A multi-period freight pricing and logistics network planning method

Abstract

The present invention provides a multi-period freight pricing and logistics network planning method, which relates to the technical field of logistics pricing and network planning. This method obtains relevant business information from a logistics planning database; preprocesses the business information to obtain model input data; determines a multi-period freight pricing and logistics network planning model based on the model input data; obtains a hybrid metaheuristic algorithm according to the particle swarm algorithm and the differential evolution algorithm; uses the hybrid metaheuristic algorithm to solve the multi-period freight pricing and logistics network planning model to obtain a logistics planning scheme; and determines the final freight pricing and the implementation scheme of the logistics network planning according to the logistics planning scheme. The present invention can determine the freight price according to the behavior decisions of different types of customers in the logistics network planning, and take into account the selection of logistics carriers, improving the practicability of the logistics network planning; and can improve the planning efficiency on the premise of ensuring the solution quality, increasing the profits of enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics pricing and network planning, and particularly to a multi-period freight pricing and logistics network planning method. Background Art

[0002] In today's business environment, the revenue management problem has become a new dimension in logistics network planning. Freight not only determines the unit revenue of each commodity, but also affects the demand in each customer area. It can change logistics network planning such as the required logistics facilities and transportation routes. However, in the real world, in addition to the logistics price affecting customer demand, the customer satisfaction with logistics services also has a great impact on demand. Different types of customers do not have the same evaluation of the logistics service level. Therefore, it is more in line with the real background to formulate freight prices according to the behaviors of different types of customers. Existing logistics network planning usually only considers the selection of routes and does not consider the selection of different logistics carriers on the same route, which will result in an increase in the cost of the entire logistics network and a decrease in the distribution efficiency. Therefore, a solution is needed that can determine freight prices according to the behaviors of different types of customers in logistics network planning and take into account the selection of logistics carriers. The logistics network planning problem under complex structures usually has a mathematical model with non-linear characteristics and exponentially growing complexity. Existing algorithms either have low solution accuracy or require a large amount of solution time and computing resources. Therefore, an algorithm that can obtain high accuracy in a short time is needed. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a multi-period freight pricing and logistics network planning method aiming at the deficiencies of the above-mentioned prior art, which can determine freight prices according to the behaviors of different types of customers in logistics network planning, take into account the selection of logistics carriers, and improve the practicability of logistics network planning.

[0004] To solve the above technical problem, the technical solution adopted by the present invention is:

[0005] A multi-period freight pricing and logistics network planning method, comprising:

[0006] Obtaining relevant business information in a logistics planning database; the business information includes: market potential freight price information, market potential customer node information, supply node information, transfer center node information, route information, and logistics carrier information;

[0007] Preprocessing the business information to obtain model input data;

[0008] Determining a multi-period freight pricing and logistics network planning model according to the model input data;

[0009] According to the particle swarm optimization algorithm and the differential evolution algorithm, a hybrid meta-heuristic algorithm is obtained;

[0010] The hybrid meta-heuristic algorithm is used to solve the multi-period freight pricing and logistics network planning model, and a logistics planning scheme is obtained; the logistics planning scheme includes a freight pricing plan, a transfer center node plan, a route plan, a logistics transporter plan, and a traffic volume plan;

[0011] The final freight pricing and the implementation plan of the logistics network planning are determined according to the logistics planning scheme.

[0012] The beneficial effects of adopting the above technical solutions are as follows: The multi-period freight pricing and logistics network planning method provided by the present invention simultaneously considers different types of customer behaviors, logistics freight prices, and logistics transporters in the multi-period logistics network planning, establishes a multi-period freight pricing and logistics network planning model, and designs a hybrid meta-heuristic algorithm based on the particle swarm optimization algorithm and the differential evolution algorithm. On the premise of ensuring the solution quality, the planning efficiency is improved, and multi-period network planning solutions such as freight pricing, transfer center nodes, routes, logistics transporters, and traffic volume plans of the complex network are obtained. Compared with the basic network planning scheme, the freight price is determined according to different types of customer behaviors in the logistics network planning, and the selection of logistics transporters is taken into account, which improves the practicability of the logistics network planning, meets the actual requirements, can improve the planning efficiency on the premise of ensuring the solution quality, and increases the profit of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a flowchart of the multi-period freight pricing and logistics network planning method provided by the embodiment of the present invention;

[0014] Figure 2 It is a relationship diagram between freight pricing and demand provided by the embodiment of the present invention;

[0015] Figure 3 It is a schematic diagram of the method for constructing the multi-period freight pricing and logistics network planning model provided by the embodiment of the present invention;

[0016] Figure 4 It is a schematic diagram of the design process of the multi-period freight pricing and logistics network planning solution algorithm provided by the embodiment of the present invention;

[0017] Figure 5 It is a schematic diagram of the logistics network planning scheme provided by the embodiment of the present invention;

[0018] Figure 6 It is a flowchart of the solution of the multi-period freight pricing and logistics network planning provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following further describes the specific implementation manners of the present invention in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0020] As Figure 1 shown, the multi-period freight pricing and logistics network planning method provided in this embodiment is described as follows.

[0021] Step 1: Obtain relevant business information in the logistics planning database. The business information includes: market potential freight price information, market potential customer node information, supply node information, transfer center node information, route information, and logistics transporter information.

[0022] Step 2: Preprocess the business information to obtain model input data. Specifically, it includes:

[0023] Preprocess the business information according to the period to obtain a period set; preprocess the business information according to the price level to obtain a price level set; preprocess the business information according to the supply nodes to obtain a supply node set; preprocess the business information according to the transfer center nodes to obtain a transfer center node set; preprocess the business information according to the logistics transporters to obtain a logistics transporter set; preprocess the business information according to the customer nodes to obtain a customer node set; determine the model input data according to the period set, the price level set, the supply node set, the transfer center node set, the logistics transporter set, and the customer node set.

[0024] As Figure 2 shown, different freight price levels correspond to different demand quantities, different supply nodes have different supply capabilities, different transfer center nodes have different processing capabilities and processing costs, and different logistics transporters have different transportation capabilities and transportation costs.

[0025] Step 3: Determine a multi-period freight pricing and logistics network planning model according to the model input data.

[0026] In this embodiment, the multi-period freight pricing and logistics network planning model takes the maximum profit as the objective function, and constraints the capabilities of the supply nodes, transfer center nodes, and logistics transporters and the logistics network structure. As Figure 3 shown, the construction method of the multi-period freight pricing and logistics network planning model in this embodiment includes the following steps:

[0027] Step 301: Obtain the model input data;

[0028] Step 302: Determine a plurality of decision variables according to the model input data.

[0029] In this embodiment, the decision variables include: the logistics carriers between different nodes of the logistics network in different periods, the construction situation of the transshipment center nodes in different periods, the logistics freight price levels in different periods, and the transportation volumes of goods between different nodes in different periods.

[0030] Decision variable 1:

[0031] In this embodiment, the decision variable 1 "the logistics carrier between any two nodes of the supply node and the transshipment center node" is represented by , then u ∈ U, k ∈ K su , t ∈ T, where T is the set of periods; S represents the set of supply nodes; U represents the set of transshipment center nodes; K su represents the set of logistics carriers between the supply node s and the transshipment center node u.

[0032] If the k-th logistics carrier between the supply node s and the transshipment center node u is selected in period t, then Otherwise,

[0033] Decision variable 2:

[0034] In this embodiment, the decision variable 2 "the logistics carrier between any two nodes of the transshipment center node and the customer node" is represented by , then d ∈ D, k ∈ K ud , t ∈ T, where T is the set of periods; U represents the set of transshipment center nodes; D represents the set of customer nodes; K ud represents the set of logistics carriers between the transshipment center node u and the customer node d.

[0035] If the k-th logistics carrier between the transshipment center node u and the customer node d is selected in period t, then Otherwise

[0036] Decision variable 3:

[0037] In this embodiment, the decision variable 3 "the transshipment center node" is represented by , then t ∈ T.

[0038] If the transshipment center node u is selected in period t, then Otherwise

[0039] Decision variable 4:

[0040] In this embodiment, the decision variable 4 "the logistics freight price level" is represented by , then l ∈ L, t ∈ T, where L is the set of logistics freight price levels.

[0041] If the l-th logistics freight price level is selected in period t, then Otherwise

[0042] Decision variable 5:

[0043] In this embodiment, the decision variable 5, "the commodity transportation volume between any two nodes of the supply node and the transfer center node", is represented by Then u ∈ U, k ∈ K su , t ∈ T, where

[0044] Decision variable 6:

[0045] In the embodiment, the decision variable 6, "the commodity transportation volume between any two nodes of the transfer center node and the customer node", is represented by Then d ∈ D, k ∈ K ud , t ∈ T, where

[0046] Step 303: Construct an objective function with the goal of maximizing profit.

[0047] In this embodiment, according to the decision variables, an objective function is constructed with the goal of maximizing profit. Specifically, the total revenue is determined based on the logistics freight price and the actual commodity transportation volume, and the total cost is determined by the construction cost of the transfer center node, the cooperation cost of the logistics transportation provider, the processing cost of the transfer center node, the transportation cost of the logistics transportation provider, and the shortage cost of the commodity. Finally, the total profit is obtained:

[0048]

[0049] Where is the freight price, is the fixed construction cost of the transfer center, is used to determine whether the transfer center has been selected before the current period, is the fixed cooperation cost of the k-th logistics transportation provider from the supply node to the transfer center node, is used to determine whether the k-th logistics transportation provider from the supply node to the transfer center node has been selected before the current period, is the fixed cooperation cost of the k-th logistics transportation provider from the transfer center node to the customer node, is used to determine whether the k-th logistics transportation provider from the transfer center node to the customer node has been selected before the current period, is the unit transportation cost of the k-th logistics carrier from the supply node to the transfer center node, is the unit transportation cost of the k-th logistics carrier from the transfer center node to the customer node, is the unit handling cost of the transfer center, is the shortage quantity of goods, is the unit shortage cost.

[0050] Step 304, determine multiple constraint conditions of the objective function.

[0051] Constraint condition 1:

[0052] In this embodiment, the logistics service level per cycle depends on the ratio of the actual delivery volume to the customer demand. Constraint condition 1 includes:

[0053]

[0054] Among them, is the logistics service level.

[0055] Constraint condition 2:

[0056] In this embodiment, the customer demand per cycle depends not only on the price but also on the customer satisfaction of the previous cycle. In the first cycle, the actual demand of the customer is equal to the customer's market share; and from the second cycle, the actual demand of the customer in the current cycle is jointly determined by the freight pricing in the current cycle and the customer satisfaction of the previous cycle. Since customers in the real world are usually irrational, most people are asymmetrically sensitive to losses and gains, and the pain of facing losses is much greater than the pleasure of facing gains. Therefore, this embodiment uses the prospect theory value function to characterize the attitudes of different types of customers towards the logistics service level, that is, customer satisfaction, so as to affect the actual demand of customers. Constraint condition 2 includes:

[0057]

[0058] Among them, is the service level required by the customer, is the customer market share, is the actual demand of the customer, α d is the risk attitude coefficient of different customers when facing gains, β d is the risk attitude coefficient when facing losses, λ d is the loss aversion coefficient when facing losses.

[0059] Constraint condition 3:

[0060] In this embodiment, for the multi-period freight pricing and logistics network planning model, if a transshipment center node is constructed in a certain period, then the construction cost of this transshipment center node does not need to be considered in subsequent periods. Constraint 3 includes:

[0061]

[0062] Constraint 4:

[0063] In this embodiment, for the multi-period freight pricing and logistics network planning model, if a cooperation is reached with a certain logistics provider on a certain line between a supply node and a transshipment center node in a certain period, then the cooperation cost of this logistics provider does not need to be considered on this line in subsequent periods. Constraint 4 includes:

[0064]

[0065] Constraint 5:

[0066] In this embodiment, for the multi-period freight pricing and logistics network planning model, if a cooperation is reached with a certain logistics provider on a certain line between a transshipment center node and a customer node in a certain period, then the cooperation cost of this logistics provider does not need to be considered on this line in subsequent periods. Constraint 5 includes:

[0067]

[0068] Constraint 6:

[0069] In this embodiment, the transportation capacity limit of the logistics provider between the supply node and the transshipment center node is considered, and the quantity of goods transported cannot exceed its transportation capacity upper limit. Constraint 6 includes:

[0070]

[0071] Among them, is the transportation capacity of the k-th logistics transporter between the supply node and the transshipment center node.

[0072] Constraint 7:

[0073] In this embodiment, the transportation capacity limit of the logistics provider between the transshipment center node and the customer node is considered, and the quantity of goods transported cannot exceed its transportation capacity upper limit. Constraint 7 includes:

[0074]

[0075] Among them, is the transportation capacity of the k-th logistics transporter between the transshipment center node and the customer node.

[0076] Constraint 8:

[0077] In this embodiment, the processing capacity limit of the transfer center node is considered, and the throughput of processed goods cannot exceed its processing capacity limit. Constraint 8 includes:

[0078]

[0079] Among them, is the processing capacity of the transfer center.

[0080] Constraint 9:

[0081] In this embodiment, the supply capacity limit of the supply node is considered, and the total amount of goods flowing out of the supply node cannot exceed its maximum supply. Constraint 9 includes:

[0082]

[0083] Among them, is the supply capacity of the supply node.

[0084] Constraint 10:

[0085] In this embodiment, the flow balance limit of the transfer center node is considered, indicating that the total amount of goods entering a certain transfer center node is equal to the total amount of goods from the transfer center node, that is, the transfer center node does not detain and store goods. Constraint 10 includes:

[0086]

[0087] Constraint 11:

[0088] In this embodiment, it is considered that if a certain transfer center node is not selected, then the logistics carrier from the supply node to the transfer center node cannot be selected either. Constraint 11 includes:

[0089]

[0090] Constraint 12:

[0091] In this embodiment, it is considered that if a certain transfer center node is not selected, then the logistics carrier from the transfer center node to the customer node cannot be selected either. Constraint 12 includes:

[0092]

[0093] Constraint 13:

[0094] In this embodiment, it is considered that if a certain transfer center node is selected, then at least one logistics carrier provides services from the supply node to the transfer center node. Constraint 13 includes:

[0095]

[0096] Constraint 14:

[0097] In this embodiment, it is considered that if a certain transfer center node is selected, then at least one logistics carrier provides services from the transfer center node to the customer node. Constraint 14 includes:

[0098]

[0099] Constraint 15:

[0100] In this embodiment, the constraint of the customer demand node is considered. The total quantity of goods actually arriving cannot exceed its customer demand quantity, that is, the customer demand quantity in the current period is equal to the sum of the shortage quantity in the current period and the actual quantity delivered in the current period. Constraint 15 includes:

[0101]

[0102] Constraint 16:

[0103] In this embodiment, it is considered that only one logistics freight price level can be selected in each period. Constraint 16 includes:

[0104]

[0105] Step 305: According to the multiple decision variables in Step 302, the objective function in Step 303, and the multiple constraint conditions in Step 304, construct a multi-period freight pricing and logistics network planning model.

[0106] Step 4: Obtain a hybrid meta-heuristic algorithm according to the particle swarm optimization algorithm and the differential evolution algorithm.

[0107] The design process of the multi-period freight pricing and logistics network planning solution algorithm is as Figure 4 shown, and specifically includes:

[0108] Step 401: Encoding process. Solve the model by using one-dimensional integer encoding, and divide the encoding process into two parts;

[0109] The first part: First, number each node in the potential logistics network in the order of the supply node, the transfer center node, and the customer node, and convert the actual distribution of each node into a potential logistics network planning diagram with numbers, that is, from Figure 5 (a) is converted to Figure 5(b). Then, an adjacency matrix is established based on the number of logistics carriers between two nodes in the logistics network diagram. The integer coding range of a particle at a certain position is determined according to the number of selectable logistics carriers between two nodes in the adjacency matrix. Suppose there are n selectable logistics carriers between two certain nodes. For the particle swarm algorithm, the particle position coding range is [0, 2 n - 1], and the particle velocity coding range is [-(2 n - 1), 2 n - 1]; for the differential evolution algorithm, the particle coding range is [0, 2 n - 1]. The coding length of the first part of the particle depends on the number of node pairs with selectable logistics carriers in the potential logistics network, that is, the number of non-zero elements in the adjacency matrix, denoted as z.

[0110] Part Two: Encode the freight price levels of different customer nodes. Suppose the number of customer nodes in the potential logistics network is d, and the freight price level is l. For the particle swarm algorithm, the particle position coding range is [0, l - 1], and the particle velocity coding range is [-(l - 1), l - 1]; for the differential evolution algorithm, the particle coding range is [0, l - 1]. The coding length of the second part of the particle depends on the number of customer nodes in the potential logistics network, which is d.

[0111] In summary, for the multi-period freight pricing and logistics network planning model, the coding length of the particle in each period is (z + d). Suppose there are T periods in total, then the total coding length of the particle is T * (z + d).

[0112] Step 402: Decoding process. According to the encoding process, the decoding process is divided into two parts. The specific decoding process includes:

[0113] Part One: Decode the first part in the encoding process in a binary manner. Just convert the integer coding to 0 - 1 binary, and then the selection situation of logistics carriers between any two nodes can be obtained. Specifically, number the logistics carriers between two nodes (i, j). Then convert the integer coding into binary 0 - 1 coding, and judge the selection situation of logistics carriers between the two nodes in the order from the low bit to the high bit. If the k-th bit is 1, it means the k-th logistics carrier between two nodes (i, j) is selected, that is, the value of the decision variable is 1; if the k-th bit is 0, it means the k-th logistics carrier between two nodes (i, j) is not selected, that is, the value of the decision variable is 0. Thus, the selection situation of the decision variable transfer center node can be determined.

[0114] Assume that there are three alternative logistics carriers between node 1 and node 3, i.e., n = 3. The binary 0-1 decoding corresponding to the integer coding and the situation of the selected logistics carriers are shown in Table 1.

[0115] Table 1 Example of the first part of decoding

[0116]

[0117] Second part: Decode the second part in the coding process. The integer coding range of the particle is [0, l - 1], then the freight price level range of the customer nodes in the potential logistics network is [1, l]. When the coding value of the particle is m (m ≤ l), then the decision variable has a value of 1, that is, the customer node j ∈ D selects the freight price level m within the period t ∈ T. Assume that the highest level of the freight price l = 4, and the coding range of the particle at each customer node is [0, 3], then the freight price level range of the customer nodes in the potential logistics network is [1, 4], and the specific corresponding relationship is shown in Table 2.

[0118] Table 2 Example of the second part of decoding

[0119]

[0120] Step 403: Initialization process, which is randomly initialized within the coding range. For the particle swarm algorithm, the initialization includes randomly initializing two elements: velocity and position; for the differential evolution algorithm, each dimension of each particle is randomly initialized.

[0121] Step 404: Update process. The update processes of the particle swarm algorithm and the differential evolution algorithm are slightly different, specifically including:

[0122] For the particle swarm algorithm, assume that there are m particles in the population, the dimension of the solution space is N, x i , v i , p i and p g respectively represent the current position, velocity, historical best solution of the i-th particle, and the global best solution passed by all particles in the population. The update formula for the velocity of the particles in the population is:

[0123]

[0124] The update formula for the position of the particles in the population is:

[0125]

[0126] Among them, the inertia weight is w, and the learning factors are c1 and c2, both of which are non-negative real numbers; ξ, η ∈ U[0, 1]. Specifically, it includes: Let \(tt\) represent the current iteration number of the algorithm, and \(TT\) represent the maximum iteration number of the algorithm.

[0127] After the particles in the population update their velocities and positions, they are first rounded, but may exceed their own value ranges. The following method is used to handle this:

[0128] If the value of a certain position encoding in a particle exceeds the upper bound of the encoding, then the upper bound value of the encoding is used as the position encoding value of that bit; if the value of a certain position encoding in a particle is lower than the lower bound of the encoding, then the lower bound value of the encoding is used as the position encoding value of that bit. If the value of a certain velocity encoding in a particle exceeds its value range, the same method is used for processing.

[0129] For the differential evolution algorithm, the update process consists of mutation operation, crossover operation, and selection strategy. The DE / best / 2 / bin method is adopted, where best represents the optimal value as the base, 2 represents the number of differential vectors, and bin represents the binomial crossover strategy, which specifically includes:

[0130] Each individual in the population generates a corresponding mutant individual according to the following formula:

[0131] \(V\) i (tt + 1)=X best (tt)+F(tt)[(X r1 (tt)-X r2 (tt))+(X r3 (tt)-X r4 (tt))]

[0132] Among them, \(i\) represents the index number of the target individual, \(r1\), \(r2\), \(r3\), \(r4\) respectively represent the index numbers of different individuals in the population, and \(i\), \(r1\), \(r2\), \(r3\), \(r4\) are mutually different, \(X\) best (tt) represents the optimal individual in the \(tt\)-th generation population,

[0133]

[0134] The differential vector is generated by randomly selecting 4 different individuals, and then a mutation operation is performed on \(X\) best (tt). Since integer encoding is used in the encoding process, the result is rounded after calculation by the above formula. After rounding the mutant individual, the situation of exceeding the encoding range may occur, so each bit of the individuals in the population is processed according to the following formula:

[0135]

[0136] Among them, \(V\) ij (tt + 1) represents the value of the \(i\)-th individual in the \((tt + 1)\)-th generation population in the \(j\)-th dimension, Denote the maximum value of the individual in the j-th dimension.

[0137] Then, the current individual X i (tt) and the mutated individual V i (tt + 1) are subjected to a crossover operation to generate a new trial individual U i (tt):

[0138]

[0139] where j represents the j-th dimension of the i-th individual, rand is a random number between [0, 1], and CR is the crossover probability, CR ∈ [0, 1].

[0140] Finally, a selection strategy of "greedy strategy" is adopted to evaluate the trial individual U i (tt) and the current individual X i (tt), and select the better individual between the two as the individual of the offspring.

[0141] Step 405: Minimum-cost maximum-flow algorithm. When the freight pricing and the basic structure of the logistics network are determined, it is necessary to allocate the transportation volume for each logistics transporter in the logistics network according to customer demands and constraint factors such as capacity and cost in the logistics network. In this embodiment, the minimum-cost maximum-flow algorithm is used to evaluate the selected network and calculate the total network cost. Since the minimum-cost maximum-flow algorithm has certain limitations in solving network planning problems, it is applicable to solving network problems with only a single source point and a single sink point, and no capacity constraints for intermediate nodes. Therefore, the potential logistics network is transformed into a network problem with a single source point, a single sink point, and no capacity constraints for intermediate nodes. Specifically, it includes:

[0142] Add a virtual source point as the starting point of the entire network and add a virtual sink point as the ending point of the entire network. The capacities of the virtual arcs from the virtual source point to the supply nodes and from the customer nodes to the virtual sink point are both set to infinity, and the cooperation cost and the unit transportation cost are both 0. In addition, since the transfer center nodes also have processing capacity constraints and cost attributes, each transfer center node is split into two points, and the processing capacity constraints and cost attributes of the transfer center node are transferred to the virtual arc between the two nodes. The processing capacity constraints and unit processing cost attributes of the transfer center node are converted into the transportation capacity constraints and unit transportation cost attributes of the corresponding virtual transportation arc, and the construction cost of the transfer center node is converted into the construction cost of the corresponding virtual transportation arc.

[0143] Step 406: Solution repair process. Since the obtained solution may not meet the actual constraint conditions, that is, the network may be disconnected, it is necessary to repair the infeasible solution.

[0144] There are three cases of network disconnection:

[0145] For the case where the customer node is disconnected from the network, one or more arcs are randomly added to the network to connect to the customer node, the positions of the disconnected node pairs in the particle are found, and finally an integer is randomly selected within the value range to replace the 0 value at the original position;

[0146] For the case where the supply node is disconnected from the network, one or more arcs are randomly added to the network to connect to the supply node, and the remaining processing methods are the same as those in the above case;

[0147] For the case where the transfer center node is disconnected from the network, if there are incoming arcs but no outgoing arcs, one or more arcs are randomly added to the network starting from the transfer center node; if there are outgoing arcs but no incoming arcs, one or more arcs reaching the transfer center node are randomly added to the network starting from the supply node, and the remaining processing methods are the same as those in the above case.

[0148] New infeasible solutions may be generated during the repair process. Therefore, the logistics network needs to be repaired multiple times, and the logistics network should be detected before and after each repair until the entire network is fully connected and the goods can reach the customers smoothly.

[0149] Step 407: Evaluation of the solution. The maximum value of the total profit in the logistics network is used as a fitness function, that is, the objective function. The aim is to find the particle that maximizes the total profit in the network among all particles that satisfy the constraints. This particle is the obtained target logistics network.

[0150] Step 408: Algorithm selection mechanism. To improve the planning efficiency while ensuring the solution quality, an optimal algorithm selection mechanism is designed to determine which algorithm to use in each iteration according to the change of individual fitness values in the population and the change of the global optimal value of the population. Specifically, it includes:

[0151] The change of individual fitness values in the population. First, the particle swarm optimization algorithm is applied to the initially randomly generated population X(0). The individuals in the original population are X(0) = (x1(0), x2(0),..., x NP (0)), and the individuals in the updated population are u = (u1, u2,..., u NP ). During the process of updating the current individual optimal value and the optimal solution, a new variable lbp is introduced. lbp is the number of individuals in the updated population u that replace the individuals in the original population X(0); Similarly, the differential evolution algorithm is applied to the initially randomly generated population X(0), and the individuals in the updated population are During the process of updating the current individual optimal value and the optimal solution, a new variable lbd is introduced. lbd is the updated population u* The number of individuals replacing those in the original population X(0); the change in the fitness values of the individuals in the population is represented by the variable lb, where

[0152] Regarding the change in the global optimal value of the population, during the global search process, the variable gb is introduced. If the particle swarm optimization algorithm is used in the tt - th generation iteration and the current global optimal value is not worse than that of the previous generation, then gb = 1; otherwise gb = 2. If the differential evolution algorithm is used in the tt - th generation iteration and the current global optimal value is not worse than that of the previous generation, then gb = 2; otherwise gb = 1.

[0153] Algorithm selection mechanism: Determine which algorithm to use in the next iteration according to the values of lb and gb in each generation. When lb ≥ 0.5 and gb = 1, it indicates that after the operation of the particle swarm optimization algorithm, more individuals' fitness values in the population are updated, and the global optimal solution is updated compared with the previous generation. Then the particle swarm optimization algorithm is used in the next generation. When lb < 0.5 and gb = 2, it means that after the operation of the differential evolution algorithm, more individuals' fitness values in the population are updated, and the global optimal solution is updated compared with the previous generation. Then the differential evolution algorithm is used in the next generation. When lb ≥ 0.5 and gb = 2, it shows that after the operation of the particle swarm optimization algorithm, more individuals' fitness values in the population are updated, but the global optimal solution is not updated. According to the roulette rule, a random number rand is generated. If rand ≤ lb, then the differential evolution algorithm is used in the next generation; if rand > lb, then the particle swarm optimization algorithm is used in the next generation. When lb < 0.5 and gb = 1, it means that after the operation of the differential evolution algorithm, more individuals' fitness values in the population are updated, but the global optimal solution is not updated. According to the roulette rule, a random number rand is generated. If rand < lb, then the differential evolution algorithm is used in the next generation; if rand ≥ lb, then the particle swarm optimization algorithm is used in the next generation. The specific selection situations are shown in Table 3:

[0154] Table 3 Algorithm selection mechanism

[0155] Individual fitness judgment condition Global optimal value judgment condition Roulette wheel judgment condition Algorithm used for the next generation lb≥0.5 gb = 1 - Particle swarm optimization algorithm lb≥0.5 gb = 2 rand ≤ lb Differential evolution algorithm lb≥0.5 gb = 2 rand > lb Particle swarm optimization algorithm lb＜0.5 gb = 2 - Differential evolution algorithm lb＜0.5 gb = 1 rand < lb Differential evolution algorithm lb＜0.5 gb = 1 rand ≥ lb Particle swarm optimization algorithm

[0156] Step 409: Algorithm termination condition. If the maximum number of iterations is reached, the algorithm ends.

[0157] Step 5: Solve the multi - period freight pricing and logistics network planning model to obtain the logistics planning scheme.

[0158] The specific process of solving the multi - period freight pricing and logistics network planning is as Figure 6 shown, including:

[0159] Step 501: Initialize the population size NP, dimension D, maximum number of iterations NG, initial population X, and parameters in the particle swarm optimization algorithm and differential evolution algorithm.

[0160] Step 502: Set the initial number of iterations, i.e., tt = 0. At the same time, use the particle swarm optimization algorithm and differential evolution algorithm to solve the multi-period freight pricing and logistics network planning models respectively according to Steps 401 - 407, and calculate the initial values of lb and gb. Select the algorithm used in the first iteration according to the algorithm selection mechanism in Step 408.

[0161] Step 503: If it is determined to select the particle swarm optimization algorithm, use the particle swarm optimization algorithm to solve the multi-period freight pricing and logistics network planning models according to Steps 401 - 407; if it is determined to select the differential evolution algorithm, use the differential evolution algorithm to solve the multi-period freight pricing and logistics network planning models according to Steps 401 - 407.

[0162] Step 504: Update the values of lb and gb according to the above solution results, and select the algorithm used in the next iteration according to the algorithm selection mechanism in Step 408.

[0163] Step 505: Update the current individual optimal value, optimal solution, current global optimal value, and optimal solution.

[0164] Step 506: Determine whether the termination condition is met. If Step 409 is satisfied, end the solution; if not, return to Step 503.

[0165] A logistics planning solution can be obtained according to the above solution algorithm process.

[0166] Step 6: Determine the final freight pricing and the implementation plan of the logistics network planning according to the logistics planning solution, as Figure 5 shown in (c).

[0167] In this embodiment, different types of customer behaviors, logistics freight prices, and logistics carriers are simultaneously considered in the multi-period logistics network planning. A multi-period freight pricing and logistics network planning model is established, and a hybrid meta-heuristic algorithm is designed based on the particle swarm optimization algorithm and differential evolution algorithm. On the premise of ensuring the solution quality, the planning efficiency is improved, and a multi-period network planning solution such as freight pricing, transfer center nodes, routes, logistics carriers, and shipment volume plans of the complex network is obtained. This solution can formulate logistics freight prices according to different types of customer behaviors in the multi-period logistics network planning, take into account the selection of logistics carriers, improve the practicability of the logistics network planning, meet the actual requirements, and increase the profits of the enterprise.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A multi-period freight pricing and logistics network planning method, characterized in that: The method includes: Obtaining relevant business information in the logistics planning database; Preprocessing the business information to obtain model input data; Determining a multi-period freight pricing and logistics network planning model according to the model input data; Obtaining a hybrid meta-heuristic algorithm according to the particle swarm optimization algorithm and the differential evolution algorithm; Solving the multi-period freight pricing and logistics network planning model by using the hybrid meta-heuristic algorithm to obtain a logistics planning scheme; the logistics planning scheme includes a freight pricing plan, a transfer center node plan, a route plan, a logistics transporter plan, and a traffic volume plan; Determining a final freight pricing and a logistics network planning implementation plan according to the logistics planning scheme; The hybrid meta-heuristic algorithm specifically includes: Solving the model by using a one-dimensional integer coding method, and dividing the coding process into two parts; Dividing the decoding process into two parts according to the coding process, where the first part corresponds to logistics network planning and the second part corresponds to freight pricing; The population initialization process adopts a random initialization method; Updating the particles in the population according to the update formulas of the particle swarm optimization algorithm and the differential evolution algorithm; According to the above coding, decoding, initialization, and update processes, the freight pricing and the basic structure of the logistics network can be determined, and the multi-period freight pricing and logistics network planning problem is converted into a problem of allocating traffic volume to each logistics transporter in the logistics network; Solving the new problem according to the minimum cost maximum flow algorithm; Repairing the infeasible solutions by using a repair strategy according to the existing real-world constraint conditions; Taking the maximum value of the total profit in the logistics network, that is, the objective function, as a fitness function; Determining an algorithm selection mechanism according to the change of the individual fitness value in the population and the change of the global optimal value of the population to decide which algorithm to use in each iteration process, that is, which algorithm to use to solve the multi-period freight pricing and logistics network planning model; Taking the maximum number of iterations as the termination condition of the algorithm, that is, when the maximum number of iterations is reached, the algorithm ends.

2. The multi-period freight pricing and logistics network planning method according to claim 1, characterized in that: The business information includes: market potential freight price information, market potential customer node information, supply node information, transfer center node information, route information, and logistics transporter information.

3. The multi-period freight rate pricing and logistics network planning method according to claim 2, characterized in that: Preprocessing the business information to obtain model input data specifically includes: Preprocessing the business information according to the period to obtain a period set; preprocessing the business information according to the price level to obtain a price level set; preprocessing the business information according to the supply node to obtain a supply node set; preprocessing the business information according to the transfer center node to obtain a transfer center node set; preprocessing the business information according to the logistics transporter to obtain a logistics transporter set; preprocessing the business information according to the customer node to obtain a customer node set; Determining the model input data according to the period set, the price level set, the supply node set, the transfer center node set, the logistics transporter set, and the customer node set.

4. The multi-period freight pricing and logistics network planning method according to claim 3, wherein: The multi-period freight pricing and logistics network planning model takes maximizing profit as the objective function, and constraints the capabilities of supply nodes, transfer center nodes, and logistics carriers and the logistics network structure. Determine the decision variables, and construct the objective function with the goal of maximizing profit. Among them, the total revenue is determined according to the logistics freight price and the actual commodity transportation volume, and the total cost is determined by the construction cost of the transfer center node, the cooperation cost of the logistics carrier, the processing cost of the transfer center node, the transportation cost of the logistics carrier, and the shortage cost of the commodity. Finally, the total profit is obtained, that is, the objective function, and its expression is: Among them, T is the set of periods, L is the set of price levels, S is the set of supply nodes, U is the set of transshipment center nodes, D is the set of customer nodes, and K su is the set of logistics carriers between supply nodes and transshipment center nodes, and K ud is the set of logistics carriers between transshipment center nodes and customer nodes, is the freight price, is the transportation volume of the k-th logistics carrier between transshipment center nodes and customer nodes, is the selection situation of the freight price level, is the fixed construction cost of the transshipment center, is the selection situation of the transshipment center, is used to judge whether a transshipment center has been selected before the current period, is the fixed cooperation cost of the k-th logistics carrier between supply nodes and transshipment center nodes, is the selection situation of the k-th logistics carrier between supply nodes and transshipment center nodes, is used to judge whether the k-th logistics carrier between supply nodes and transshipment center nodes has been selected before the current period, is the fixed cooperation cost of the k-th logistics carrier between transshipment center nodes and customer nodes, is the selection situation of the k-th logistics carrier between transshipment center nodes and customer nodes, is used to judge whether the k-th logistics carrier between transshipment center nodes and customer nodes has been selected before the current period, is the unit transportation cost of the k-th logistics carrier between supply nodes and transshipment center nodes, is the transportation volume of the k-th logistics carrier between supply nodes and transshipment center nodes, is the unit transportation cost of the k-th logistics carrier between transshipment center nodes and customer nodes, is the unit handling cost of the transshipment center, is the shortage quantity of goods, is the unit shortage cost; The expression of the constraint condition is: Among them, is the logistics service level, is the service level required by customers, is the customer market share, is the actual demand of customers, α d is the risk attitude coefficient of different customers when facing benefits, β d is the risk attitude coefficient when facing losses, λ d is the loss aversion coefficient when facing losses, is the transportation capacity of the k-th logistics carrier between the supply node and the transfer center node, is the transportation capacity of the k-th logistics carrier between the transfer center node and the customer node, is the processing capacity of the transfer center, is the supply capacity of the supply node.

5. The multi-period freight rate pricing and logistics network planning method according to claim 4, wherein: The specific method of the algorithm selection mechanism includes: The change of individual fitness values in the population. First, perform the particle swarm optimization algorithm on the initially randomly generated population X(0). The individuals in the original population are X(0) = (x1(0), x2(0),..., x NP (0)), and the individuals in the updated population are u = (u1, u2,..., u NP ), where NP is the population size. During the process of updating the current individual optimal value and the optimal solution, a new variable lbp is introduced. lbp is the number of individuals in the updated population u that replace the individuals in the original population X(0); Similarly, perform the differential evolution algorithm on the initially randomly generated population X(0). The individuals in the updated population are During the process of updating the current individual optimal value and the optimal solution, a new variable lbd is introduced. lbd is the number of individuals in the updated population u * that replace the individuals in the original population X(0); The change of individual fitness values in the population is represented by the variable lb, where The change of the global optimal value of the population. In the global search process, introduce the variable gb. If the particle swarm optimization algorithm is used in the tt-th generation iteration process and the current global optimal value is not worse than the previous generation's global optimal value, then gb = 1; otherwise gb = 2. If the differential evolution algorithm is used in the tt-th generation iteration process and the current global optimal value is not worse than the previous generation's global optimal value, then gb = 2; otherwise gb = 1. The algorithm selection mechanism determines which algorithm to use in the next iteration process according to the values of lb and gb in each generation. Specifically as follows: When lb ≥ 0.5 and gb = 1, it means that after the particle swarm optimization algorithm operation, the number of updated individual fitness values in the population is more, and the global optimal solution has been updated compared with the previous generation, so the particle swarm optimization algorithm is used in the next generation. When lb < 0.5 and gb = 2, it means that after the differential evolution algorithm operation, the number of updated individual fitness values in the population is more, and the global optimal solution has been updated compared with the previous generation, so the differential evolution algorithm is used in the next generation. When lb ≥ 0.5 and gb = 2, it means that after the particle swarm optimization algorithm operation, the number of updated individual fitness values in the population is more, but the global optimal solution has not been updated. According to the roulette rule, generate a random number rand. If rand ≤ lb, the differential evolution algorithm is used in the next generation. If rand > lb, the particle swarm optimization algorithm is used in the next generation. When lb < 0.5 and gb = 1, it means that after the differential evolution algorithm operation, the number of updated individual fitness values in the population is more, but the global optimal solution has not been updated. According to the roulette rule, generate a random number rand. If rand < lb, the differential evolution algorithm is used in the next generation. If rand ≥ lb, the particle swarm optimization algorithm is used in the next generation.

6. The multi-period freight rate pricing and logistics network planning method according to claim 5, wherein: The specific process of solving the multi-period freight pricing and logistics network planning model includes: Step 501: Initialize the population size NP, dimension D, maximum number of iterations NG, initial population X, and parameters in the particle swarm optimization algorithm and the differential evolution algorithm. Step 502: Set the initial number of iterations, that is, tt = 0. At the same time, use the particle swarm optimization algorithm and the differential evolution algorithm to solve the multi-period freight pricing and logistics network planning model respectively, and calculate the initial values of lb and gb. Select the algorithm used in the first iteration according to the algorithm selection mechanism. Step 503: If it is determined to select the particle swarm optimization algorithm, use the particle swarm optimization algorithm to solve the multi-period freight pricing and logistics network planning model; if it is determined to select the differential evolution algorithm, use the differential evolution algorithm to solve the multi-period freight pricing and logistics network planning model; Step 504: Update the values of lb and gb according to the above solution results, and select the algorithm to be used in the next iteration according to the algorithm selection mechanism; Step 505: Update the current individual optimal value, optimal solution, current global optimal value, and optimal solution; Step 506: Determine whether the termination condition is satisfied. If it is satisfied, end the solution and obtain the logistics planning scheme; if it is not satisfied, return to Step 503.

Images