Port bulk cargo storage location intelligent distribution method and system

Abstract

The invention discloses a port bulk cargo storage location intelligent distribution method and system, and belongs to the technical field of port storage location management and intelligent distribution. The method comprises the following steps: collecting historical data related to port cargo scheduling, and defining a storage location distribution demand; designing constraint conditions for a storage location distribution process; constructing a multi-objective optimization model based on a genetic algorithm according to the constraint condition; generating an initial feasible solution meeting all hard constraints by adopting a greedy pre-screening algorithm; taking the initial feasible solution as the input of a multi-objective optimization model, and solving the optimal solution of storage location allocation by using an improved NSGA-II algorithm; and according to an optimization result, distributing the goods to the optimal storage location. By adopting the technical scheme of the invention, the problems of low storage location utilization rate, load imbalance and high delivery cost in port multi-type cargo scheduling can be effectively solved, so that the port resource utilization rate and the operation efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of port storage location management and intelligent allocation, and particularly relates to an intelligent allocation method and system for port bulk cargo storage locations based on constraint hierarchical processing and multi-objective optimization. Background Art

[0002] With the rapid development of global trade, ports, as the core hubs of the logistics system, are facing problems such as low efficiency of bulk cargo storage location allocation, insufficient resource utilization rate, and weak dynamic adjustment ability. Traditional storage location management relies on manual experience and is difficult to cope with challenges such as diverse cargo forms, complex turnover times, and dynamic spatial changes, resulting in pain points such as low storage location utilization rate, long cargo detention time, and high scheduling costs. Especially in the scenario of bulk cargo management, storing goods without considering the outbound time will lead to an increase in the rehandling rate, which in turn increases costs. Therefore, how to achieve the intelligence, dynamics, and visualization of storage location allocation has become a key issue in improving port operation efficiency.

[0003] After retrieval, a patent application case with the Chinese patent application number CN202410459782.9 discloses a method for allocating steel coil storage locations based on multi-objective optimization. This method constructs an optimization model by analyzing the steel coil turnover rate and within-class dispersion, converts the multi-objective problem into a single objective by combining expert scoring and weighting, and uses the hunting algorithm to solve the vacant storage location allocation scheme to improve the steel coil scheduling efficiency and storage identity. However, its model design focuses on steel coil characteristics (such as within-class dispersion calculation and crane travel speed optimization), and its application scope is limited to the steel industry. Moreover, relying on static weighting of expert experience results in fixed target weights and is difficult to adapt to the dynamic constraint scenarios of multiple cargo types in ports. In addition, although the hunting algorithm enhances the global search ability through the cosine-type motion formula, its solution space depends on random initialization and lacks a high-quality initial solution generation mechanism, making it easy to fall into local optima or have insufficient convergence speed. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent allocation method and system for port bulk cargo storage locations. By dynamically managing the priorities of hard constraints and soft constraints and combining greedy pre-screening with an improved NSGA-II algorithm, it can effectively solve the problems of low storage location utilization rate, load imbalance, and high outbound costs in the scheduling of multiple types of cargo (such as containers, steel coils, and bulk cargo) in ports, thereby improving port resource utilization rate and operation efficiency.

[0005] To achieve the above object, the technical solution provided by the present invention is as follows:

[0006] In the first aspect of the present invention, an intelligent allocation method for port bulk cargo storage locations is provided, including:

[0007] Collect historical data related to port cargo scheduling, including cargo type, volume, weight, inbound and outbound times, and clarify the requirements for storage location allocation;

[0008] Design constraint conditions for the storage location allocation process, where the constraint conditions include hard constraints and soft constraint conditions;

[0009] Based on the constraint conditions, construct a multi-objective optimization model using the genetic algorithm;

[0010] Combined with the historical data related to port cargo scheduling, use the greedy pre-screening algorithm to generate an initial feasible solution that satisfies all hard constraints;

[0011] Use the above initial feasible solution as the input of the multi-objective optimization model, and use the improved NSGA-II algorithm to solve the optimal solution of storage location allocation;

[0012] According to the optimization result, allocate the cargo to the optimal storage location.

[0013] According to any of the technical solutions described in the first aspect of the present invention, the hard constraints include capacity, stacking shape, stacking order, same-type constraint, and load-bearing constraint, where:

[0014] Capacity constraint: The volume of the cargo allocated in storage location k does not exceed its maximum capacity C k , that is

[0015] Stacking order constraint: The outbound time of the upper-layer cargo is greater than or equal to the outbound time of the lower-layer cargo;

[0016] Same-type cargo constraint: Only the specified cargo types can be stacked in storage location k;

[0017] Storage location load-bearing constraint: The total weight W k after storage location scheduling does not exceed the maximum load-bearing limit W max,k ;

[0018] Stacking shape constraint: Ensure that the stacking shape of the cargo conforms to physics to prevent affecting the stability of the cargo. Specifically, the impacts of slip, stress, displacement, etc. generated when the cargo is stacked can be comprehensively considered.

[0019] According to any of the technical solutions described in the first aspect of the present invention, the soft constraints include proximity constraint, outbound time concentration constraint, storage location load balance constraint, stacking height recommendation constraint, transportation path optimization constraint, and dynamic adjustment ability constraint, where:

[0020] The objective function of the proximity constraint is: minimize∑ {i,j∈B} (d(k i ,k j )×S {i,j} ), where, d(ki , k j ) is the physical distance between storage locations i and j; S {i,j} is the cargo correlation matrix;

[0021] The objective function for the outbound time concentration constraint is: minimize ∑ {k∈K} σ t (T b | b ∈ k), σ t is the time standard deviation, and T b is the outbound timestamp of cargo b;

[0022] The objective function for the storage location load balance constraint is: W k is the total weight of the cargo in storage location k; W max,k represents the maximum load limit of storage location k;

[0023] The objective function for the stacking height recommendation constraint is: h k is the actual stacking height of storage location k, and H recommend,k is the recommended safe stacking height of storage location k;

[0024] The objective function for the transportation path optimization constraint is: minimize ∑ {b∈B} (f b × d(k b , G)), f b is the expected inbound and outbound frequency of the cargo, and d(k b , G) is the physical distance between the storage location k where cargo b is located b and the main road location G;

[0025] The objective function for the dynamic adjustment ability constraint is: V b is the volume of cargo b, C k is the maximum capacity of storage location k, and P k is the reserved value coefficient of storage location k.

[0026] According to any one of the technical solutions described in the first aspect of the present invention, according to the constraint conditions, a multi-objective optimization model is constructed based on a genetic algorithm. The total objective function of this multi-objective optimization model is:

[0027]

[0028] In the above formula:

[0029] C in,b,k is the inbound cost of cargo b entering storage location k, and w in is the weight of the inbound cost;

[0030] C st,b,kis the storage cost of goods b in storage location k, w st is the weight of the storage location storage cost;

[0031] represents the storage location utilization rate, where V k is the volume actually occupied in storage location k, C k is the maximum capacity of storage location k, w u is the weight of the storage location utilization rate;

[0032] represents the storage location load balance, where W k is the total weight of the goods in storage location k; W max,k represents the maximum load-bearing limit of storage location k, |K| is the total number of storage locations, w l is the weight of the storage location load balance.

[0033] According to any one of the technical solutions described in the first aspect of the present invention, the use of the greedy pre-screening algorithm to generate an initial feasible solution that satisfies all hard constraints includes:

[0034] (1) Dynamic priority formula design

[0035] Combining spatio-temporal coupling attenuation, dynamic load-bearing risk, and spatial fitness, construct the following dynamic priority function:

[0036] Priority(b) = α·E b +β·B l +γ·U l

[0037] where E b , B l , U l are spatio-temporal coupling attenuation, dynamic load-bearing risk, and spatial fitness respectively, and α, β, γ are the corresponding dynamic weight coefficients;

[0038] Spatio-temporal coupling attenuation E b is calculated as follows:

[0039]

[0040] In the above formula, λ is the time decay coefficient, which determines the decay rate of the overall fitness E b due to the goods outbound time; T b is the outbound timestamp of goods b; d b is the physical distance from the goods to the storage location, d0 is the distance threshold, γ d is the distance sensitivity coefficient, which determines the sensitivity of the distance d b from the goods to the storage location to the spatial fitness;

[0041] Storage location dynamic load-bearing risk Bl The calculation is as follows:

[0042]

[0043] In the above formula, σ risk (l) is the load-bearing risk coefficient of storage location l, and σ threshold is the load-bearing safety threshold;

[0044] The space adaptability is calculated as follows:

[0045]

[0046] U l is the space utilization rate, V b is the volume of the current goods b to be allocated, V i is the volume of the existing goods i in storage location l, C l is the maximum volume of storage location l, Atop is the available area at the top layer of the current storage location, A l is the area of the reference layer of the storage location;

[0047] (2) Traverse the storage locations, screen the feasible storage locations that meet all hard constraints to obtain the initial feasible solution; calculate the dynamic priorities of all feasible storage locations, sort them in descending order of weights, and update the storage location status dynamically.

[0048] According to any one of the technical solutions described in the first aspect of the present invention, the dynamic weights α, β, γ are adjusted according to the storage location status:

[0049]

[0050] According to any one of the technical solutions described in the first aspect of the present invention, the improved NSGA-II algorithm adopts an adaptive crossover probability, and the crossover probability calculation formula is as follows:

[0051]

[0052] Among them, η represents the crossover sensitivity coefficient, which determines the non-linear mapping strength between the cosine similarity (similarity of the parent paths) and the crossover probability; and are the path matrices of the parent individuals x a and x b respectively; represents and the cosine similarity between;

[0053] According to any one of the technical solutions described in the first aspect of the present invention, when p c →1, the PathCopy crossover operation is performed, and when p c →0, the uniform crossover operation is adopted.

[0054] According to any of the technical solutions described in the first aspect of the present invention, when the improved NSGA-II algorithm performs the mutation operation, it first verifies the capacity and load-bearing in the hard constraints, and synchronously checks the current load ∑V path of the child storage location k i and W b when generating the Mask k . If it exceeds the limit, it is forced to be set to 0 to ensure compliance with the C H constraint set;

[0055] If the goods outbound time T i <T j , but it is stacked on the lower layer, then the patching strategy is triggered after crossover, and the positions are automatically exchanged to meet the C 顺序 :

[0056] Child = Mask path ⊙x a +(1 - Mask path )⊙x b

[0057]

[0058] The second aspect of the present invention also provides an intelligent storage location allocation system for port bulk cargo, including:

[0059] A data collection and storage location allocation requirement analysis module, which is used to collect historical data related to port cargo scheduling, including cargo type, volume, weight, inbound and outbound times, and clarify the storage location allocation requirements;

[0060] A constraint condition design module, which is used to design constraint conditions for the storage location allocation process, and the constraint conditions include hard constraints and soft constraint conditions;

[0061] A multi-objective optimization model construction module, which is used to construct a multi-objective optimization model based on the genetic algorithm according to the constraint conditions;

[0062] An initial feasible solution generation module, which is used to generate an initial feasible solution that meets all hard constraints by using a greedy pre-screening algorithm; and

[0063] A storage location allocation scheme optimization module, which is used to solve the multi-objective optimization model by using the improved NSGA-II algorithm to obtain the optimal storage location allocation scheme.

[0064] By using the storage location intelligent allocation system of the present invention, the storage locations of port cargo can be allocated according to any of the intelligent allocation methods described in the first aspect of the present invention, so as to realize the intelligent, efficient and accurate allocation of bulk cargo storage locations.

[0065] Adopting the technical solution provided by the present invention, compared with the prior art, the following beneficial effects can be achieved:

[0066] (1) The present invention proposes a storage location allocation method based on constraint hierarchical processing and multi-objective optimization. Through a hard and soft constraint hierarchical dynamic adjustment mechanism (such as hard rules like capacity, stacking order, load-bearing, etc. and flexible objectives like outbound cost, utilization rate, etc.), combined with a greedy pre-screening algorithm to generate a high-quality initial solution, and an improved NSGA-II algorithm is used to directly optimize the multi-objective Pareto solution set. Thus, it not only supports multiple types of goods such as steel coils, containers, and bulk cargo, but also can dynamically respond to the real-time operation requirements of the port (such as emergency order insertion, equipment load fluctuation), significantly improving the storage location utilization rate, load balance, and scheduling efficiency.

[0067] (2) Specifically, through the hierarchical processing of hard and soft constraints and dynamic adjustment of weights in the present invention, the rigid and flexible requirements of port operations can be taken into account, avoiding the optimization imbalance caused by static weights in traditional methods; by quickly generating a feasible solution that meets the hard constraints through the greedy pre-screening algorithm, the convergence speed of NSGA-II can be effectively improved; the improved NSGA-II algorithm combines hybrid coding and constraint-aware operations to output a diverse Pareto front solution set, supporting decision-makers to select the optimal solution according to scenario requirements; through strict solution set verification and dynamic termination mechanism, the feasibility and stability of the solution in actual port operations can be ensured, reducing manual intervention and scheduling costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the intelligent allocation process of port bulk cargo storage locations according to the present invention;

[0069] Figure 2 It is a comparison chart of the algorithm convergence speed in the embodiment of the present invention;

[0070] Figure 3 It is a comparison chart of the Pareto distribution in the embodiment of the present invention;

[0071] Figure 4 It is a comparison chart of the constraint condition satisfaction rate in the embodiment of the present invention;

[0072] Figure 5 It is a comparison chart of the algorithm efficiency in the embodiment of the present invention;

[0073] Figure 6 It is a comparison chart of the solution density in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0074] As an important hub of global trade, ports undertake huge tasks of cargo transfer. By integrating a hybrid coding strategy, an improved non-dominated sorting genetic algorithm (NSGA-II), and dynamic weight adjustment technology, the present invention can achieve multi-objective collaborative optimization of port berth allocation, especially applicable to the warehousing scheduling of highly volatile goods such as containers and bulk commodities (such as steel).

[0075] To further understand the content of the present invention, the present invention will be described in detail below in conjunction with specific embodiments.

[0076] As Figure 1 shown, the intelligent bulk cargo berth allocation method of the embodiment of the present invention includes the following steps:

[0077] S1. Data collection and demand analysis

[0078] Collect historical data on port cargo scheduling (cargo type, volume, weight, inbound and outbound time, etc.), analyze the cargo turnover rate and stacking rules, and clarify the berth allocation requirements.

[0079] S2. Hierarchical constraint condition design

[0080] For the berth allocation process, set hard constraints and soft constraints. Hard constraints include capacity, stacking shape, stacking order, same-type constraint, and load-bearing constraint. Soft constraints include proximity constraint, outbound time concentration constraint, berth load balance constraint, stacking height recommendation constraint, transportation path optimization constraint, and dynamic adjustment ability constraint. Specifically:

[0081] Hard constraints:

[0082] (1) Capacity constraint: The volume of the cargo allocated in berth k does not exceed its maximum capacity C k , that is

[0083] (2) Stacking order constraint: The outbound time of the upper-layer cargo is greater than or equal to that of the lower-layer cargo. That is, if cargo i is on the upper layer of cargo j, it must satisfy T i ≥T j , where T i represents the outbound time of cargo i;

[0084] (3) Same-type cargo constraint (consistency of same-type cargo types): Ensure the stacking limit of the same-type cargo in berth k to reduce possible confusion and errors, defined as: If then it cannot be stacked; if then it can only be stacked according to the specified cargo type;

[0085] (4) Berth load-bearing constraint: The total weight W k after berth scheduling does not exceed the maximum load-bearing limit Wmax,k , i.e., W k ≤W max,k ;

[0086] (5) Stacking shape constraint: Ensure that the stacking shape of the goods conforms to physics to prevent affecting the stability of the goods. In actual operation, for different bulk goods in the port, for example, H-beams are usually placed flat or horizontally, and are stacked in layers crosswise to achieve uniform weight distribution and stable stacking; steel coils are mostly stacked horizontally, supplemented with anti-slip and anti-rolling measures; in special cases, they may be stacked semi-vertically on special brackets. Therefore, the stacking shape constraints of the goods vary according to different goods.

[0087] Soft constraint:

[0088] (1) Transportation path optimization constraint: High-frequency incoming and outgoing goods are close to the main transportation artery, and the objective function is:

[0089]

[0090] f b is the expected incoming and outgoing frequency of the goods, and d(k b , G) is the physical distance between the storage location k where the goods b are located and the location G of the main artery. b

[0091] More preferably, through the distance attenuation term the soft constraint of bringing high-frequency goods close to the main artery is transformed into a computable exponential decay weight. For example, the decay coefficient of the storage location near the main artery (d(k, G) is small) approaches 1, giving a higher priority. The objective function of the optimized transportation path optimization constraint is:

[0092] (2) Proximity constraint (priority of goods correlation): That is, preferentially allocate associated goods (such as the same consignor / destination / transport vehicle) to adjacent storage locations, and the objective function is: minimize∑ {i,j∈B} (d(k i , k j ) × S {i,j} ), where d(k i , k j ) is the physical distance between storage locations i and j; S {i,j} is the goods correlation.

[0093] Furthermore, the embodiment of the present invention directly encodes the requirement of preferentially allocating adjacent storage locations according to the goods correlation through the reachability matrix . If the storage location k i is physically connected to k j and is not occupied, then otherwise it is 0 to avoid path conflicts. ​

[0094] The specific implementation method is as follows: First, perform path cost pre-calculation, offline calculate the shortest distance d(k, G) from all storage locations to the main road, and generate a distance matrix; secondly, dynamically update the reachability matrix. After each allocation, update the occupied storage location combinations are blocked.

[0095] More preferably, through the distance decay term, d(k i , k j ) is transformed into a computable exponential decay weight. Then the optimized proximity constraint objective function is:

[0096]

[0097] Through the above processing, the number of inventory reorganizations can be significantly reduced. Through explicit optimization of the path cost, the secondary handling of goods caused by unreasonable positions can be reduced; at the same time, through the reachability matrix filtering out unreachable storage locations can effectively ensure that the allocation plan meets the physical layout restrictions of the storage yard.

[0098] (3) Outbound time concentration constraint: The outbound times of goods in the same storage location should be as concentrated as possible to reduce the number of searches; that is, minimize the standard deviation of the outbound times of goods in the same storage location. The objective function is: minimize ∑ {k∈K} σ t (T b |b ∈ k), where σ t is the standard deviation of time, and T b is the outbound timestamp of goods b;

[0099] (4) Storage location load balance constraint: Minimize the difference in the weight load rates of storage locations (the weight loads of each storage location are relatively balanced to avoid local overload). The objective function is: W k is the total weight of the goods in storage location k; W max,k represents the maximum load-bearing limit of storage location k; the balance degree is measured by the difference between the maximum / minimum load rates;

[0100] (5) Stacking height recommendation constraint: Maximize the stacking height within the safe range to improve space utilization; the objective function is: h k is the actual stacking height of storage location k, and H recommend,k is the recommended safe stacking height of storage location k;

[0101] (6) Dynamic adjustment ability constraint: Maximize the stacking height within the safe range to reserve adjustment space for future inbound storage. The objective function is: V b is the volume of goods b, C k is the maximum capacity of storage location k, Pk The reserved value coefficient for storage location k (determined based on location, equipment accessibility, etc.).

[0102] S3. Construction of multi-objective optimization model

[0103] Construct a multi-objective optimization model according to the constraints. This model adopts the objective function of multi-objective optimization, specifically including inbound cost, storage cost, storage location utilization rate, and load balance, so as to ensure that multiple objectives are considered simultaneously in the optimization process.

[0104] 1) Inbound cost (C in )

[0105] The inbound cost is mainly related to the handling and storage of goods and can be expressed as:

[0106] C in = ∑ (k∈K) ∑ ∈b C in,b,k

[0107] where C in,b,k is the inbound cost of goods b entering storage location k, and K: the set of all storage locations.

[0108] 2) Storage cost of storage location (C st )

[0109] The storage cost of the storage location is mainly related to the residence time of goods in the storage location and the storage cost and can be expressed as:

[0110] where, b k is the set of goods stored in storage location k, and C st,b,k is the storage cost of goods b stored in storage location k.

[0111] 3) Storage location utilization rate (U)

[0112] The storage location utilization rate is an important indicator to measure the use efficiency of the storage location and can be expressed as:

[0113]

[0114] where V k is the actually occupied volume in storage location k, and C k is the maximum capacity of storage location k.

[0115] 4) Storage location load balance (L)

[0116] The storage location load balance aims to ensure that the load distribution of each storage location is uniform, avoiding overloading or idling of some storage locations and can be expressed as:

[0117]

[0118] where W k is the total weight of the goods in storage location k, W max,k is the maximum load-bearing limit of storage location k, and |K| is the total number of storage locations.

[0119] To design a weight function for the distribution characteristics of prediction errors, in order to balance these different objectives, the embodiments of the present invention introduce weights to adjust the importance of each objective. Assume the weights are w in , w st , w u and w l , then the comprehensive objective function is expressed as:

[0120] F(x) = w in ·C in + w st ·C st + w u ·(1 - U)+ w l ·L

[0121] where: the weight w u of the storage location utilization rate takes a negative value, so as to be used to maximize the utilization rate.

[0122] The selection of weights can be determined according to specific application scenarios and priorities. In actual applications, these weights can be dynamically adjusted according to different constraint conditions and optimization stages. To sum up, the total objective function is:

[0123]

[0124] S4. Greedy pre-screening to generate an initial solution

[0125] Adopt a greedy pre-screening algorithm to quickly generate an initial feasible solution that satisfies all hard constraints to ensure the feasibility of the solution. Filter the solution of the goods through the hard constraint conditions in the hierarchical constraint conditions, and combine the historical data in S1 to screen out the storage locations that do not meet the requirements for the goods to be stored. After finding the storage locations that meet the conditions for all goods, this set of solutions is the greedy initial solution.

[0126] Design of the dynamic priority formula:

[0127] Combined with spatio-temporal coupling attenuation, dynamic load-bearing risk, and spatial fitness, construct a dynamic priority function:

[0128] Priority(b) = α·E b + β·B l + γ·U l

[0129] (1) Spatio-temporal coupling attenuation

[0130]

[0131] where: λ is the time decay coefficient, and T b is the outbound timestamp of goods b; d b is the physical distance from the goods to the storage location, d0 is the distance threshold, and γ d is the distance sensitivity coefficient, which determines the sensitivity of the distance d of the goods to the storage location b to the spatial fitness;

[0132] The spatio-temporal coupling decay is the core mechanism for dynamically evaluating the priority of goods handling in this storage location allocation method. Its essence is to couple and model the time urgency and spatial fitness, and through the combination of the exponential decay function and the distance sensitivity function, realize the quantitative evaluation of the urgency degree of goods batches. The specific analysis is as follows:

[0133] 1) Physical meaning of spatio-temporal coupling

[0134] Time dimension: Through the exponential decay function reflects the urgency of the goods outbound time, and T b is smaller (more urgent), the time decay factor is larger, and the priority is higher.

[0135] Spatial dimension: Through the distance sensitivity function measures the spatial fitness between the goods and the target storage location, and d b is the physical distance from the goods to the storage location, and d0 is the distance threshold. The closer the distance, the higher the spatial fitness, and γ d is the distance sensitivity coefficient, which determines the sensitivity of the distance d of the goods to the storage location b to the spatial fitness, and γ d and the threshold d0 jointly define the "effective influence range" of the distance. By adjusting γ d , it is possible to control whether the priority change of the goods near the threshold is sensitive.

[0136] 2) Coupling method

[0137] Through multiplicative coupling

[0138] transforms the combined influence of time and space into a comprehensive index to ensure that:

[0139] Urgent and nearby goods: High spatio-temporal coupling value, preferentially allocated;

[0140] Urgent but distant goods: High time component but low spatial component, need to be weighed;

[0141] Non-urgent but nearby goods: May be delayed in allocation to free up resources for more urgent tasks.

[0142] (2) Dynamic load-bearing risk of storage location

[0143]

[0144] In the in - stock location dynamic load - bearing risk model, the parameter design is used to quantify the load - bearing safety risk of the in - stock location, ensuring that the goods allocation does not exceed the physical bearing capacity of the in - stock location. Among them:

[0145] 1) σ risk (l): In - stock location dynamic load - bearing risk coefficient

[0146] Physical meaning: Reflects the risk degree of the current load - bearing state of in - stock location l, considering the following factors comprehensively:

[0147] Current load: The total weight W of the goods stacked in the in - stock location l 。

[0148] Historical load fluctuation: The standard deviation of the weight change of the in - stock location within the past time period (measuring the dynamic load - bearing stability).

[0149] Structural strength attenuation: The structural fatigue coefficient of the in - stock location due to long - term use (for example, the load - bearing capacity of an aging shelf decreases).

[0150] Calculation formula:

[0151] Among them, η w : Fluctuation penalty coefficient (amplifying the impact of historical load fluctuation);

[0152] σW l : The standard deviation of the weight of in - stock location l within the past week.

[0153] 2) σ threshold : Load - bearing safety threshold

[0154] Physical meaning: The maximum risk coefficient that the in - stock location allows to bear, usually less than 1 (for example, 0.8), ensuring a safety margin;

[0155] Setting basis: The maximum load - bearing W of the in - stock location design max,l , dynamic risk tolerance (the threshold may be reduced during peak hours for conservative allocation.

[0156] 3) The core objective of the dynamic load - bearing risk formula is:

[0157] Quantify risk:

[0158] When σ risk (l) → σ threshold When, B l → 0, indicating that the in - stock location is approaching the load - bearing limit and careful allocation is required;

[0159] When σ risk(l) << σ threshold When B l → 1, it indicates that the storage location is safe and can be preferentially allocated.

[0160] Dynamic priority adjustment:

[0161] In the greedy pre-screening algorithm, B l As a scoring factor for dynamic load-bearing risk, it directly affects the priority of the storage location. If B l is low (high risk), the total priority of storage location l decreases, reducing the probability of being selected.

[0162] (3) Space fitness (space utilization rate)

[0163]

[0164] U l is the space utilization rate, V b is the volume of the current cargo b to be allocated, V i is the volume of the existing cargo i in storage location l, C l is the maximum volume of storage location l, Atop is the available area at the top layer of the current storage location, A l is the area of the reference layer of the storage location.

[0165] 1) Volume utilization rate

[0166] is the volume V of the current cargo b to be allocated b and the total volume of the existing cargo in storage location l ∑i∈S l V i ; C l is the maximum volume of storage location l.

[0167] Significance: It reflects the filling degree of the storage location in the vertical direction, avoiding space waste caused by incomplete filling.

[0168] 2) Area fitness

[0169] A top : The available area at the top layer of the current storage location (affected by the shape of the stacked goods);

[0170] A l : The area of the reference layer of the storage location A l (the maximum bottom area of the storage location design);

[0171] Significance: It measures the flatness of the top layer space of the storage location to ensure the stacking stability of the goods (such as avoiding suspension or inclination).

[0172] Therefore, the dynamic priority formula becomes:

[0173]

[0174] Furthermore, the dynamic weights α, β, and γ are adjusted according to the bin status:

[0175] And α + β + γ = 1.

[0176] Specifically, when generating the initial solution based on greedy pre-screening:

[0177] (1) Priority sorting: Sort in ascending order of outbound time, descending order of volume, and clustering by cargo type;

[0178] (2) Hard constraint verification: Traverse the bins to filter out the feasible bins that meet all hard constraints;

[0179] (3) Greedy allocation strategy: Prioritize selecting the bin with the lowest outbound cost, the highest utilization rate, or the lowest load rate, that is, calculate the dynamic priorities of all feasible bins, sort them in descending order of weights, and dynamically update the bin status.

[0180] The greedy pre-screening algorithm can quickly generate an initial feasible solution on the premise of meeting all hard constraints. This initial solution provides a high-quality candidate solution set for subsequent multi-objective optimization, which helps to improve the convergence speed and solution quality of the NSGA-II algorithm.

[0181] S5. Multi-objective optimization based on the improved NSGA-II

[0182] Take the greedy initial solution as the input of the genetic algorithm, and use the improved NSGA-II algorithm to handle the multi-objective optimization problem in bin allocation to ensure finding multiple optimal solutions. During the solution process, dynamically adjust the weights of the objective function and the constraint conditions to ensure the rapid convergence of high-quality solutions. Strictly verify in the final solution set to ensure that all solutions meet the preset hard constraint conditions.

[0183] The present invention mainly optimizes the calculation formula of the crossover probability, the selection of the crossover operator, and the mutation operation on the basis of the existing NSGA-II algorithm.

[0184] Specifically, in some embodiments, the multi-objective optimization based on the improved NSGA-II includes:

[0185] (1) Initialize the population

[0186] Randomly generate an initial parental population with a size of N. The decision variables of each individual are generated through a uniform distribution or a specific distribution to ensure covering the entire search space, avoiding the initial solution being too concentrated, and providing a diversity basis for subsequent evolution.

[0187] (2) Fast non-dominated sorting

[0188] According to the objective function values, the population is divided into multiple non-dominated fronts. If for an individual x i all objectives are not worse than those of x j and at least one objective is better, then x i dominates x j . The first front F1 contains all non-dominated individuals, and subsequent fronts are formed by successively screening the remaining individuals, forming a hierarchical structure.

[0189] (3) Crowding degree calculation

[0190] Within the same front, after sorting the objective function values for each, calculate the sum of the normalized distances of the neighbors around the individual. The crowding degree of boundary individuals is set to infinity to ensure diversity. The higher the crowding degree, the sparser the region where the solution is located, and it is preferentially retained.

[0191] (4) Select, crossover, and mutate to generate offspring

[0192] Through tournament selection: randomly select two individuals, and preferentially select individuals with a lower non-dominated level; if the levels are the same, then select the individual with a greater crowding degree to ensure that excellent and evenly distributed individuals are more likely to be selected for reproduction.

[0193] Crossover: Commonly used simulated binary crossover (SBX), generate the positions of offspring through random numbers, making the offspring tend to be distributed around the parents, and the distribution index controls the concentration degree of the offspring. In the embodiments of the present invention, when implementing the crossover step, explore and calculate the similarity of each pair of parent chromosomes. Driven by the path similarity, for parent individuals x a and x b the larger the inner product of the path matrices A x indicates that they are more similar in terms of the main road distance and the accessibility of the storage location. With a high probability (p c →1), perform path optimization-oriented crossover (PathCopy) to retain high-quality path genes; on the other hand, maintain the diversity of the population. When the similarity is low (p c →0), use uniform crossover to introduce randomness to avoid premature convergence and cover the exploration requirements for the global optimal storage location selection in the patent.

[0194] The specific process of the crossover operation in the embodiments of the present invention is as follows: first perform similarity calculation, for each pair of parent chromosomes, extract their path matrices and , and calculate the cosine similarity. Calculate the probability mapping method, and convert the similarity into the crossover probability p c through the Sigmoid function . The parameter η s controls the slope to adapt to different optimization stages (η s is smaller in the early stage to enhance exploration and increases in the later stage to accelerate convergence), ηs ∈ [2, 10].

[0195] Furthermore, the parameter η s has the following value distribution:

[0196] Low-sensitivity mode (η s = 2 - 4): Suitable for complex path spaces (multi-modal, multi-constrained), requires strong exploration ability.

[0197] Balanced mode (η s = 5 - 7): Balances convergence and diversity, the general recommended value.

[0198] High-sensitivity mode (η s = 8 - 10): Suitable for simple path spaces or accelerating convergence in the later stage of the algorithm.

[0199] For the selection of the crossover operator, PathCopy crossover or uniform crossover is dynamically selected according to the p c value. For example, when p c > 0.7, the path features of the parent generation are preferentially retained. Formula implementation:

[0200]

[0201] Compared with the standard NSGA-II algorithm, the convergence speed of the embodiment of the present invention is effectively improved: The similarity-driven directional crossover enables the algorithm to quickly approach the Pareto front, the number of convergence generations is greatly reduced, and the convergence speed is doubled. Soft constraint satisfaction rate: PathCopy crossover inherits the high-scoring path pattern, reduces soft constraint conflicts, and greatly improves the accuracy.

[0202] Mutation: Random perturbations are introduced through polynomial mutation, and small adjustments are made to some dimensions of the solution to avoid falling into local optima. For example, the variables are slightly increased or decreased with a certain probability while ensuring that the variable range is not exceeded. In order to reduce errors occurring during the crossover process of the offspring, in the embodiment of the present invention, two parameters crucial for hard constraints are verified. First, the capacity and load-bearing are verified. When generating Mask path s, the current load ∑V i of the offspring storage location k b and W k are simultaneously checked. If it exceeds the limit, it is forced to be set to 0 to ensure compliance with the C H constraint set. To ensure the stacking order, a repair strategy is triggered after crossover. If the goods outbound time T i < T j but is stacked in the lower layer, the positions are automatically exchanged to meet the C 顺序 .

[0203] Child = Mask path ⊙ x a+(1 - Mask path )⊙x b

[0204]

[0205] It is equivalent to setting up a layer of insurance, reducing the error probability of offspring crossover, resulting in an increase in errors, without increasing a large algorithm overhead, reducing invalid solutions, performing built-in hard constraint checks, and reducing the invalid solution generation rate.

[0206] (5) Merge the parent generation and the offspring generation:

[0207] Merge the parent population and the offspring population into a temporary population of size 2N, and then through non-dominated sorting and crowding degree screening, retain N optimal and evenly distributed individuals as the new generation population.

[0208] (6) Screen the new generation population

[0209] Select individuals from low to high non-dominated levels to fill the new population. If a certain front exceeds N after being added, then select some individuals in this front in descending order of crowding degree to ensure the population size and diversity.

[0210] (7) Termination condition: Repeat the iteration (selection, crossover, mutation, merge, screening) until the maximum number of iterations is reached or the convergence condition is satisfied (such as the Pareto front is stable), and finally output the optimal solution set.

[0211] S6. Output the result of storage location allocation

[0212] According to the optimization result, allocate the goods efficiently and reasonably to the appropriate storage locations, thereby improving the port scheduling efficiency and resource utilization rate.

[0213] The embodiment of the present invention also provides an intelligent storage location allocation system for port bulk cargo, including:

[0214] A data collection and storage location allocation requirement analysis module, which is used to collect historical data related to port cargo scheduling, including cargo type, volume, weight, inbound and outbound time, and clarify the storage location allocation requirements;

[0215] A constraint condition design module, which is used to design constraint conditions for the storage location allocation process, and the constraint conditions include hard constraints and soft constraint conditions;

[0216] A multi-objective optimization model construction module, which is used to construct a multi-objective optimization model based on the genetic algorithm according to the constraint conditions;

[0217] An initial feasible solution generation module, which is used to generate initial feasible solutions that meet all hard constraints by using a greedy pre-screening algorithm; and

[0218] The storage location allocation scheme optimization module is used to solve the multi-objective optimization model by using the improved NSGA-II algorithm to obtain the optimal storage location allocation scheme.

[0219] Combined with Figure 2 It can be seen that the improved NSGA-II algorithm of the present invention has a 25.6% increase in the convergence speed in terms of the hypervolume index. When iterating to the 40th generation, the improved algorithm has reached the convergence level of the standard algorithm at the 60th generation. The output curve shows that the improved algorithm has a steeper convergence slope in the early iteration stage (1 - 30 generations), indicating that the adaptive crossover probability mechanism effectively accelerates the optimization process.

[0220] Combined with Figure 3 It can be seen that the solution set of the improved algorithm forms a better distribution in three dimensions: storage cost, loading balance, and throughput efficiency, and the coverage range of the solution set expands by 42%. There is an obvious aggregation phenomenon in the solution set of the standard algorithm (variance 0.1), while the variance of the solution set of the improved algorithm drops to 0.05, verifying the effectiveness of the dynamic crowding distance operator.

[0221] Combined with Figure 4 It can be seen that the average satisfaction rate of the improved algorithm for the five core constraints (capacity limit, stacking order, cargo category isolation, load limit, and adjacency constraint) is increased to 95.2%, and the satisfaction rate of the adjacency constraint is increased by 41.5%. The 65% satisfaction rate of the standard algorithm for the Proximity constraint is increased to 92% after improvement, verifying the effectiveness of the priority-based constraint handling mechanism.

[0222] Combined with Figure 5 It can be seen that the median running time of the improved algorithm for a single run is 82 seconds (120 seconds for the standard algorithm), and the computing efficiency is increased by 31.7%. It shows that the efficiency improvement is statistically significant, verifying the effectiveness of the population partitioning strategy based on GPU acceleration.

[0223] Combined with Figure 6 It can be seen that the standard deviation of the density distribution of the solution set of the improved algorithm is 0.15 (0.23 for the standard algorithm), and the coverage rate of the solution space is increased by 39%. The improved algorithm forms a more uniform grid-like distribution in the objective space (grid density 0.8), while the standard algorithm shows a central aggregation feature, verifying the effectiveness of the diversity preservation strategy based on information entropy.

[0224] In summary, the present invention comprehensively considers the port yard layout, cargo characteristics, and outbound rules, optimizes the storage location of goods through reasonable storage location allocation, so as to reduce the number of inventory reorganizations and detention time as much as possible. Through this innovative method, the port can manage the storage location allocation more efficiently, achieve dynamic optimization of cargo scheduling, improve resource allocation efficiency, reduce logistics costs, and enhance the overall business benefits.

Claims

1. An intelligent allocation method for bulk cargo storage positions in a port, characterized in that Including: Collect historical data related to port cargo scheduling, including cargo type, volume, weight, inbound and outbound times, and clarify the requirements for storage location allocation; Design constraint conditions for the storage location allocation process, where the constraint conditions include hard constraints and soft constraint conditions; Based on the constraint conditions, construct a multi-objective optimization model using a genetic algorithm; Combined with historical data related to port cargo scheduling, use a greedy pre-screening algorithm to generate an initial feasible solution that satisfies all hard constraints; Use the above initial feasible solution as the input of the multi-objective optimization model, and use an improved NSGA-II algorithm to solve the optimal solution of storage location allocation; According to the optimization results, allocate the cargo to the optimal storage location.

2. The intelligent allocation method for port bulk cargo storage locations according to claim 1, wherein The hard constraints include capacity, stacking shape, stacking order, same-type constraint, and load-bearing constraint, where: Capacity constraint: The volume of the goods allocated in storage location k does not exceed its maximum capacity C k , that is Stacking order constraint: The outbound time of the upper-layer cargo is greater than or equal to the outbound time of the lower-layer cargo; Same-type cargo constraint: Only the specified cargo types can be stacked in storage location k; Warehouse location load-bearing constraint: the total weight W after warehouse location scheduling k shall not exceed the maximum load-bearing limit W max,k ; Stacking shape constraint: Ensure that the stacking shape of the cargo conforms to physics to prevent affecting the stability of the cargo.

3. The intelligent allocation method of port bulk cargo storage locations according to claim 1, wherein The soft constraints include proximity constraint, outbound time concentration constraint, storage location load balance constraint, stacking height recommendation constraint, transportation path optimization constraint, and dynamic adjustment ability constraint, where: The objective function of the proximity constraint is: minimize ∑ {i,j∈B} (d(k i ,k j ) × S {i,j} ), where d(k i ,k j ) is the physical distance between storage locations i and j; S {i,j} is the degree of goods association; The objective function for the concentration constraint of the outbound time is: minimize ∑ {k∈K} σ t (T b | b ∈ k), σ t is the time standard deviation, and T b is the outbound timestamp of the goods b; The objective function of the storage location load balance constraint is as follows: W k is the total weight of the goods in storage location k; W max,k represents the maximum load-bearing limit of storage location k; The objective function of the stacking height recommended constraint is as follows: h k is the actual stacking height of storage location k, and H recommend,k is the recommended safe stacking height of storage location k; The objective function for optimizing the transportation route constraints is: minimize ∑ {b∈B} (f b ×d(k b , G)), where f b is the expected incoming and outgoing frequency of the goods, and d(k b , G) is the physical distance between the storage location k b where the goods b is located and the main road location G; The objective function for dynamically adjusting the capacity constraint is as follows: V b is the volume of goods b, and C k is the maximum capacity of storage location k, and P k is the reserved value coefficient of storage location k.

4. The intelligent allocation method for port bulk cargo storage locations according to any one of claims 1-3, characterized in that, Based on the constraint conditions, construct a multi-objective optimization model using a genetic algorithm. The total objective function of this multi-objective optimization model is: In the above formula: C in,b,k is the inbound cost for goods b to enter storage location k, w in is the weight of the inbound cost; C st,b,k is the storage cost of goods b stored in storage location k, w st is the weight of the storage location storage cost; Indicates the utilization rate of the storage location, where V k is the actually occupied volume in storage location k, and C k is the maximum capacity of storage location k, and w u is the weight of the utilization rate of the storage location; Indicates the storage location load balance, where W k is the total weight of the goods in storage location k; W max,k represents the maximum load-bearing limit of storage location k, |K| is the total number of storage locations, and w l is the weight of the storage location load balance.

5. The intelligent allocation method for port bulk cargo storage locations according to claim 4, characterized in that The use of the greedy pre-screening algorithm to generate an initial feasible solution that satisfies all hard constraints includes: (1) Design of the dynamic priority formula Combined with spatio-temporal coupling attenuation, dynamic load-bearing risk, and spatial fitness, construct the following dynamic priority function: Priority(b) = α·E b + β·B l + γ·U l Among them, E b , B l , U l are the spatio-temporal coupling attenuation, dynamic load-bearing risk, and spatial adaptability respectively; α, β, and γ are the corresponding dynamic weight coefficients; Space-time coupled attenuation E b The calculation is as follows: In the above formula, λ is the time decay coefficient, which determines the decay rate of the goods outbound time on the overall fitness E b , T b is the outbound timestamp of goods b; d b is the physical distance from the goods to the storage location, d0 is the distance threshold, and γ d is the distance sensitivity coefficient, which determines the sensitivity of the distance d b from the goods to the storage location to the spatial fitness; Dynamic load-bearing risk of storage location B l The calculation is as follows: In the above formula, σ risk (l) is the load-bearing risk coefficient of storage location l, and σ threshold is the load-bearing safety threshold; The spatial fitness is calculated as follows: U l is the space utilization rate, V b is the volume of the goods b to be allocated currently, V i is the volume of the existing goods i in the storage location l, C l is the maximum volume of the storage location l, Atop is the available area at the top layer of the current storage location, A l is the area of the reference layer of the storage location; (2) Traverse the storage locations, screen the feasible storage locations that satisfy all hard constraints to obtain the initial feasible solution; calculate the dynamic priorities of all feasible storage locations, sort them in descending order of weights, and dynamically update the storage location status.

6. The intelligent allocation method for port bulk cargo storage locations according to claim 5, wherein The dynamic weights α, β, γ are adjusted according to the storage location status: and α + β + γ = 1.

7. The intelligent allocation method of port bulk cargo storage location according to claim 5, characterized in that, The improved NSGA-II algorithm uses an adaptive crossover probability, and the crossover probability calculation formula is as follows: Among them, η represents the cross-sensitivity coefficient; and are respectively the path matrices of the parental individuals x a and x b ; represents and the cosine similarity between them.

8. The intelligent allocation method for port bulk cargo storage locations according to claim 7, characterized in that, When p c → 1, perform the PathCopy crossover operation. When p c → 0, perform the uniform crossover operation.

9. The intelligent allocation method for port bulk cargo storage locations according to claim 8, characterized in that, When the improved NSGA-II algorithm performs the mutation operation, it first verifies the capacity and load-bearing in the hard constraints. When generating Mask path it synchronously checks the current load ∑V i of the offspring storage location k b and W k . If it exceeds the limit, it is forced to be set to 0 to ensure that the C H constraint set is satisfied; If the goods outbound time is T i <T j , but they are stacked at the lower layer, then after intersection, the patching strategy is triggered to automatically swap positions to meet C 顺序 : Child=Mask path ⊙x a +(1 - Mask path )⊙x b 10. An intelligent allocation system for port bulk cargo storage locations, characterized in that, Including: Data acquisition and storage location allocation requirement analysis module, used to collect historical data related to port cargo scheduling, including cargo type, volume, weight, inbound and outbound times, and clarify the requirements for storage location allocation; Constraint condition design module, used to design constraint conditions for the storage location allocation process, where the constraint conditions include hard constraints and soft constraint conditions; Multi-objective optimization model construction module, used to construct a multi-objective optimization model based on the genetic algorithm according to the constraint conditions; Initial feasible solution generation module, used to generate an initial feasible solution that satisfies all hard constraints using a greedy pre-screening algorithm; And Storage location allocation scheme optimization module, which is used to solve the multi-objective optimization model using an improved NSGA-II algorithm to obtain the optimal storage location allocation scheme.

Images