FCM-PSO-FWA algorithm-based site selection method for multiple vegetable bird stations
Through the FCM-PSO-FWA algorithm, combined with fuzzy C-mean clustering and fireworks-particle clustering algorithm, the location of Cainiao Station was optimized, and the problem of site selection of multiple Cainiao Stations was solved, and effective site selection in the community was achieved, which was better than The prior art has improved site selection accuracy and efficiency.
Patent Information
- Application Number
- CN202411076184.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-07
AI Technical Summary
The existing technology mainly focuses on the location selection of a single logistics distribution center in the selection of a single logistics distribution center, and lacks effective solutions for multiple Cainiao stations, especially in the location selection of multiple Cainiao stations in the community. The existing research has been carried out. Imature.
Using multiple Cainiao Station site selection methods based on FCM-PSO-FWA algorithm, the 0-1 integer planning model is constructed through the fuzzy C-mean clustering algorithm and the fireworks-particle swarm algorithm, and the location of Cainiao Station is optimized to minimize the total number of collectors. The distance.
It effectively solves the site selection problem of multiple Cainiao Stations, which is better than using the particle swarm algorithm alone, and can find the optimal solution under the conditions of any required points, improves the site selection accuracy and efficiency, and verifies that it has the ability to use it in the practical application of Jiangnan University. Feasibility and effectiveness.
Smart Images

Figure CN119090391A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the application fields of linear programming, statistics, geographic information processing, network analysis, etc. More specifically, it relates to a method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm. Background Art
[0002] With the popularity of e-commerce, consumers' demand for express delivery services has also grown rapidly. As a new express delivery service model, the location of Cainiao Post Station is crucial to improving the efficiency of express delivery services and user experience. Therefore, the location of Cainiao Post Station needs to comprehensively consider consumers' pickup needs and express delivery service efficiency, so as to meet the growing market demand.
[0003] Foreign scholars have studied the location selection problem very early. Awasthi et al. used multi-criteria decision making, hierarchical analysis method and fuzzy TOPSIS method to study the location selection problem of urban logistics distribution centers from four aspects: vehicle size, congestion charging plan, access time limit and distribution center service. Eryuruk et al., experts in the clothing and logistics industries, evaluated the establishment of logistics centers in the Turkish clothing industry. Based on the data of the questionnaire survey database of the Turkish clothing industry, the hierarchical analysis method was used to evaluate the questionnaire survey results. Tomic et al. conducted logistics center location selection for the Balkan Peninsula, using the environmental strategy-performance paradigm as a means to understand the correlation between environmental complexity and logistics centers, and using mathematical methods such as greedy heuristic algorithms and hierarchical analysis methods to solve the model. Ozmen et al. designed a three-stage method logistics center location selection model in the context of Kayseri Logistics Development Plan. The ranking results of the location selection schemes of various logistics centers were obtained. Georgijevic et al. considered the logistics location problem involving the optimal number, size and location of public logistics centers, established a mathematical model combining loading and unloading costs and temporary storage costs of goods in logistics centers, and evaluated the effectiveness of the proposed method using a numerical example of an important public logistics center in the Republic of Serbia. Foreign logistics location research integrates multiple disciplines, including logistics, geography, sociology and economics, to develop a more scientific logistics center location plan.
[0004] Most of the research on logistics site selection by domestic scholars also focuses on the site selection of logistics distribution centers. Yang Ming solved the site selection problem in reverse logistics based on the optimized discrete particle swarm algorithm, which made the algorithm converge faster and more accurate, and improved the operation efficiency of reverse logistics. Zhang Yu et al. constructed a logistics center site selection model and a logistics distribution area division model by improving the clustering algorithm and the minimum supporting tree method, and then simulated and verified the algorithm. In order to solve the problem of blindly pursuing the minimum distribution cost and ignoring the service level, Tang Xifeng et al. proposed a calculation method for the reliability of the distribution center system, established a multi-objective optimization model, and converted the established multi-objective model into a single-objective model through the main objective method, and then used the greedy take algorithm to obtain the optimal logistics site selection plan. Lu Shichang et al. constructed a planning model that minimizes the comprehensive cost of carbon emissions logistics for the decision-making problem of the combination of distribution center site selection and multi-model vehicle path optimization under the background of low-carbon logistics, and then used a two-stage heuristic algorithm to solve it. Huang Zhiwen and others studied the location selection of emergency material distribution centers in wartime under uncertain conditions, established a fuzzy reliability location selection model, and combined immune and genetic algorithms to solve the problem. Peng Hongchun established a model with the goal of minimizing cost, solved the logistics location selection problem by improving the wolf pack algorithm, and finally proved the optimality of the improved algorithm by comparing it with other algorithms. Domestic researchers have found that traditional location analysis methods have great limitations in practical applications, so in recent years, modern technologies such as artificial intelligence have been introduced into the model to improve the location selection effect.
[0005] Most of the current research focuses on the location selection of logistics distribution centers, and the research on the location selection of Cainiao Post Stations is still immature. In addition, the existing literature mainly considers the location selection of a single logistics distribution center, and there are relatively few literatures on the location selection of multiple logistics centers. In real life, communities such as university campuses and large residential areas often need to establish multiple Cainiao Post Stations. Therefore, this paper focuses on the location selection of multiple Cainiao Post Stations in community life, integrates multiple machine learning methods, and solves the location selection problem of multiple Cainiao Post Stations in a community. Summary of the invention
[0006] In view of the above problems existing in the prior art, the present invention proposes a method for site selection of multiple Cainiao stations based on the FCM-PSO-FWA algorithm. The present invention focuses on the site selection problem of multiple Cainiao stations in a community, and establishes a Cainiao station site selection model based on 0-1 integer programming from the perspective of the shortest total pickup distance for all pickups. The FCM-PSO-FWA algorithm that can effectively solve the site selection problem of multiple Cainiao stations is designed, and the algorithm is simulated.
[0007] To achieve the above object, the present invention provides the following technical solutions:
[0008] The location selection method of multiple Cainiao stations based on the FCM-PSO-FWA algorithm includes the following steps: Assume that m Cainiao stations are established in a community, and the location of the i-th Cainiao station is (X i ,Y i ), assuming that there are n demand points in the community, temporarily ignoring the number of people picking up items at each demand point and the cost of building a Cainiao Post Station, and assuming that the total one-way distance for all pick-ups to go to the Cainiao Post Station to pick up items is f, then there is the following 0-1 integer linear programming model:
[0009]
[0010] Where (x max ,y max ) represents the maximum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, (x min ,y min ) represents the minimum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, d ij represents the distance that the recipient needs to travel from the jth demand point to the i-th Cainiao station to pick up the item. The FCM-PSO-FWA algorithm is used to find the global optimal solution of the above model.
[0011] The technical solution is further optimized, and the steps of the FCM-PSO-FWA algorithm are as follows:
[0012] (1) Initialize the parameters of the fuzzy C-means clustering algorithm and calculate the cluster center;
[0013] (2) Update the membership matrix and iterate, output the final cluster center and the membership matrix of each demand point to each cluster, so as to obtain the classification results of each demand point. For each type of demand point, the location of the Cainiao station for each type of demand point is solved by the Fireworks-Particle Swarm Algorithm.
[0014] The technical solution is further optimized. Step (1) includes selecting the locations of m Cainiao stations, that is, first using the fuzzy C-means clustering algorithm to cluster n demand points into m categories, with the number of clusters being m and the fuzzification parameter being h, and randomly initializing the membership matrix U = [u ij ], where u ij Indicates the degree to which the jth demand point belongs to the i-th cluster, and for each demand point, the membership degree u ij Must satisfy Then calculate the center m of each cluster i =(m ix ,m iy ), the calculation formula of cluster center is as follows:
[0015]
[0016] Where (x j,y j ) represents the coordinate position of the jth demand point.
[0017] In a further optimization of the technical solution, step (2) includes updating the membership degree u of each demand point to each cluster ij , calculated as follows:
[0018]
[0019] where d ij represents the Euclidean distance from the jth demand point to the ith Cainiao station. Repeat formulas (2), (3) and (4) to update the cluster center position and the membership matrix until the change of the membership matrix U is less than a predetermined threshold or the maximum number of iterations is reached. Output the final cluster center and the membership matrix of each demand point to each cluster, thereby obtaining the classification results of each demand point. For each type of demand point, the fireworks-particle swarm algorithm is used to solve the Cainiao station position of each type of demand point. The classification result of the demand point corresponds to K in formula (1). ij The value of K, that is, if demand point j belongs to the i-th category, then ij =1, otherwise K ij =0.
[0020] The technical solution is further optimized, and the fireworks-particle swarm algorithm includes the following steps:
[0021] (3) Initialize the particle swarm and determine the initial position of the particles, that is, the initial position of the Cainiao Station (X i ,Y i ))、initial speed (V xi ,V yi ), inertia factor w, the boundary range of the particle's horizontal and vertical coordinates (x max ,y max ) and (x min ,y min ) and particle search speed range [V min ,V max ] parameter value, for the inertia factor w, generally in the early stage of the algorithm, a larger inertia weight is used to enhance the global search capability of the algorithm, and a smaller inertia weight is used to improve the local search capability of the algorithm in the later stage of the algorithm. Therefore, this embodiment adopts a time-varying inertia weight linear decreasing strategy to update the inertia factor value, and its formula is:
[0022] w(t)=(w max -w min )(1-t / T)+w min (5)
[0023] where w max is the maximum value of random weight, w minis the minimum value of random weight, t is the current iteration number, and T is the maximum number of algorithm iterations;
[0024] (4) Calculate the fitness of each particle. The fitness function is the objective function f in formula (1). According to the fitness value f of each particle i Update the local and global optimal solutions and optimal values. For each particle, compare its current fitness value with its known best fitness value. If the current fitness value is better, that is, the total pickup distance is shorter, then update the personal optimal solution of the particle and its corresponding position, that is, update the shortest total pickup distance and the Cainiao Station location. Then find the particle with the best fitness value among all the personal optimal solutions of the particle swarm, and compare this fitness value with the current global optimal solution. If it is better than the current global optimal solution, then update the global optimal solution and global optimal value.
[0025] (5) Treat each particle as a firework, and use the fireworks algorithm to search for the local optimal solution of each particle using the sparks produced by the explosion within the explosion radius. Fireworks x i The explosion radius A i and the number of explosion sparks S i The calculation formulas are as follows:
[0026]
[0027] where y min =min(f(x i )),y max =max(f(x i )), i = 1, 2…N represent the minimum and maximum values of all spark fitness respectively, R and M are constants used to adjust the explosion radius and the number of explosion sparks respectively, ε is the minimum amount of the machine set to prevent the denominator of the formula from being zero, in order to prevent particles with good fitness from generating too many explosion sparks and particles with poor fitness from generating too few sparks, so the number of sparks generated by particle i S i The following constraints exist:
[0028]
[0029] Where a and b are constants, round() represents the rounding function, and then randomly select some sparks from the explosion sparks generated by each particle to generate variant sparks through Gaussian mutation, fireworks x i The variation formula in the kth dimension is shown in formula (9). If the sparks generated by the explosion or the variation sparks exceed the feasible region, the spark position is adjusted according to formula (10).
[0030] x ik =x ik ×e (9)
[0031] x ik =x LB,k +|x ik |%(x UB,k -x LB,k ) (10)
[0032] Where e~N(1,1), x UB,k 、x LB,k They are the upper and lower boundaries of the spark in the kth dimension;
[0033] (6) Compare the fitness value of each particle with the corresponding spark, update the local and global optimal solutions and optimal values, calculate the fitness value of the spark and the mutated spark generated by each particle, and if there is a spark whose fitness value is better than the particle fitness value, use the fitness value of the spark to update the local optimal value of the particle, and use the position of the spark to update the local optimal solution of the particle. Similarly, update the local optimal solutions of all remaining particles, and update the global optimal solution after updating all local optimal solutions;
[0034] (7) Update the speed and position of each particle according to the local and global optimal solutions, and according to the boundary range of the horizontal and vertical coordinates of the Cainiao Station location in formula (1) (x max ,y max ) and (x min ,y min ) and the set particle speed range are used to adjust particles whose speed or position exceeds the boundary conditions. The particle speed and coordinate update formulas are shown in formulas (11) and (12). The adjustment method for speed or position exceeding the boundary is shown in formulas (13), (14) and (15):
[0035]
[0036] V ki =rd×(V max -V min )+V min (13)
[0037] X i =rd×(x max -x min )+x min (14)
[0038] Y i =rd×(y max -y min )+y min (15)
[0039] In the formula represents the velocity of the i-th particle in the k-th dimension at the t-th iteration, p pk and pgk are the local and global optimal solutions in the kth dimension, c1 and c2 are the self-learning factor and social learning factor, r1 and r2 are random numbers in [0,1], α is the constraint factor used to control the weight, rd represents a random number in the range of [0,1), V ki represents the velocity of the i-th particle in the k-th dimension;
[0040] (8) Determine whether the result meets the termination condition. If so, output the optimal solution and end the algorithm. If not, proceed to the next step.
[0041] (9) Return to step (4) and continue iterating until the optimal location (X, Y) of this type of Cainiao Station is generated. Similarly, the location selection method for other types of Cainiao Stations is also as shown in steps (3)-(9).
[0042] Different from the prior art, the above technical solution has the following beneficial effects:
[0043] (1) Since the particle swarm algorithm can only solve a global optimal solution to the problem, there is a limitation on the number of optimal solutions. The particle swarm algorithm alone cannot solve the location selection problem of multiple Cainiao stations. FCM can divide the location selection problem of multiple Cainiao stations into multiple single Cainiao stations, thus making machine learning methods similar to the particle swarm algorithm applicable to the multi-center location selection problem.
[0044] (2) Under the condition of any number of demand points, directly using the cluster center of FCM as the location of Cainiao Post Station is not the optimal solution. Based on the FCM clustering results, PSO and FWA are combined to solve the optimal location of Cainiao Post Station for each type of demand point. At this time, the total distance for all recipients to go to the corresponding Cainiao Post Station to pick up the parcel is the shortest, that is, the location selection result of FCM-PSO-FWA is better than the location selection result of FCM.
[0045] (3) According to the actual data of all demand points of Jiangnan University, the site selection model and algorithm proposed in this paper are applied to the site selection of Cainiao Station of Jiangnan University. It is found that the two Cainiao Stations of Jiangnan University should be located in the North District Student Activity Center and the South District Runyuan Student Dormitory, which is consistent with the actual site selection of Cainiao Station of Jiangnan University, further verifying the feasibility and effectiveness of FCM-PSO-FWA algorithm. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 It is the overall flow chart of FCM-PSO-FWA algorithm;
[0047] Figure 2 This is a schematic diagram of the solution results of FCM for two Cainiao stations with different scale demand points;
[0048] Figure 3 This is a schematic diagram of the location of the Cainiao Station obtained by using FCM-PSO-FWA;
[0049] Figure 4 This is a map of Jiangnan University;
[0050] Figure 5 This is a schematic diagram of the clustering results of Jiangnan University;
[0051] Figure 6 This is a schematic diagram of the site selection results of the Cainiao Station, the first type of demand point within Jiangnan University;
[0052] Figure 7 Schematic diagram of the site selection results of the Cainiao Station, the second type of demand point of Jiangnan University;
[0053] Figure 8 This is the actual location map of the Cainiao Station at Jiangnan University; DETAILED DESCRIPTION
[0054] In order to explain the technical content, structural features, achieved objectives and effects of the technical solution in detail, the following is a detailed description in conjunction with specific embodiments and accompanying drawings.
[0055] The present invention mainly aims at the site selection problem of multiple Cainiao stations in communities such as newly built schools and newly built communities. For newly built communities or campuses, the location where the recipient sets out to pick up the items (hereinafter referred to as the demand point) is known, but the road has not been planned. Therefore, the Euclidean distance from the demand point to the Cainiao station is used as the pickup distance between the two points during calculation. Assuming that the recipient goes to the Cainiao station to pick up the items and then returns along the original route, the goal is to minimize the total one-way distance of all recipients to the Cainiao station, which is equivalent to minimizing the total round-trip distance of all recipients to the Cainiao station. Then use FCM to classify the demand points according to the number of Cainiao stations to be established as needed, and then use PSO-FWA to search for the location of the Cainiao station for each type of demand point. PSO-FWA performs a small-scale explosion on each particle based on the PSO particle's own update iteration, so that each particle will generate random sparks each iteration, which improves the search efficiency and precision, and effectively improves the problem that PSO is prone to fall into local optimal solutions. The specific steps of site selection are as follows.
[0056] Assume that m Cainiao Post Stations are established in a community, and the location of the i-th Cainiao Post Station is (X i ,Y i ), assuming that there are n demand points in the community, and temporarily ignoring the number of people picking up parcels at each demand point and the cost of building a Cainiao Post Station. Let the total one-way distance for all parcel pickers to go to the Cainiao Post Station to pick up parcels be f, then there is the following 0-1 integer linear programming model:
[0057]
[0058] Where (x max ,y max ) represents the maximum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, (x min ,y min ) represents the minimum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, (x j ,y j ) represents the coordinate position of the jth demand point, d ij Represents the distance the recipient travels from the jth demand point to the i-th Cainiao Station to pick up the package.
[0059] The present invention mainly studies the situation of establishing multiple Cainiao stations in a community, and uses particle swarm optimization (PSO) to solve the optimal location of the Cainiao station. However, PSO can only solve the optimal location for establishing a Cainiao station through iteration, and cannot solve the site selection problem of multiple Cainiao stations. In response to this problem, the present invention first uses the fuzzy C-means (FCM) clustering method to cluster all demand points according to their geographical location coordinates, and obtains the cluster center of each type of demand point. Each cluster center is regarded as the initial site selection position of the Cainiao station of each type of demand point, so that the site selection problem of multiple Cainiao stations in a community is divided into the site selection problem of a single Cainiao station in multiple sub-communities.
[0060] However, using the cluster center location as the Cainiao Station location is not the optimal solution. In the present invention, FCM is only a tool to divide multiple Cainiao Station location problems into multiple single Cainiao Station location problems. After that, PSO needs to be used to iteratively search for the optimal Cainiao Station location for each type of demand point. Although PSO has the advantages of fast convergence speed, few parameters, and simple and easy-to-implement algorithm, it is also more dependent on good initialization. Otherwise, it is very easy to fall into the local optimal solution and cannot find the global optimal solution. The Fireworks Algorithm (FWA) is an algorithm that finds the optimal solution by simulating the fireworks explosion process. Fireworks are regarded as a feasible solution in the solution space of the optimization problem. Then the process of fireworks exploding to produce a certain number of sparks is the process of searching the neighborhood. FWA has the characteristics of randomness, explosiveness, and diversity. Therefore, integrating FWA into PSO can improve the phenomenon that PSO is prone to "premature maturity" (falling into the local optimal solution) and the problem of slow search speed. Therefore, the fireworks-particle swarm algorithm (PSO-FWA) is designed to find the global optimal solution. The overall process of the FCM-PSO-FWA algorithm is as follows Figure 1 As shown, the implementation steps of the FCM-PSO-FWA algorithm are as follows:
[0061] (1) Initialize the FCM algorithm parameters and calculate the cluster center. To select the location of m Cainiao stations, first use the FCM algorithm to cluster n demand points into m categories, with the number of clusters being m and the fuzzification parameter being h. Randomly initialize the membership matrix U = [u ij ], where u ij Indicates the degree to which the jth demand point belongs to the i-th cluster, and for each demand point, the membership degree u ij Must satisfy Then calculate the center m of each cluster i =(m ix ,m iy ), the calculation formula of cluster center is as follows:
[0062]
[0063] Where (x j ,y j ) represents the coordinate position of the jth demand point.
[0064] (2) Update the membership matrix and iterate. Update the membership u of each demand point to each cluster ij , calculated as follows:
[0065]
[0066] where d ij Represents the Euclidean distance from the jth demand point to the ith Cainiao station. Repeat formulas (2), (3) and (4) to update the cluster center position and membership matrix until the change in the membership matrix U is less than a predetermined threshold or the maximum number of iterations is reached. Output the final cluster center and the membership matrix of each demand point to each cluster, thereby obtaining the classification results of each demand point. For each type of demand point, the PSO-FWA algorithm is used to solve the Cainiao station location of each type of demand point. The classification result of the demand point corresponds to K in formula (1). ij The value of K, that is, if demand point j belongs to the i-th category, then ij =1, otherwise K ij =0.
[0067] (3) Initialize the particle swarm and determine the initial position of the particle (i.e. the initial position of the Cainiao Station (X i ,Y i ))、initial speed (V xi ,V yi ), inertia factor w, the boundary range of the particle's horizontal and vertical coordinates (x max ,y max ) and (x min ,y min ) and particle search speed range [V min ,Vmax ] and other parameter values. For the inertia factor w, generally, a larger inertia weight is used in the early stage of the algorithm to enhance the global search capability of the algorithm, and a smaller inertia weight is used in the later stage of the algorithm to improve the local search capability of the algorithm. Therefore, this embodiment adopts a time-varying inertia weight linear decreasing strategy to update the inertia factor value, and its formula is:
[0068] w(t)=(w max -w min )(1-t / T)+w min (5)
[0069] where w max is the maximum value of random weight, w min is the minimum value of random weight, t is the current iteration number, and T is the maximum number of algorithm iterations.
[0070] (4) Calculate the fitness of each particle. The fitness function is the objective function f in formula (1). According to the fitness value f of each particle i Update local and global optimal solutions and optimal values. For each particle, compare its current fitness value with its known best fitness value. If the current fitness value is better, that is, the total pickup distance is shorter, then update the personal optimal solution of the particle and its corresponding position, that is, update the shortest total pickup distance and Cainiao Station location. Then find the particle with the best fitness value among all personal optimal solutions of the particle swarm, compare this fitness value with the current global optimal solution, and if it is better than the current global optimal solution, update the global optimal solution and global optimal value.
[0071] (5) Treat each particle as a firework, and use the fireworks algorithm to search for the local optimal solution of each particle using the sparks produced by the explosion within the explosion radius. Fireworks x i The explosion radius A i and the number of explosion sparks S i The calculation formulas are as follows:
[0072]
[0073] where y min =min(f(x i )),y max =max(f(x i )), i = 1, 2…N represent the minimum and maximum values of all spark fitness respectively, R and M are constants used to adjust the explosion radius and the number of explosion sparks respectively, and ε is the minimum value of the machine set to prevent the denominator of the formula from being zero. In order to prevent particles with good fitness from generating too many explosion sparks and particles with poor fitness from generating too few sparks, the number of sparks generated by particle i S i The following constraints exist:
[0074]
[0075] Among them, a and b are constants, and round() represents the rounding function. Then, among the explosion sparks generated by each particle, some sparks are randomly selected to generate variant sparks through Gaussian mutation. Fireworks x i The variation formula in the kth dimension is shown in formula (9). If the sparks generated by the explosion or the variation sparks exceed the feasible domain, the spark position is adjusted according to formula (10).
[0076] x ik =x ik ×e (9)
[0077] x ik =x LB,k +|x ik |%(x UB,k -x LB,k ) (10)
[0078] Where e~N(1,1), x UB,k 、x LB,k are the upper and lower boundaries of the spark in the kth dimension respectively.
[0079] (6) Compare the fitness values of each particle and the corresponding spark, and update the local and global optimal solutions and optimal values. Calculate the fitness values of the sparks and mutated sparks generated by each particle. If there is a spark whose fitness value is better than the particle fitness value, then use the fitness value of the spark to update the local optimal value of the particle, and use the position of the spark to update the local optimal solution of the particle. Similarly, update the local optimal solutions of all remaining particles, and update the global optimal solution after updating all local optimal solutions.
[0080] (7) Update the speed and position of each particle according to the local and global optimal solutions, and according to the boundary range of the horizontal and vertical coordinates of the Cainiao Station location in formula (1) (x max ,y max ) and (x min ,y min ) and the set particle speed range are used to adjust particles whose speed or position exceeds the boundary conditions. The particle speed and coordinate update formulas are shown in formulas (11) and (12), and the adjustment method for speed or position exceeding the boundary is shown in formulas (13), (14) and (15):
[0081]
[0082]
[0083] V ki =rd×(V max -Vmin )+V min (13)
[0084] X i =rd×(x max -x min )+x min (14)
[0085] Y i =rd×(y max -y min )+y min (15)
[0086] In the formula represents the velocity of the i-th particle in the k-th dimension at the t-th iteration, p pk and p gk are the local and global optimal solutions in the kth dimension, c1 and c2 are the self-learning factor and social learning factor, r1 and r2 are random numbers in [0,1], α is the constraint factor used to control the weight, rd represents a random number in the range of [0,1), V ki Represents the velocity of the i-th particle in the k-th dimension.
[0087] (8) Determine whether the result meets the termination condition. If so, output the optimal solution and end the algorithm. If not, proceed to the next step.
[0088] (9) Return to step (4) and continue iterating until the optimal location (X, Y) of this type of Cainiao Station is generated. Similarly, the location selection method for other types of Cainiao Stations is also as shown in steps (3)-(9).
[0089] Firstly, the results of FCM and FCM-PSO-FWA in solving multiple Cainiao Station location selection problems are compared to verify the feasibility and effectiveness of the algorithm proposed in this paper. Secondly, the proposed algorithm is applied to the site selection of Cainiao Station in Jiangnan University, and the site selection results are analyzed.
[0090] Algorithm simulation
[0091] The following takes two Cainiao Station location selection problems as examples, and uses FCM and FCM-PSO-FWA to solve the location of the Cainiao Station respectively. Similarly, the location selection problems of more than two Cainiao Stations can also be solved by the same method. Some articles in existing research use FCM to select locations for multiple logistics distribution centers. Although it is a feasible method to use the cluster center obtained by FCM as the location selection result of the logistics distribution center, this method has limitations. The clustering results are sensitive to the selection of the initial cluster center and are prone to fall into the local optimal solution. Therefore, different initializations may lead to different clustering results. The FCM-PSO-FWA proposed in the present invention uses an intelligent algorithm to further search for the location of the Cainiao Station based on the FCM clustering results. After multiple iterations, a Cainiao Station location selection result that is better than the FCM location selection result is obtained, thereby avoiding over-reliance on initialization parameter settings and falling into the local optimal solution.
[0092] 1FCM site selection results
[0093] For the location selection problem of two Cainiao Post Stations, FCM can be used to directly cluster all demand points into two categories according to their geographical locations, and the cluster center of each category is used as the location selection result of the Cainiao Post Station for this category of demand points. Each category of Cainiao Post Station stores the express items of all demand points of this category. The following randomly generates the locations of 20, 40, 60, and 80 demand points in a square area of 100×100 meters, and uses FCM to select two Cainiao Post Stations for demand points of different sizes. The solution results are as follows: Figure 2 shown. Figure 2 (a) is the FCM site selection result of 20 demand points. Figure 2 (b) is the FCM site selection result of 40 demand points. Figure 2 (c) is the FCM site selection result of 60 demand points. Figure 2 (d) is the FCM site selection result of 80 demand points. The solid dots in the figure are randomly generated demand point locations. According to the FCM clustering results, different categories of demand points are distinguished by color, where all black dots are the first category of demand points, and all gray dots are the other category of demand points. The two five-pointed stars are the FCM clustering centers of the two Cainiao stations, which are the best locations of the Cainiao stations obtained by FCM. The demand points covered by express delivery in the first category of Cainiao stations are connected by black straight lines, and the demand points covered by express delivery in the second category of Cainiao stations are connected by gray dotted lines. The lines in the figure can be considered for future road construction. Figure 2For example, 20 randomly generated demand points are clustered into two categories using the FCM algorithm. The results show that the 3rd, 6th, 7th, 8th, 11th, 14th, 16th, and 17th demand points are clustered into the first category, and the 1st, 2nd, 4th, 5th, 9th, 10th, 12th, 13th, 15th, 18th, 19th, and 20th demand points are clustered into the second category. The FCM obtains the location of the Cainiao Station for the first type of demand point as (29.5439, 48.8715), and the location of the Cainiao Station for the second type of demand point as (88.8778, 47.9812). At this time, the total distance that the recipients at each demand point travel to the corresponding Cainiao Station to pick up the parcels is 546.4508 meters. Similarly, when the number of demand points is 40, the locations of the two Cainiao stations corresponding to the two types of demand points are (22.3306, 53.1286) and (82.1541, 44.813), respectively. At this time, the total pickup distance is 1149.0322 meters; when there are 60 demand points, the locations of the two Cainiao stations are at coordinates (26.5075, 35.7034) and (75.6512, 56.0436), respectively. The total pickup distance is 1845.7843 meters; when there are 80 demand points, the locations of the Cainiao stations are (28.2803, 32.6445) and (74.8563, 56.9393), respectively. At this time, the total pickup distance is 2405.1499 meters.
[0094] 2FCM-PSO-FWA site selection results
[0095] Next, the method proposed in the present invention is used to divide all demand points into two categories based on the clustering results, and then multiple Cainiao Station location problems are divided into two single Cainiao Station location problems. Then, the improved particle swarm algorithm is used to solve the optimal location of the Cainiao Station for each type of demand point. Still taking the locations of 20, 40, 60, and 80 demand points generated in the previous text as an example, the location of the Cainiao Station obtained by FCM-PSO-FWA is as follows: Figure 3 shown.
[0096] Figure 3 (a)-(d) are the location selection results of Cainiao Post Stations corresponding to each type of demand point when 20, 40, 60, and 80 demand points are randomly generated. Figure 3 Taking (a) as an example, when the locations of 20 demand points are randomly generated, the 20 demand points are divided into two categories according to the FCM clustering results. The first category contains 8 demand points, and the second category contains 12 demand points. Figure 3The left picture in (a) shows the result of selecting the location of the Cainiao Station for the first type of demand points based on the clustering results using the improved particle swarm algorithm, and the right picture shows the result of selecting the location of the Cainiao Station for the second type of demand points. At this time, the optimal location of the Cainiao Station for the first type of demand points is at the coordinates (30.68, 55.8184), and the optimal location of the Cainiao Station for the second type of demand points is at the coordinates (92, 47). The sum of the shortest pickup distances is 537.0421 meters. Similarly, when the number of demand points is 40, 60, and 80, the optimal Cainiao Station location for each type of demand point and the corresponding total pickup distance are as follows: Figure 3 (b) To Figure Figure 3 (d) as shown.
[0097] 3 Algorithm Comparison
[0098] The results of using FCM and FCM-PSO-FWA to solve the two Cainiao Station location problems are shown in the following table:
[0099] Table 1 Comparison results of the total pickup distance
[0100]
[0101] Comparing the results of solving the location selection problem of multiple Cainiao stations using FCM and FCM-PSO-FWA in the above table, it can be seen that when the number of demand points gradually increases, the total distance for picking up the goods also increases continuously, and no matter how many demand points there are, the total distance for the recipient to pick up the goods using FCM-PSO-FWA is always shorter than the total distance directly using the cluster center as the location of the Cainiao station. Therefore, the FCM-PSO-FWA proposed in the present invention has the feasibility and effectiveness of solving the location selection problem of multiple Cainiao stations. The above simulation only takes the location selection of two Cainiao stations as an example, and the FCM-PSO-FWA proposed in the present invention is still applicable to the location selection of more than two Cainiao stations.
[0102] Site selection for Jiangnan University Cainiao Station
[0103] 2.1 Problem Analysis
[0104] The e-commerce industry is becoming increasingly popular, and online shopping has brought convenience to people's lives. College students are one of the main forces in the era of online shopping. Taking Jiangnan University as an example, the following study will examine the location selection of its Cainiao Station. The Cainiao Station of Jiangnan University adopts an operating mode in which students go to the Cainiao Station to pick up items. The Cainiao Station does not provide delivery services, so students need to go to the Cainiao Station from various places in the school to pick up items. Although the roads in Jiangnan University have been built, due to human subjective initiative, different students have multiple options for going from the same location to the same Cainiao Station. Therefore, it is impossible to accurately judge the specific distance from the demand point to the Cainiao Station. Therefore, the Euclidean distance between the two points is still used as the actual walking distance to solve, that is, the actual walking distance is ignored. The map of Jiangnan University is as follows Figure 4 shown.
[0105] Since there are relatively few students or faculty members in places such as bus stops, campus gates, sports training grounds, and ornamental gardens on campus, these places are not considered as demand points on campus. The present invention only considers places on campus with a large flow of people as demand points, such as dormitories, teaching buildings, administrative buildings, colleges, libraries, and other venues. That is, only 56 sites with serial numbers 04, 05, 08, 09, 10, 11, 12, 13, 14, 20, 21, 22, 23, 24, 28, 29, 30, 32, 34, 35, 37, 38, 39, 41, 42, 43, 45, 46, 47, 48, 49, 50, 51, 52, 57, 59, 60, 61, 62, 64, 65, 66, 67, 68, 73, 75, 79, 80, 84, 85, 87, 89, 93, 94, 95, 99 in the map are taken as demand points. To facilitate subsequent calculations, the coordinate data of each site is preprocessed. The geographic data of each demand point on Jiangnan University are obtained from Amap. In the preprocessing stage, the real geographic location of each demand point on the map of Jiangnan University is scaled to a two-dimensional coordinate system with a horizontal and vertical coordinate range of [0,100]. That is, the position of each demand point on the map is converted into coordinates in the coordinate system. The coordinates of the converted demand points are shown in Table 2.
[0106] Table 2 Coordinates of demand points of Jiangnan University
[0107]
[0108]
[0109] 2.2.2 Site selection results of Jiangnan University Cainiao Station
[0110] There are 56 demand points in Jiangnan University. Two Cainiao stations are selected on campus. FCM is used to cluster the 56 demand points into two categories. The schematic diagram of the clustering results of Jiangnan University demand points is as follows: Figure 5As shown in the figure, FCM divides 56 demand points into two categories. The first category contains 27 demand points, and the second category contains 29 demand points. In the coordinate axis with the horizontal and vertical coordinate range of [0,100], the coordinates of the two cluster centers are (68.3155,76.7325) and (78.2307,23.1491), respectively. If the cluster center is used as the Cainiao Station, the total pickup distance in the coordinate is 1056.5729. If FCM-PSO-FWA is used for solution, the location selection results of the Cainiao Station for the two types of demand points are shown in the following figure. Figure 6 and Figure 7 The following are schematic diagrams of the site selection results of the Cainiao Station for the first and second types of demand points on the Jiangnan University campus. The gray circle at the lowest point of the three-dimensional surface is the optimal site coordinate of the Cainiao Station. The optimal site coordinates of the Cainiao Station for the first type of demand point are (69.9971, 76.4206), and the coordinates of the Cainiao Station for the second type of demand point are (78.3313, 21.6386). At this time, the total pickup distance in the coordinates is 1053.9224. Compared with the total pickup distance of 1056.5729 with the cluster center as the Cainiao Station, the total pickup distance obtained by FCM-PSO-FWA is shorter, which confirms the effectiveness of the algorithm. The coordinates of the two Cainiao Stations are proportionally restored to the satellite map of Jiangnan University. The locations of the two Cainiao Stations are as follows: Figure 8 Shown is the actual location map of the Cainiao Station at Jiangnan University.
[0111] according to Figure 8 As shown in the figure, the best location for the Cainiao Post Station for the first type of demand point is at the 08 North District Student Activity Center, and the best location for the Cainiao Post Station for the second type of demand point is at the 64 South District Runyuan Student Dormitory. At this time, the total distance for faculty and students to go to the Cainiao Post Station to pick up express delivery is the smallest. The actual locations of the two Cainiao Post Stations of Jiangnan University are currently located at the 44 Express Supermarket and 98 Zhenshan Building, which are next to the Xiaodong Gate of the North District of Jiangnan University and near the South Gate of Jiangnan University. Considering the convenience of express delivery vehicles entering and leaving the campus and not affecting the lives of teachers and students on campus, the actual location of the Cainiao Post Station is relatively remote and close to the campus gate. From the perspective of the pickup distance, it is not convenient for teachers and students to go back and forth to pick up parcels. The location of the Cainiao Post Station obtained by the present invention is similar to the actual location of the Cainiao Post Station, one is located in the north district of the school and the other is located in the south district of the school. Compared with the location of the current Cainiao Post Station, the location of the Cainiao Post Station of the present invention is located near the center of the demand point group in the south and north districts. Both places are locations where teachers and students frequently pass by and have a large flow of people, which makes it more convenient for teachers and students to pick up items, and these two location selection locations will not affect the teaching and daily life of teachers and students. Therefore, the location selection results of the present invention are reasonable and feasible, and also verify the effectiveness of the algorithm, providing a reference for the location selection of the Cainiao Post Station of Jiangnan University.
[0112] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "include..." or "comprise..." do not exclude the existence of other elements in the process, method, article or terminal device including the elements. In addition, in this article, "greater than", "less than", "exceed" and the like are understood to exclude the number itself; "above", "below", "within" and the like are understood to include the number itself.
[0113] Although the above embodiments have been described, once those skilled in the art know the basic creative concepts, they can make additional changes and modifications to these embodiments. Therefore, the above description is only an embodiment of the present invention and does not limit the patent protection scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the specification and drawings of the present invention, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm, characterized in that: The following steps are involved: Assume that m Cainiao Post Stations are established in a community, and the location of the i-th Cainiao Post Station is (X i ,Y i ), assuming that there are n demand points in the community, temporarily ignoring the number of people picking up items at each demand point and the cost of building a Cainiao Post Station, and assuming that the total one-way distance for all pick-ups to go to the Cainiao Post Station to pick up items is f, then there is the following 0-1 integer linear programming model: Where (x max ,y max ) represents the maximum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, (x min ,y min ) represents the minimum boundary value of the horizontal and vertical coordinate range of the Cainiao Station location, d ij represents the distance that the recipient needs to travel from the jth demand point to the i-th Cainiao station to pick up the item. The FCM-PSO-FWA algorithm is used to find the global optimal solution of the above model.
2. The method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm according to claim 1, characterized in that: The steps of the FCM-PSO-FWA algorithm are as follows: (1) Initialize the parameters of the fuzzy C-means clustering algorithm and calculate the cluster center; (2) Update the membership matrix and iterate, output the final cluster center and the membership matrix of each demand point to each cluster, so as to obtain the classification results of each demand point. For each type of demand point, the location of the Cainiao station for each type of demand point is solved by the Fireworks-Particle Swarm Algorithm.
3. The method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm as claimed in claim 2, characterized in that: The step (1) includes selecting the locations of m Cainiao stations, that is, first using the fuzzy C-means clustering algorithm to cluster n demand points into m categories, with the number of clusters being m and the fuzzification parameter being h, and randomly initializing the membership matrix U = [u ij ], where u ij Indicates the degree to which the jth demand point belongs to the i-th cluster, and for each demand point, the membership degree u ij Must satisfy Then calculate the center m of each cluster i =(m ix ,m iy ), the calculation formula of cluster center is as follows: Where (x j ,y j ) represents the coordinate position of the jth demand point.
4. The method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm as claimed in claim 3, characterized in that: The step (2) includes updating the membership degree u of each demand point to each cluster ij , calculated as follows: where d ij represents the Euclidean distance from the jth demand point to the ith Cainiao station. Repeat formulas (2), (3) and (4) to update the cluster center position and the membership matrix until the change of the membership matrix U is less than a predetermined threshold or the maximum number of iterations is reached. Output the final cluster center and the membership matrix of each demand point to each cluster, thereby obtaining the classification results of each demand point. For each type of demand point, the fireworks-particle swarm algorithm is used to solve the Cainiao station position of each type of demand point. The classification result of the demand point corresponds to K in formula (1). ij The value of K, that is, if demand point j belongs to the i-th category, then ij =1, otherwise K ij =0.
5. The method for selecting multiple Cainiao stations based on the FCM-PSO-FWA algorithm as claimed in claim 4, characterized in that: The fireworks-particle swarm algorithm includes the following steps: (3) Initialize the particle swarm and determine the initial position of the particles, that is, the initial position of the Cainiao Station (X i ,Y i ))、initial speed (V xi ,V yi ), inertia factor w, the boundary range of the particle's horizontal and vertical coordinates (x max ,y max ) and (x min ,y min ) and particle search speed range [V min ,V max ] parameter value, for the inertia factor w, generally in the early stage of the algorithm, a larger inertia weight is used to enhance the global search capability of the algorithm, and a smaller inertia weight is used to improve the local search capability of the algorithm in the later stage of the algorithm. Therefore, this embodiment adopts a time-varying inertia weight linear decreasing strategy to update the inertia factor value, and its formula is: w(t)=(w max -w min )(1-t / T)+w min (5) where w max is the maximum value of random weight, w min is the minimum value of random weight, t is the current iteration number, and T is the maximum number of algorithm iterations; (4) Calculate the fitness of each particle. The fitness function is the objective function f in formula (1). According to the fitness value f of each particle i Update the local and global optimal solutions and optimal values. For each particle, compare its current fitness value with its known best fitness value. If the current fitness value is better, that is, the total pickup distance is shorter, then update the personal optimal solution of the particle and its corresponding position, that is, update the shortest total pickup distance and the Cainiao Station location. Then find the particle with the best fitness value among all the personal optimal solutions of the particle swarm, and compare this fitness value with the current global optimal solution. If it is better than the current global optimal solution, then update the global optimal solution and global optimal value. (5) Treat each particle as a firework, and use the fireworks algorithm to search for the local optimal solution of each particle using the sparks produced by the explosion within the explosion radius. Fireworks x i The explosion radius A i and the number of explosion sparks S i The calculation formulas are as follows: where y min =min(f(x i )),y max =max(f(x i )), i = 1, 2…N represent the minimum and maximum values of all spark fitness respectively, R and M are constants used to adjust the explosion radius and the number of explosion sparks respectively, ε is the minimum amount of the machine set to prevent the denominator of the formula from being zero, in order to prevent particles with good fitness from generating too many explosion sparks and particles with poor fitness from generating too few sparks, so the number of sparks generated by particle i S i The following constraints exist: Where a and b are constants, round() represents the rounding function, and then randomly select some sparks from the explosion sparks generated by each particle to generate variant sparks through Gaussian mutation, fireworks x i The variation formula in the kth dimension is shown in formula (9). If the sparks generated by the explosion or the variation sparks exceed the feasible region, the spark position is adjusted according to formula (10). x ik =x ik ×e (9) x ik =x LB,k +|x ik |%(x UB,k -x LB,k ) (10) Where e~N(1,1), x UB,k 、x LB,k They are the upper and lower boundaries of the spark in the kth dimension; (6) Compare the fitness value of each particle with the corresponding spark, update the local and global optimal solutions and optimal values, calculate the fitness value of the spark and the mutated spark generated by each particle, and if there is a spark whose fitness value is better than the particle fitness value, use the fitness value of the spark to update the local optimal value of the particle, and use the position of the spark to update the local optimal solution of the particle. Similarly, update the local optimal solutions of all remaining particles, and update the global optimal solution after updating all local optimal solutions; (7) Update the speed and position of each particle according to the local and global optimal solutions, and according to the boundary range of the horizontal and vertical coordinates of the Cainiao Station location in formula (1) (x max ,y max ) and (x min ,y min ) and the set particle speed range are used to adjust particles whose speed or position exceeds the boundary conditions. The particle speed and coordinate update formulas are shown in formulas (11) and (12). The adjustment method for speed or position exceeding the boundary is shown in formulas (13), (14) and (15): V ki =rd×(V max -V min )+V min (13) X i =rd×(x max -x min )+x min (14) AND i =rd×(y max -and min )+y min (15) In the formula represents the velocity of the i-th particle in the k-th dimension at the t-th iteration, p pk and p gk are the local and global optimal solutions in the kth dimension, c1 and c2 are the self-learning factor and social learning factor, r1 and r2 are random numbers in [0,1], α is the constraint factor used to control the weight, rd represents a random number in the range of [0,1), V ki represents the velocity of the i-th particle in the k-th dimension; (8) Determine whether the result meets the termination condition. If so, output the optimal solution and end the algorithm. If not, proceed to the next step. (9) Return to step (4) and continue iterating until the optimal location (X, Y) of this type of Cainiao Station is generated. Similarly, the location selection method for other types of Cainiao Stations is also as shown in steps (3)-(9).
Citation Information
Patent Citations
Cai Niao post site selection method based on improved k-means algorithm
CN110619351A
EVCS site selection planning method based on improved satin blue garden-dingbird algorithm
CN117788050A
Leakage detection system and method for long petroleum pipeline based on afpso-k-means
US20210033484A1
