating the time in the scheduling that is produced in every kind of tug dynamic scheduling scheme.

To calculate the gained optimum solution simultaneously and note, each optimum solution has all comprised the value of four penalty functions and total adaptive value, and the round of sending tugboat.

So for this model, the numerical value of adaptive value is more little, just mean that also the time that this tug dynamic scheduling scheme wasted is few more, scheme is excellent more.

Population is evolved and is upgraded:

Each particle after the iteration, all can produce an adaptive value each time, if the adaptive value of current this generation less than the adaptive value of previous generation, this life of particle will be upgraded its particle optimum solution automatically so.And the optimum solution of population then is to be upgraded by the optimum solution in the optimum solution of all particles.

This is expression just, if the tugboat of certain tug dynamic scheduling scheme sent changes, postpone if can obtain less time, just show that it is feasible that such sending adjusted, promptly retain the scheduling scheme after the current improvement, continue next round is improved.

According to particle optimum solution p (best) and population optimum solution g (best) and current position and the speed of each particle, decide next particle position and speed.

In the various candidate schemes of tugboat scheduling, each scheme is all passed through and is constantly sent order and tugboat round, reduces the delay of the time that occurs in the scheduling as much as possible.

Judge the termination condition:

Satisfy any one condition in following 3 conditions, program run is promptly ended, and shows the result that current operation produces.

(1) iterations reaches 3000

The many iteration that can carry out 3000 times are done in the operation of representation program, if in 3000 iteration, there all do not have to find certain adaptive value to remain after through 600 iteration to be constant, and program run was ended immediately to 3000 generations so, and the optimum solution of getting current population is that algorithm output is separated.

(2) adaptive value＞1, and continuous 600 generations remain unchanged

In program operation process, if adaptive value, does not still change after 600 generations at subsequent iteration, the present invention just thinks that this adaptive value at this moment is as satisfactory solution so.

(3) adaptive value＜1

If adaptive value less than 1 situation, is just thought and is found optimum solution.

Improvement to particle cluster algorithm

According to the particle cluster algorithm of standard, have only the optimum solution of each particle in all generations before, just can be retained and be used as the foundation that particle is evolved.This way has been missed some valuable information in the iterative process, because the suboptimal solution of each particle or also may exist the possibility more excellent than the optimum solution of another particle than suboptimal solution, therefore the way of the optimality of single consideration particle itself in the standard particle group algorithm, may cause the loss of the valuable suboptimal solution of other particles in the population, make algorithmic procedure convergence of algorithm in early stage speed be restricted.In addition, in the algorithmic procedure later stage, when have a few all when extremely approaching the population optimum solution, if the population optimum solution of this moment is locally optimal solution but not globally optimal solution, algorithm is easy to be absorbed in locally optimal solution so, and can not find globally optimal solution.The present invention attempts to collect the method that combines with entropy function with the elite, improves these two problems of standard particle group algorithm.

Elite's collection is during each particle all in whole iterative process are separated, the set of separating of the some of adaptive value optimum.Some are valuable in this set meeting reservation, still uncared-for separating in standard particle group algorithm iteration process.The suitable moment in iterative process, the elite concentrate some separate some that can be used to replace in the population and separate.

Program is at first moved after the generation, separates for preceding 20 that get the adaptive value minimum, forms best 20, and according to from small to large series arrangement.Program and then operation then, false code is write as follows:

If?fitness?value＜best?20

Then?update?the?current?new?fitness?value?in?best?20

Kick?out?the?original?No20?and?put?the?new?fitness?value?into?the?queue

Be that original elite concentrates the 20th value of previous generation to be excluded, this optimum solution of this moment according to the arrangement of its size, is substituted on the position of corresponding 1-20, elite's collection constantly obtains upgrading like this, and the elite concentrates and enumerated 20 kinds of optimum scheduling schemes in all tugboat scheduling possibility schemes.

Entropy is chaotic and unordered tolerance.Entropy is big more, and confusing degree is big more.In algorithm, can represent the distribution range of whole population in solution space with entropy.Population distributes extensively more, and entropy is just big more.Along with the carrying out of iteration, each particle levels off to optimum solution gradually and gathers relatively, and the entropy of population also reduces gradually with this process.The present invention wishes suitably to keep the range of distribution of particles on the basis that guarantees algorithm the convergence speed, to keep the hunting zone of population, avoids algorithm convergence in locally optimal solution.

The present invention is with the distance between each particle in the population and other all particles and be defined as entropy, and its value is entropy.Therefore entropy function also can be expressed as the range formula of particle in the N dimension space, that is:

Shang＝[(a1-a2) ²+(b1-b2) ²+…+(x1-x2) ²] ^1/2

1 and 2 represent 2 particles respectively in the following formula, and x represents the dimension in this space, 2 particle places.

The present invention sets up a standard entropy variable in iterative process, the entropy situation of iterative process before preserving.The mean value of entropy of at first getting preceding ten generations is as the standard entropy, according to the variation tendency of average entropy of per ten generations later on, the standard entropy made adjustment then.If the average entropy in certain ten generation is less than the standard entropy, then the optimum solution of concentrating with the elite once goes to replace randomly separating arbitrarily in the population, and hope can be kept entropy.If in 40 continuous generations, four replacements make that all entropy diminishes, and at this moment just according to the average entropy in last ten generations the standard entropy are adjusted.This way has more reliability and evenness.

In the operational process of system, in case entropy is less than standard value, then the optimum solution that the elite is concentrated is substituted in the separating of original group randomly, the purpose of this way is to wish can accomplish to control artificially the variation range of entropy, it is too fast to avoid entropy to reduce system's speed of convergence of being caused rapidly, and can't find the reasonably situation of optimum solution.

The introducing of elite collection and entropy function is two defectives of wishing to improve to a certain extent the particle cluster algorithm that preamble mentions.In the early stage of algorithm iteration process, separating that the introducing elite concentrates can be accelerated convergence of algorithm speed; And, introduce the hunting zone that can suitably keep algorithm of separating that the elite concentrates in the later stage of algorithm iteration process, and make algorithm have more opportunity to jump out local optimum, find globally optimal solution.In this process, weigh the range of population with entropy, the coverage of promptly separating is as an index of algorithm performance.

Description of drawings

Further specify the present invention below in conjunction with accompanying drawing and case study on implementation.

Fig. 1 tugboat operation synoptic diagram;

Fig. 2 coding is separated read procedure;

Tugboat grade classification table in Fig. 3 example;

Fig. 4 task string encoding;

Fig. 5 tugboat is assisted big ship alongside synoptic diagram;

Fig. 6 tugboat big ship synoptic diagram that leaves the port that escorts;

Fig. 7 experimental result record;

Fig. 8 algorithm improves the adaptive value situation of change of back optimum solution;

Fig. 9 optimum solution value condition;

Figure 10 optimum solution correspondence is sent the tugboat number table of work;

Figure 11 task scheduling scheme example;

Figure 12 experimental result record;

Figure 13 optimum solution value condition;

The adaptive value situation of change of optimum solution contrast before and after Figure 14 algorithm improves;

Situation of change contrast before and after Figure 15 entropy is replaced;

The contrast of scheduling scheme before and after Figure 16 improves.

Embodiment

Below by a case study on implementation, further specify the present invention.

In this example, T _D=14400 seconds, M=16, N=8d, t=2, v _Max=20, most=3000.N=8d is exactly the mode that obtains of population quantity.Population quantity is a wheel 8 times of sub-quantity in the task.T be the tugboat grade that allows overflow place's quantity.The grade classification of each tugboat is shown in tugboat grade classification table in Fig. 3 example, and used data are the harbour service real data.The task coding of this example is shown in Fig. 4 task coding:

Illustrate the implication of above-mentioned task string encoding:

B, 22,2,2 represent escort mission respectively, No. 22 operations in the berth, the anchorage is 2, the tugboat grade that need send operation is 2, it is the tugboat of 3000-3500 horsepower, corresponding harbour can be numbered 5,6,7 for the tugboat of allotment, wherein Tiao Pei tugboat grade can be greater than required grade, but can not be less than required grade.The operation situation is assisted shown in the big ship alongside synoptic diagram as Fig. 5 tugboat: C, 4,0,2,2 represent the task of leaving the port, No. 4 operations in the berth respectively, need not escort, need send the tugboat of operation to want 2 altogether, grade is all 2, it is the tugboat of 3000-3500 horsepower, corresponding harbour can be numbered 5,6,7 for the tugboat of allotment, wherein Tiao Pei tugboat grade can be greater than required grade, but can not be less than required grade.The operation situation is left the port shown in the synoptic diagram as Fig. 6 tugboat big ship that escorts:

This selected task dynamically has 15 job tasks among the figure, and each capitalization English letter is represented an operation, and will send corresponding tugboat respectively and go to assist big ship to finish the entering and leaving port.

Solving result after algorithm improves

With algorithm model operation ten times, 10 operation results that obtain are shown in Fig. 7 experimental result record:

L wherein, Y, what C, R represented respectively is the calculated value of four fitness functions, and f represents total adaptive value, and evolutionary generation is represented when program run finishes, the iterations that particle upgraded.And mean value has been noted this average result of 10 times, be used for and improve before the result make comparisons, relatively have reliability.

Ten computings all obtain feasible solution, and all do not have to surpass 2.00, get the once analysis of total adaptive value minimum, and wherein the final iterations of particle is 3000 times, and corresponding adaptive value situation of change such as Fig. 8 algorithm improve the adaptive value situation of change of back optimum solution; The value condition of optimum solution is as shown in Figure 9: after each numerical value rounded, shown in the tugboat number table that Figure 10 optimum solution correspondence is sent work, numeral was the numbering of sending the tugboat that works out in the table.Optimum solution is corresponding with plan, can be clear which bar tugboat is each task send go work respectively.Figure 11 is a selected parts example of scheduling scheme, and this plan target is an escort mission, No. 22 operations in the berth, and the anchorage is 2, the tugboat grade that need send operation is 2.Scheduling is sent and is numbered 8 tugboat, this tugboat grade is 3, satisfy the operation needs, but may since other tugboat of little level carry out other work and can't bear this task, the perhaps tugboat of rank 3 task that can fulfil assignment more efficiently than the tugboat of rank 2.Send time representation to send No. 8 tugboats to go to carry out convoy operations the 1112.6th second the moment, finish timetable be shown in the 11461.3rd second the time after tugboat finish this job task.The activity duration that whole escort mission continues is 10348.7 seconds, adds up to 2.87 hours.Solving result before algorithm improves: with algorithm model operation ten times, the result is shown in Figure 12 experimental result record: ten computings all obtain feasible solution, and 9 times result is in 5.Get the once analysis of total adaptive value minimum, the value condition of optimum solution is shown in Figure 13 optimum solution value condition.

The convergence after algorithm improves and the improvement of scheduling scheme

Constringent improvement

The adaptive value situation of change of optimum solution can see that both restrain the comparison of situation before and after the contrast algorithm improved.Can see the adaptive value situation of change contrast of optimum solution before and after algorithm improves by Figure 14

Solid line is represented among the figure is to add elite's collection and entropy function, and the result who obtains after particle cluster algorithm is improved is the result who uses standard particle group algorithm before improving and dotted line is represented.Can see that the speed of convergence that obtains after particle cluster algorithm is improved is faster, and can find more excellent adaptive value.

The present invention passes through after introducing entropy function, the replacement that utilizes the elite to collect, program before and after replacing is moved 10 later average case separately respectively done analysis, observe changes of entropy for the influence of searching optimum solution and the variation tendency that produces replacement back entropy, the changes of entropy situation of replacing front and back is to shown in the situation of change contrast of replacing front and back such as Figure 15 entropy:

Darker line is represented among the figure is to add elite's collection and entropy function, and the result who obtains after particle cluster algorithm is improved is the result who uses standard particle group algorithm before improving and more shallow line is represented.Can see that obviously faster than before improving, promptly speed of convergence is significantly improved the arithmetic result entropy decline after iterative process is improved in earlier stage compared with before-improvement; Arithmetic result entropy after the iterative process later stage improves shows more stablely, and decline rate is more slow compared with before-improvement.In addition, not only the algorithm after the improvement has found better separates, and the x axle span of dark line is significantly less than shallow line among the figure, and this shows that to improve iterations that the back algorithm finds the result to experience also a lot of than reducing before improving.

Improve the contrast of front and back scheduling scheme by Figure 16, numerical results before and after algorithm improves as can be known, tug dynamic scheduling scheme after the improvement is relatively large in the improvement that postpones on the fitness function, 427.6 be reduced to 35.34 before original improvement, (the data overstriking shows in the table) also shortened with regard to the time that boats and ships wait tugboat assistance alongside has been described, therefore from the angle of efficient, the scheduling scheme after the improvement must be more efficiently.

In addition, the quantity of overflowing from grade also as can be seen, before algorithm improves, need overflow 11 grades, the utilization factor that tugboat has been described still exists very big waste, and after the algorithm improvement, overflowing of grade is reduced to 9, grade overflow the waste that is bound to cause tugboat horsepower, go work with the tugboat than high-power, cost also can correspondingly increase, under market economy environment of today, harbour can be paid attention to the saving of cost more, therefore, we need accomplish to reduce as much as possible to send the overflowing of tugboat grade of operation, accomplish to carry out schedule job really, from the angle of cost, the scheme after the improvement also is excellent more.

Above result's running environment is under the Windows XP, and in the eM-Plant program of writing voluntarily, the computing machine relevant configuration is AMD double-core 1800MHz, and 2G internal memory, calculating process are handled about 20 minutes of averaging time of CPU.Consider that the dynamic plan of tugboat just was delivered to tugboat traffic department in general about 10 hours in advance, therefore, it is suitable that this algorithm application is dispatched in actual tugboat.

Claims (4)

1. port tug dynamic scheduling method is characterized in that: may further comprise the steps:

1) on the basis of analyzing the dynamic plan of tugboat, sets up the tug dynamic scheduling model, propose a kind of dynamic coding mode;

2) with particle swarm optimization algorithm problem is carried out fitness function design and scheduling scheme is found the solution;

3) on the basis of standard particle group algorithm, introduce the notion of elite's collection and entropy function, algorithm is improved;

4) by with improve before program run result's comparing result analysis, find that the result obtains optimization and improvement to a certain degree really; Further proof elite collection and the Application feasibility of entropy function in particle cluster algorithm find the dispatching method that makes the boats and ships entering and leaving port more smooth.

2. dispatching method according to claim 1 is characterized in that: the condition of described step 1) is: ships entered harbor always can arrive the limit, berth smoothly, can not stop because of the obstruction of preceding ship; Mission Operations before dynamically escort mission begins to dispatch is all finished; The activity duration can not conflict before and after the tugboat.

3. dispatching method according to claim 1 is characterized in that: described step 2) numerical value of fitness function adaptive value is more little, just means that also the time that the tug dynamic scheduling scheme wasted is few more, and scheme is excellent more.

4. dispatching method according to claim 1 is characterized in that: choosing of described step 3) elite's collection is in order to keep the optimum solution of all particles of whole population in whole iterative process, to be the population optimum solution; Replace general solution with separating of concentrating of elite, iterative process early stage can accelerating algorithm speed of convergence, can keep the stable of entropy in the iterative process later stage simultaneously.

Images