Electric logistics vehicle path optimization method based on improved adaptive large neighborhood search algorithm

Abstract

The invention discloses an electric vehicle end distribution path optimization method based on an improved adaptive large neighborhood search algorithm. The method comprises the following steps: 1) constructing an electric logistics vehicle end distribution path optimization model; 2) defining a plurality of destruction operators and repair operators to improve the adaptive large neighborhood search algorithm; 3) adjusting operator weights by adopting a roulette method, and selecting different operators; and 4) controlling the acceptance process of the solution according to a simulated annealing criterion so as to obtain an optimal path scheme. According to the invention, distribution and energy supplement routes can be arranged for large-scale end distribution demand points in a short time, so that the distribution efficiency is effectively improved, and the total time of end distribution of the electric logistics vehicle is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle path planning, and in particular to an electric logistics vehicle path optimization method based on an improved adaptive large neighborhood search algorithm. Background Art

[0002] The vehicle routing problem (VRP) was first proposed by Dantzig and Ramser in 1959. With the continuous advancement of science and technology, many related problems have been widely studied, most of which are based on traditional vehicle models. Traditional vehicles have long cruising range and there are dense gas stations in cities, so refueling is usually not considered in these traditional VRPs. However, the range of electric vehicles is not yet comparable to that of traditional vehicles, and even in cities, the number of charging stations is relatively small. These factors make managing electric fleets more challenging than traditional vehicles, prompting the emergence of a new class of VRP problems, namely the electric vehicle routing problem (EVRP). In EVRP, two main aspects need to be considered: 1) customer service sequence and 2) charging route plan for each vehicle. Since these two aspects are closely related, most traditional VRP methods cannot be directly applied to EVRP.

[0003] At present, the methods for solving the problem of electric vehicle delivery tasks mainly include precise algorithms and heuristic algorithms. Precise algorithms can quickly find the optimal solution in small-scale scenarios, but when the number of customer points increases to about 50, the solution time will increase significantly, and it is impossible to find a feasible route within a limited time.

[0004] Heuristic algorithms have been widely used to solve the electric vehicle routing problem (EVRP). Common heuristic algorithms for solving EVRP include the variable neighborhood search (VNS) algorithm, the iterative local search (ILS) method, and the genetic algorithm (GA). However, the VNS algorithm is highly dependent on the initial solution. If the initial solution is not good, even if the neighborhood structure search is changed, it is difficult to find the global optimal solution within the time limit, resulting in delivery delays and affecting customer satisfaction. The ILS method is inefficient in large-scale distribution. Faced with complex tasks across cities and with many customers, it takes a long time to find a better solution, which increases operating costs, reduces vehicle utilization efficiency, and cannot meet the needs of fast and efficient distribution. In the complex and changeable electric vehicle distribution environment, the genetic algorithm lacks clear principles for parameter selection (crossover probability, mutation probability, etc.), and the results of different combinations vary greatly. Practical applications require a large number of experiments to determine the parameters, which increases the difficulty and cost of application.

[0005] Among many heuristic algorithms, the ALNS algorithm shows a strong advantage in solving the electric vehicle delivery problem because it can flexibly adjust the operator weights during the iteration process. However, it also has disadvantages in actual scenarios. For example, if the neighborhood search operator is poorly designed or lacks a reasonable adaptive adjustment mechanism, the ALNS algorithm may fall into a local optimal solution, resulting in low delivery efficiency, increasing the time and energy consumption of terminal delivery, and affecting the quality of delivery services. Summary of the invention

[0006] The present invention aims to address the deficiencies of the above-mentioned prior art and proposes a path optimization method for electric logistics vehicles based on an improved adaptive large-scale neighborhood search algorithm, in order to arrange distribution and refueling routes for a larger scale of terminal distribution demand points in a shorter time, thereby effectively improving distribution efficiency and reducing the total time of terminal distribution of electric logistics vehicles.

[0007] In order to achieve the above-mentioned object, the present invention adopts the following technical solution:

[0008] The electric logistics vehicle terminal delivery path optimization method based on the improved adaptive large neighborhood search algorithm is characterized in that it is applied to a distribution center. , Customers, A distribution service area consisting of charging stations and K electric logistics vehicles; in the distribution service area, Customers are recorded as customer node set , ; The charging station is recorded as the charging station node set , ;in, Indicates Client nodes, represents the j'th charging station node;

[0009] In the delivery service area, the distribution center The customer node set V and the charging station node set F together constitute the node set , the node set The straight line paths between nodes in the ;in, Indicates any Nodes and The path between nodes; record the path The distance is ; Assume that all electric logistics vehicles are used in distribution centers As the starting point and the end point, the electric logistics vehicle is fully charged when it leaves the distribution center S, and when it visits any charging station node The electric logistics vehicle route optimization method is performed according to the following steps:

[0010] Step 1: Construct a terminal delivery path planning model for electric logistics vehicles:

[0011] Step 1.1: Construct the objective function of the electric logistics vehicle terminal delivery path planning model according to formula (1): :

[0012] (1)

[0013] In formula (1), For the Electric logistics vehicles from Node to The driving distance of each node, For the Whether an electric logistics vehicle passes through the path , if passed, let , otherwise, let ;

[0014] Step 1.2: Construct the constraint conditions of the electric logistics vehicle terminal delivery path planning model according to equations (2) to (5):

[0015] (2)

[0016] (3)

[0017] (4)

[0018] (5)

[0019] In formula (2) to formula (5), Indicates whether the kth electric logistics vehicle comes from the distribution center Departure to nodes, Indicates whether the kth electric logistics vehicle comes from the Nodes return to the distribution center ; Indicates Whether the electric logistics vehicle passes the Nodes and The path between nodes , To indicate the Client Nodes The demand for goods, is the maximum load capacity of the electric logistics vehicle, Indicates whether the kth electric logistics vehicle serves the Client Nodes If the service is is 1, otherwise, let is 0; The kth electric logistics vehicle arrives at The power consumption of each node is The kth electric logistics vehicle arrives at The power consumption of each node is Represents the path of the kth electric logistics vehicle from the ith node to the jth node The power consumption rate per unit distance; is the maximum power of the electric logistics vehicle; Indicates that the kth electric logistics vehicle comes from the distribution center Battery level at departure;

[0020] Step 2: Construct an initial path plan based on the electric logistics vehicle terminal delivery path planning model:

[0021] Step 2.1: Set and initialize the current number of iterations , define the maximum number of iterations as ;

[0022] Generate the current individual of the I generation using the nearest neighbor greedy algorithm ;and Is The serial number of the customer node and the serial number of at least one charging station node form an integer code length greater than The current path plan;

[0023] Let the global optimal path solution for ;

[0024] Step 2.2 Calculate using formula (1) The objective function value is recorded as ;

[0025] Step 3: Execute the IALNS algorithm Perform iterative optimization to obtain the best path solution :

[0026] Step 3.1: Construct an operator set:

[0027] Step 3.1.1: Define the set of destruction operators ,in, Remove operators for random clients, Remove the operator for correlation, Remove the operator for the worst charging station; Any destruction operator is denoted as

[0028] Step 3.1.2: Define the repair operator set ,in, is the random repair operator, is the greedy repair operator, is the regret repair operator; let Any repair operator in is denoted as ,y

[0029] Step 3.1.3: and Any operator in , let any nth operator The weight of ; and initialize , ;

[0030] Step 3.2: Use the roulette wheel selection method to select and Select the destruction operator and repair operator under the I iteration and form the operator pair under the I iteration ;

[0031] Using the destruction operator under the Ith iteration right After the destruction, the path plan after the destruction in the Ith iteration is obtained And the set of nodes removed in the Ith iteration ;

[0032] according to , using the repair operator of the Ith iteration right After repair, until Until it is empty, the candidate path solution under the Ith iteration is obtained ;

[0033] Step 3.3: Determine whether to accept the candidate path solution under the Ith iteration , thereby updating ;

[0034] Step 3.4: Dynamically adjust the weight of the nth operator ;

[0035] Step 3.5, after assigning I+1 to I, if , then directly output the best path solution , otherwise, return to step 3.2 and execute sequentially.

[0036] The method for optimizing the terminal delivery path of an electric logistics vehicle based on an improved adaptive large neighborhood search algorithm described in the present invention is also characterized in that any destruction operator in step 3.1.1 For the current path plan The destruction process is as follows:

[0037] Step 3.1.1.1: Define the random customer removal operator Is randomly removed In client nodes, thus obtaining the path plan after destruction in the Ith iteration , and remove the client nodes are added to the list of removed nodes in the Ith iteration ; ;

[0038] Step 3.1.1.2: Define the correlation removal operator It starts with Randomly select a client node , and change it from the path After removing, traverse and calculate any other m-th customer node , and use formula (1) to calculate and The correlation ; thus will be with The client node with the highest correlation is removed, and the destroyed path is obtained in the Ith iteration. and the path plan after destruction in the Ith iteration , add the removed customer node with the highest correlation to the removed node list in the Ith iteration ; ;

[0039] (1)

[0040] In formula (1), A and B represent the distance relevance weight and demand relevance weight respectively. Represents the client node Driving distance to distribution center S; express Driving distance to distribution center S; Represents the mth customer node The demand for goods;

[0041] Step 3.1.1.3: Define the worst charging station removal operator It is the computing charging station node from The damaged path plan after removal The change in the objective function value , and from The charging station node with the largest change value is removed, so as to obtain the damaged path plan under the Ith iteration. , and add the charging station node with the largest change value to the removed node list under the Ith iteration .

[0042] Furthermore, any repair operator in step 3.1.2 right The repair process is as follows:

[0043] Step 3.1.2.1. Define the random repair operator Is a random selection A node in , then from Randomly select a path in the Ith iteration and will insert Thus, the new path under the I-th iteration is obtained , until All nodes in are inserted into the randomly selected path, thus obtaining the candidate path solution under the Ith iteration ;

[0044] Step 3.1.2.2: Define the greedy repair operator is from Randomly select a node , and from Randomly select a path , the calculation will Insert into The change in the objective function at different positions in ,in, Indicates insertion forward The objective function value of Indicates that insert Middle The objective function value at the position, so as to find the position that minimizes the change of the objective function , and Insert into Middle positions and obtain a new path And assign it to ; Repeat the process of node and path selection and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration ;

[0045] Step 3.1.2.3: Define the regret repair operator is from Randomly select a path ,from Select any node , and use formula (2) to calculate Regret value :

[0046] (2)

[0047] In formula (2), Indicates that Insert into any path The change in the objective function after the position where the objective function value increases the least is, For the general The change in the objective function after inserting it to the position that causes the second smallest increase in the objective function value;

[0048] from Remove the node with the largest regret value and insert into The position where the objective function value increases the least in , thereby updating , repeat the process of node and path selection, regret value calculation and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration .

[0049] Furthermore, the roulette wheel selection method in step 3.2 is performed as follows:

[0050] Step 3.2.1: Use formula (6) to calculate the nth operator in the Ith iteration: The proportion of the total weight

[0051] (6)

[0052] Step 3.2.2: Use formula (7) to calculate the value of any operator pair in the Ith iteration: Weight :

[0053] (7)

[0054] In formula (7), express The weight of

[0055] Step 3.2.3: Use formula (8) to calculate the value of any operator pair in the Ith iteration: The proportion of the total weight :

[0056] (8)

[0057] Step 3.2.4: Use formula (9) to calculate the value of any operator pair in the Ith iteration: The cumulative probability :

[0058] (9)

[0059] Step 3.2.5: Select the operator pair for the Ith iteration ;

[0060] Generate a pseudo-random number in the interval [0,1] for the Ith iteration ;

[0061] Traverse the cumulative probability of all operator pairs under the Ith iteration :If we find an operator pair that satisfies the Ith iteration The cumulative probability , then select and As the destruction operator in the Ith iteration and repair operator ;when or season ; represents the x-1th destruction operator, Represents the y-1th repair operator.

[0062] Further, the step 3.3 includes:

[0063] Step 3.3.1: Calculate according to formula (1) The objective function value of and The objective function value of , and then according to formula (10) calculate the acceptance under the I iteration Probability :

[0064] (10)

[0065] In formula (10), represents the current temperature, and , is the initial temperature, is the cooling coefficient, and ;

[0066] Step 3.3.2: Generate a pseudo-random number in the interval [0,1] for the Ith iteration ;

[0067] like or , then Assign to ,Will Assign to Otherwise, keep constant;

[0068] Step 3.3.3, if , then Assign to Otherwise, keep constant.

[0069] Further, the step 3.4 includes:

[0070] Step 3.4.1. Define 4 weight scores ,and ; Record any weight score as ;

[0071] Step 3.4.2: Update the sth weight score :

[0072] If the operator pair selected in the Ith iteration ( ) generated candidate path solutions for , then it is the operator pair selected in the Ith iteration ( ) Assign a score ;

[0073] like , then it is the operator pair selected in the Ith iteration ( ) Assign a score ;

[0074] like ,and , then it is the operator pair selected in the Ith iteration ( ) Assign a score ;

[0075] like ,and , then it is the operator selected in the Ith iteration ( ) Assign a score ;

[0076] Step 3.4.3: Assign to ,in, is the operator decay rate.

[0077] An electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the electric vehicle terminal delivery path planning method, and the processor is configured to execute the program stored in the memory.

[0078] The present invention provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, which is characterized in that when the computer program is run by a processor, the steps of the electric vehicle terminal delivery path planning method are executed.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. This paper proposes an improved adaptive large neighborhood search (IALNS) algorithm framework, which achieves efficient, flexible and powerful path planning capabilities by dynamically adjusting search strategies and parameters, combining diverse destruction and repair operators, and the characteristics of randomness and local search. This method can quickly locate the global optimal path solution, significantly improve delivery efficiency, and thus improve the quality of delivery services.

[0081] 2. The present invention adopts the simulated annealing acceptance criterion to decide whether to adopt a new solution. This criterion allows the algorithm to accept lower-quality solutions to a certain extent, thereby enhancing the robustness and global search capabilities of the algorithm, helping to prevent the path solution from falling into the local optimum, and effectively reducing the total time and energy consumption of distribution.

[0082] 3. During the destruction and repair process, the present invention uses a roulette wheel method to determine the selection probability of the operation operator. This method can dynamically adjust the selection probability according to the weight of each operator, better balance exploration and utilization, and improve the efficiency of distribution path planning.

[0083] In summary, with these advantages, the electric logistics vehicle path optimization method based on the improved adaptive large neighborhood search algorithm demonstrates high efficiency, flexibility and global search capability, and is suitable for solving complex electric logistics vehicle path optimization problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 is a flow chart of the method of the present invention;

[0085] Figure 2 A node distribution diagram of the data set used by the method of the present invention;

[0086] Figure 3 This is a path planning result diagram of the method of the present invention. DETAILED DESCRIPTION

[0087] In this embodiment, a method for planning the path for terminal delivery of electric vehicles based on an improved adaptive large neighborhood search algorithm is provided. First, a neighborhood search operator is customized for the electric vehicle delivery path problem solved by the present invention to better adapt to the structure and characteristics of the problem. Secondly, an automatic parameter selection mechanism is introduced, combined with a roulette wheel selection method, so that the algorithm can automatically screen and adjust parameters to reduce the burden of manual parameter adjustment and improve the robustness and versatility of the algorithm. Finally, combined with the simulated annealing algorithm, the ALNS algorithm is helped to escape from the local optimal solution and accelerate the convergence speed, thereby improving the performance and stability of the algorithm. The current solution is continuously disturbed by the neighborhood search operator (destruction operator and repair operator), the range of the solution is expanded, and the weight scores of each operator are updated after each iteration, so that the model can better utilize the operator and improve the quality of the path plan. Specifically, this planning method is applied to a distribution center consisting of 1 , Customers, The distribution service area consists of charging stations and K electric delivery vehicles. In the distribution service area, Customer Set , ; Charging station set , ;in, Indicates Customers, represents the j'th charging station.

[0088] Distribution centers in the distribution service area The customer set V and the charging station set F together constitute the point set , this point set covers all possible path nodes. The straight line paths between nodes in the ;in, Indicates any Nodes and The straight line path between nodes; The distance is ; Assume that all vehicles are based at the distribution center is the starting point and the end point, and the vehicle is fully charged when it leaves the distribution center S. Fully charged immediately.

[0089] In this embodiment, refer to Figure 1 As shown, the vehicle path planning method is carried out in the following steps:

[0090] Step 1: Construct a terminal delivery path planning model for electric logistics vehicles:

[0091] Step 1.1: Construct the objective function of the electric logistics vehicle terminal delivery path planning model according to formula (1): :

[0092] (1)

[0093] In formula (1), For the Electric logistics vehicles from Node to The driving distance of each node, For the Whether an electric logistics vehicle passes through the path , if passed, let , otherwise, let ;

[0094] Step 1.2: Construct the constraint conditions of the electric logistics vehicle terminal delivery path planning model according to equations (2) to (5):

[0095] (2)

[0096] (3)

[0097] (4)

[0098] (5)

[0099] In formula (2) to formula (5), Indicates whether the kth electric logistics vehicle comes from the distribution center Departure to nodes, Indicates whether the kth electric logistics vehicle comes from the Nodes return to the distribution center ; Indicates Whether the electric logistics vehicle passes the Nodes and The path between nodes , To indicate the Client Nodes The demand for goods, is the maximum load capacity of the electric logistics vehicle, Indicates whether the kth electric logistics vehicle serves the Client Nodes If the service is is 1, otherwise, let is 0; The kth electric logistics vehicle arrives at The power consumption of each node is The kth electric logistics vehicle arrives at The power consumption of each node is Represents the path of the kth electric logistics vehicle from the ith node to the jth node The power consumption rate per unit distance; is the maximum power of the electric logistics vehicle; Indicates that the kth electric logistics vehicle comes from the distribution center Battery level at departure.

[0100] Step 2: Construct an initial path plan based on the electric logistics vehicle terminal delivery path planning model:

[0101] Step 2.1: Set and initialize the current number of iterations , define the maximum number of iterations as ;

[0102] Generate the current individual of the I generation using the nearest neighbor greedy algorithm ;and Is The serial number of the customer node and the serial number of at least one charging station node form an integer code length greater than The current path plan;

[0103] Let the global optimal path solution for ;

[0104] Step 2.2 Calculate using formula (1) The objective function value is recorded as ;

[0105] Step 3: Execute the IALNS algorithm Perform iterative optimization:

[0106] Step 3.1: Construct an operator set:

[0107] Step 3.1.1: Define the set of destruction operators ,in, Remove operators for random clients, Remove the operator for correlation, Remove the operator for the worst charging station; Any operator in

[0108] Step 3.1.1.1: Define the random customer removal operator Is randomly removed In client nodes, thus obtaining the path plan after destruction in the Ith iteration , and remove the client nodes are added to the list of removed nodes in the Ith iteration ; ;

[0109] For example, if the path plan under the Ith iteration Contains two paths and , randomly select the node to be removed as and , then after the random customer removal operator After the operation, the path planning scheme after the destruction is obtained ,in becomes , becomes .

[0110] Step 3.1.1.2: Define the correlation removal operator :

[0111] Defining the Correlation Removal Operator It starts with Randomly select a client node , and change it from the path After removing, traverse and calculate any other m-th customer node , and use formula (1) to calculate and The correlation ; thus will be with The client node with the highest correlation is removed, and the destroyed path is obtained in the Ith iteration. and the path plan after destruction in the Ith iteration , add the removed customer node with the highest correlation to the removed node list in the Ith iteration ; ;

[0112] (1)

[0113] In formula (1), A and B represent the distance relevance weight and demand relevance weight respectively. Represents the client node Driving distance to distribution center S; express Driving distance to distribution center S; Represents the mth customer node The demand for goods;

[0114] For example, if the path plan under the Ith iteration There is a path , where R(1,5) is the largest, then after the correlation removal operator After operation, Remove and , get the path after destruction .

[0115] Step 3.1.1.3: Define the worst charging station removal operator It is the computing charging station node from The damaged path plan after removal The change in the objective function value , and from The charging station node with the largest change value is removed, so as to obtain the damaged path plan under the Ith iteration. , and add the charging station node with the largest removed change value to the list of removed nodes in the Ith iteration .

[0116] Assume the path plan under iteration I In Charging station nodes The original path is , remove the operator after the worst charging station After the operation, the destroyed path will be obtained .

[0117] Step 3.1.2: Define the repair operator set ,in, is the random repair operator, is the greedy repair operator, is the regret repair operator; let Any operator in ,y

[0118] Step 3.1.2.1. Define the random repair operator Is a random selection A node in , then from Randomly select a path in the Ith iteration and will insert Thus, the new path under the I-th iteration is obtained , until All nodes in are inserted into the randomly selected path, thus obtaining the candidate path solution under the Ith iteration ;

[0119] For example, suppose the There are three paths after the destruction in the iteration , , , remove the node list has , , . Randomly select nodes first , and then randomly select a path Insert and get the new path . Then insert them one by one in the same way To the appropriate path, finally get Candidate path solutions under iteration .

[0120] Step 3.1.2.2: Define the greedy repair operator is from Randomly select a node , and from Randomly select a path , the calculation will Insert into The change in the objective function at different positions in ,in, Indicates insertion forward The objective function value of Indicates that insert Middle The objective function value at the position, so as to find the position that minimizes the change of the objective function , and Insert into Middle positions and obtain a new path And assign it to ; Repeat the process of node and path selection and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration ;

[0121] For example, to remove a node from the node list , in the path Insert the calculation objective function at different positions. Suppose we insert it at position 2. Minimum, the node Insert at this location and update the path for And so on, until all nodes are inserted, the candidate path solution for the Ith iteration is obtained. .

[0122] Step 3.1.2.3: Define the regret repair operator is from Randomly select a path ,from Select any node , and use formula (2) to calculate Regret value :

[0123] (2)

[0124] In formula (2), Indicates that Insert into any path The change in the objective function after the position where the objective function value increases the least is, For the general The change in the objective function after inserting it to the position that causes the second smallest increase in the objective function value;

[0125] from Remove the node with the largest regret value and insert into The position where the objective function value increases the least in , thereby updating , repeat the process of node and path selection, regret value calculation and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration ;

[0126] For example, remove the node list has nodes , , , calculate the node On the path The regret value is the largest when the position is 2, and the path is obtained , then calculate the node , The regret value is repeated until all nodes are inserted to obtain the candidate path solution under the Ith iteration. .

[0127] Step 3.1.3: and Any operator in , let any nth operator The weight is ; and initialize , ;

[0128] Step 3.2: Use the roulette wheel selection method to select the path destruction operator and the path repair operator to obtain the operator pair under the Ith iteration: :

[0129] Step 3.2.1: Use formula (6) to calculate the nth operator in the Ith iteration: The proportion of the total weight

[0130] (6)

[0131] Step 3.2.2: Use formula (7) to calculate the value of any operator pair in the Ith iteration: Weight :

[0132] (7)

[0133] Step 3.2.3: Use formula (8) to calculate the value of any operator pair in the Ith iteration: The proportion of the total weight :

[0134] (8)

[0135] Step 3.2.4: Use formula (9) to calculate the value of any operator pair in the Ith iteration: The cumulative probability :

[0136] (9)

[0137] Step 3.2.5: Select the operator pair for the Ith iteration ;

[0138] Generate a pseudo-random number in the interval [0,1] for the Ith iteration ;

[0139] Traverse the cumulative probability of all operator pairs under the Ith iteration :If we find an operator pair that satisfies the Ith iteration The cumulative probability , then select and As the destruction operator in the Ith iteration and repair operator ;when or season ; represents the x-1th destruction operator, Represents the y-1th repair operator.

[0140] Step 3.2.6: Use the removal operator under the Ith iteration right Destroy and obtain the path plan after destruction in the Ith iteration And the set of nodes removed in the Ith iteration ;

[0141] Step 3.2.7, according to , using the repair operator under the Ith iteration right Repair until Until it is empty, thus obtaining the candidate path solution under the Ith iteration ;

[0142] Step 3.2.8: Determine whether to accept the candidate path solution under the Ith iteration ;

[0143] Step 3.2.8.1. Calculate according to formula (1) The objective function value of and The objective function value of , and then according to formula (10) calculate the acceptance under the I iteration Probability :

[0144] (10)

[0145] In formula (10), represents the current temperature, and , is the initial temperature, is the cooling coefficient, and .

[0146] Step 3.2.8.2: Generate a pseudo-random number in the interval [0,1] for the Ith iteration ;

[0147] like or , then Assign to ,Will Assign to Otherwise, keep constant;

[0148] Step 3.2.8.3, if , then Assign to Otherwise, keep constant;

[0149] Step 3.2.9: Dynamically adjust the weight of the nth operator :

[0150] Step 3.2.9.1. Define 4 weight scores ,and ; Record any weight score as ;

[0151] Step 3.2.9.2: Update the sth weight score :

[0152] If the operator pair selected in the Ith iteration ( ) generated candidate path solutions Is the new global optimal path solution , then it is the operator pair selected in the Ith iteration ( ) Assign a score ;

[0153] If the operator pair selected in the Ith iteration ( ) generated candidate path solutions is better than the current path solution, but not the new global optimal path solution, that is, , then it is the operator pair selected in the Ith iteration ( ) Assign a score ;

[0154] If the operator pair selected in the Ith iteration ( ) generated candidate path solutions No better solution was found, i.e. ,and , then it is the operator pair selected in the Ith iteration ( ) Assign a score .

[0155] If the operator pair selected in the Ith iteration ( ) generated candidate path solutions The acceptance condition of the simulated annealing criterion is not met, i.e. ,and , then it is the operator selected in the Ith iteration ( ) Assign a score ;

[0156] Step 3.2.9.3: Based on the obtained weight score , the selected destruction operator in the Ith iteration and repair operator The weights are updated according to the following rules:

[0157] (11)

[0158] In formula (11), is the operator decay rate ( ), which is used to balance the performance of each operator during the search process.

[0159] Step 3.3, after assigning I+1 to I, if , then directly output the best path solution , otherwise, return to step 3.2 and execute sequentially.

[0160] In this embodiment, an electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0161] In this embodiment, a computer-readable storage medium stores a computer program on the computer-readable storage medium, and the computer program executes the steps of the above method when executed by a processor.

[0162] In order to verify the effectiveness and operating efficiency of the method of the present invention, a specific experiment will be used on a real data set for verification and explanation. Experimental environment: Windows 11 operating system, Intel (R) Core (TM) i5-6200 CPU (2.30GHz), 16GB memory. The algorithm code involved is implemented in Python language.

[0163] Experimental data: In order to study the performance of the proposed method, the newly proposed benchmark dataset of the IEEE WCCI2020 EVRP EC competition is used. The benchmark contains a set of small-scale instances and a set of large-scale instances. The small-scale group consists of 7 instances with a maximum of 100 customers. The large-scale group consists of 10 instances with a maximum of 1000 customers (see the literature M.Mavrovouniotis, C. Menelaou, S. Timotheou, Christos Panayiotou, G. Ellinas,M. Polycarpou. Benchmark Set for the IEEE WCCI-2020 Competition onEvolutionary Computation for the Electric Vehicle Routing Problem, TechnicalReport 2020, KIOS CoE, University of Cyprus, Cyprus, March 2020) The detailed information of the benchmark is shown in Table 1:

[0164] surface EVRP Dataset Information

[0165]

[0166] In Table 1, the Name column represents the name of the dataset, the Customer column shows the number of customer nodes in each dataset, the Depot column indicates the number of distribution centers, the Station column indicates the number of charging station nodes, and the Route column represents the maximum number of routes, that is, the upper limit of the number of electric logistics vehicles. According to the data in Table 1, there are 1,000 customer nodes, 1 distribution center, 9 charging stations, and 43 electric logistics vehicles available in the X1001 dataset. The specific distribution is as follows: Figure 2 After applying the electric logistics vehicle path optimization method based on the improved adaptive large neighborhood search algorithm proposed in the present invention on the X1001 dataset, the following is obtained: Figure 3 The path planning scheme shown.

Claims

1. A method for optimizing the terminal delivery path of electric logistics vehicles based on an improved adaptive large neighborhood search algorithm, characterized in that: It is applied to 1 distribution center , Customers, A distribution service area consisting of charging stations and K electric logistics vehicles; In the delivery service area, Customers are recorded as customer node set , ; The charging station is recorded as the charging station node set , ;in, Indicates Client nodes, represents the j'th charging station node; In the delivery service area, the distribution center The customer node set V and the charging station node set F together constitute the node set , the node set The straight line paths between nodes in the ;in, Indicates any Nodes and The path between nodes; record the path The distance is ; Assume that all electric logistics vehicles are used in distribution centers As the starting point and the end point, the electric logistics vehicle is fully charged when it leaves the distribution center S and visits any charging station node The electric logistics vehicle route optimization method is performed according to the following steps: Step 1: Construct a terminal delivery path planning model for electric logistics vehicles: Step 1.1: Construct the objective function of the electric logistics vehicle terminal delivery path planning model according to formula (1): : (1) In formula (1), For the Electric logistics vehicles from Node to The driving distance of each node, For the Whether an electric logistics vehicle passes through the route , if passed, let , otherwise, let ; Step 1.2: Construct the constraint conditions of the electric logistics vehicle terminal delivery path planning model according to equations (2) to (5): (2) (3) (4) (5) In formula (2) to formula (5), Indicates whether the kth electric logistics vehicle comes from the distribution center Departure to nodes, Indicates whether the kth electric logistics vehicle comes from the Nodes return to the distribution center ; Indicates Whether the electric logistics vehicle passes the Nodes and The path between nodes , To indicate the Client Nodes The demand for goods, is the maximum load capacity of the electric logistics vehicle, Indicates whether the kth electric logistics vehicle serves the Client Nodes If the service is is 1, otherwise, let is 0; The kth electric logistics vehicle arrives at The power consumption of each node is The kth electric logistics vehicle arrives at The power consumption of each node is Represents the path of the kth electric logistics vehicle from the ith node to the jth node The power consumption rate per unit distance; is the maximum power of the electric logistics vehicle; Indicates that the kth electric logistics vehicle comes from the distribution center Battery level at departure; Step 2: Construct an initial path plan based on the electric logistics vehicle terminal delivery path planning model: Step 2.1: Set and initialize the current number of iterations , define the maximum number of iterations as ; Generate the current individual of the I generation using the nearest neighbor greedy algorithm ;and Is The serial number of the customer node and the serial number of at least one charging station node form an integer code length greater than The current path plan; Let the global optimal path solution for ; Step 2.2 Calculate using formula (1) The objective function value is recorded as ; Step 3: Execute the IALNS algorithm Perform iterative optimization to obtain the best path solution : Step 3.1: Construct an operator set: Step 3.1.1: Define the set of destruction operators ,in, Remove operators for random clients, Remove the operator for correlation, Remove the operator for the worst charging station; Any destruction operator is denoted as ; Step 3.1.2: Define the repair operator set ,in, is the random repair operator, is the greedy repair operator, is the regret repair operator; let Any repair operator is denoted as ,y ; Step 3.1.3: and Any operator in , let any nth operator The weight of ; and initialize , ; Step 3.2: Use the roulette wheel selection method to select and Select the destruction operator and repair operator under the I iteration and form the operator pair under the I iteration ; Using the destruction operator under the Ith iteration right After the destruction, the path plan after the destruction in the Ith iteration is obtained And the set of nodes removed in the Ith iteration ; according to , using the repair operator of the Ith iteration right After repair, until Until it is empty, the candidate path solution for the Ith iteration is obtained. ; Step 3.3: Determine whether to accept the candidate path solution under the Ith iteration , thereby updating ; Step 3.4: Dynamically adjust the weight of the nth operator ; Step 3.5, after assigning I+1 to I, if , then directly output the best path solution , otherwise, return to step 3.2 and execute sequentially.

2. According to claim 1, a method for optimizing the terminal delivery path of an electric logistics vehicle based on an improved adaptive large neighborhood search algorithm is characterized in that: Any destruction operator in step 3.1.1 For the current path plan The destruction process is as follows: Step 3.1.1.1: Define the random customer removal operator Is randomly removed In client nodes, thus obtaining the path plan after destruction in the Ith iteration , and remove the client nodes are added to the list of removed nodes in the Ith iteration ; ; Step 3.1.1.2: Define the correlation removal operator It starts with Randomly select a client node , and change it from the path After removing, traverse and calculate any other m-th customer node , and use formula (1) to calculate and The correlation ; thus will be with The client node with the highest correlation is removed, and the destroyed path is obtained in the Ith iteration. and the path plan after destruction in the Ith iteration , add the removed customer node with the highest correlation to the removed node list in the Ith iteration ; ; (1) In formula (1), A and B represent the distance relevance weight and demand relevance weight respectively. Represents the client node Driving distance to distribution center S; express Driving distance to distribution center S; Represents the mth customer node The demand for goods; Step 3.1.1.3: Define the worst charging station removal operator It is the computing charging station node from The damaged path plan after removal The change in the objective function value , and from The charging station node with the largest change value is removed, so as to obtain the damaged path plan under the Ith iteration. , and add the charging station node with the largest change value to the removed node list under the Ith iteration .

3. The electric logistics vehicle path optimization method based on an improved adaptive large neighborhood search algorithm according to claim 2 is characterized in that: Any repair operator in step 3.1.2 right The repair process is as follows: Step 3.1.2.

1. Define the random repair operator Is a random selection A node in , then from Randomly select a path in the Ith iteration and will insert Thus, the new path under the I-th iteration is obtained , until All nodes in are inserted into the randomly selected path, thus obtaining the candidate path solution under the Ith iteration ; Step 3.1.2.2: Define the greedy repair operator is from Randomly select a node , and from Randomly select a path , the calculation will Insert into The change in the objective function at different positions in ,in, Indicates insertion forward The objective function value of Indicates that insert Middle The objective function value at the position, so as to find the position that minimizes the change of the objective function , and Insert into Middle positions and obtain a new path And assign it to ; Repeat the process of node and path selection and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration ; Step 3.1.2.3: Define the regret repair operator is from Randomly select a path ,from Select any node , and use formula (2) to calculate Regret value : (2) In formula (2), Indicates that Insert into any path The change in the objective function after the position where the objective function value increases the least is, For the general The change in the objective function after inserting it to the position that causes the second smallest increase in the objective function value; from Remove the node with the largest regret value and insert into The position where the objective function value increases the least in , thereby updating , repeat the process of node and path selection, regret value calculation and insertion until All nodes in are inserted into the path, thus obtaining the candidate path solution under the Ith iteration .

4. The electric logistics vehicle terminal delivery path optimization method based on the improved adaptive large neighborhood search algorithm according to claim 3 is characterized in that: The roulette wheel selection method in step 3.2 is performed as follows: Step 3.2.1: Use formula (6) to calculate the nth operator in the Ith iteration: The proportion of the total weight : (6) Step 3.2.2: Use formula (7) to calculate the value of any operator pair in the Ith iteration: Weight : (7) In formula (7), express The weight of Step 3.2.3: Use formula (8) to calculate the value of any operator pair in the Ith iteration: The proportion of the total weight : (8) Step 3.2.4: Use formula (9) to calculate the value of any operator pair in the Ith iteration: The cumulative probability : (9) Step 3.2.5: Select the operator pair for the Ith iteration ; Generate a pseudo-random number in the interval [0,1] for the Ith iteration ; Traverse the cumulative probability of all operator pairs under the Ith iteration :If we find an operator pair that satisfies the Ith iteration The cumulative probability , then select and As the destruction operator in the Ith iteration and repair operator ;when or season ; represents the x-1th destruction operator, represents the y-1th repair operator.

5. The method for optimizing the terminal delivery path of electric logistics vehicles based on the improved adaptive large neighborhood search algorithm according to claim 4 is characterized in that: The step 3.3 comprises: Step 3.3.1: Calculate according to formula (1) The objective function value of and The objective function value of , and then according to formula (10) calculate the acceptance under the I iteration Probability : (10) In formula (10), represents the current temperature, and , is the initial temperature, is the cooling coefficient, and ; Step 3.3.2: Generate a pseudo-random number in the interval [0,1] for the Ith iteration ; like or , then Assign to ,Will Assign to Otherwise, keep constant; Step 3.3.3, if , then Assign to Otherwise, keep constant.

6. The method for optimizing the terminal delivery path of electric logistics vehicles based on the improved adaptive large neighborhood search algorithm according to claim 5 is characterized in that: The step 3.4 comprises: Step 3.4.

1. Define 4 weight scores ,and ; Record any weight score as ; Step 3.4.2: Update the sth weight score : If the operator pair selected in the Ith iteration ( ) generated candidate path solutions for , then it is the operator pair selected in the Ith iteration ( ) Assign a score ; like , then it is the operator pair selected in the Ith iteration ( ) Assign a score ; like ,and , then it is the operator pair selected in the Ith iteration ( ) Assign a score ; like ,and , then it is the operator selected in the Ith iteration ( ) Assign a score ; Step 3.4.3: Assign to ,in, is the operator decay rate.

7. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the electric vehicle terminal delivery path planning method described in any one of claims 1-6, and the processor is configured to execute the program stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the electric vehicle terminal delivery path planning method described in any one of claims 1 to 6 are executed.

Images