Cold chain distribution path optimization method based on improved whale optimization algorithm

Abstract

The invention discloses a cold chain distribution path optimization method based on an improved whale optimization algorithm, and the method is characterized in that the method comprises the following steps: obtaining the initial information of a demand point and a distribution center, and constructing an information matrix; acquiring historical road condition information around the distribution center and each demand point in a traffic management system; constructing a cold-chain logistics distribution path optimization mathematical model; calculating corresponding transportation time; a search proxy address composed of random variables is converted into a vehicle path matrix through a modeling method; calculating oil consumption of the vehicle during normal driving and idling; and classifying the whale population, and solving the model by using a clustering whale optimization algorithm. And vehicle scheduling and path planning are guided. According to the method, factors such as refrigeration, cargo damage, carbon emission, load capacity and traffic conditions are comprehensively considered, a cold-chain low-carbon logistics distribution path optimization model taking the minimum total cost as a target is established, and an improved whale optimization algorithm is provided for optimization solution.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics distribution path optimization, and in particular to a cold chain distribution path optimization method based on an improved whale optimization algorithm. Background Art

[0002] With economic growth and rising purchasing power, demand for cold-chain foods like fruits, vegetables, and dairy products continues to rise, and so do people's expectations for their quality. While the distribution of general products only requires fixed vehicle and transportation costs, the perishable nature of fresh produce creates fundamental differences in the distribution process.

[0003] The Vehicle Routing Problem (VRP) for cold chain distribution has long been a hot topic in academic research. While some scholars have established early basic models for cold chain logistics and distribution, most studies have failed to consider the impact of vehicle load on fuel consumption. Fuel consumption has a decisive impact on cold chain transportation costs and carbon emissions; ignoring this factor would undoubtedly cause cold chain distribution routing optimization models to deviate from reality. In addition to load, scholars have also considered relevant factors influencing the cold chain distribution process, such as carbon emissions, energy consumption, carbon taxes and limits, and food freshness. Heuristic algorithms such as genetic algorithms and ant colony algorithms have also been widely used to solve these models, but there is room for further improvement in the accuracy of the solutions.

[0004] Optimizing cold chain logistics routing can make transportation more efficient, ensure product quality, reduce cargo damage, and ensure timely delivery. This patent aims to study the vehicle delivery routing optimization problem in cold chain logistics. Taking into account factors such as refrigeration, cargo damage, carbon emissions, load capacity, and traffic conditions, this patent establishes a cold chain low-carbon logistics delivery routing optimization model with the goal of minimizing total cost. It also proposes an improved whale optimization algorithm for optimization and solution. This will help improve delivery efficiency and fresh product sales, thereby reducing costs and increasing corporate profitability. It will also help improve the operational efficiency and economic benefits of the entire logistics system, and enhance service quality and customer satisfaction. Summary of the Invention

[0005] The present invention provides a cold chain distribution path optimization method based on the improved whale optimization algorithm. It comprehensively considers factors such as refrigeration, cargo damage, carbon emissions, load capacity, traffic conditions, etc., establishes a cold chain low-carbon logistics distribution path optimization model with the goal of minimizing total cost, and proposes an improved whale optimization algorithm for optimization and solution.

[0006] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0007] The present invention relates to a cold chain distribution path optimization method based on an improved whale optimization algorithm, which comprises the following steps:

[0008] (1) Obtain initial information on demand points and distribution centers and construct an information matrix;

[0009] (2) Obtain historical traffic information around the distribution center and each demand point in the traffic management system. Based on this information, a data-driven approach is used to predict the traffic conditions around each demand point and distribution center over a period of time. Based on the relationship between distance and travel time, a corresponding traffic congestion function is designed.

[0010] (3) Based on the information matrix mentioned above, the objective function with the minimum total cost as the goal is to set relevant constraints and build a mathematical model for cold chain logistics distribution path optimization;

[0011] (4) Calculate the corresponding transportation time by using the traffic congestion function value and the Euclidean distance between the distribution center and each demand point;

[0012] (5) The search agent address composed of random variables is converted into a vehicle path matrix through modeling methods, making the construction of the model and the calculation of the algorithm easier.

[0013] (6) Dynamically analyze the actual load rate of each route, and combine it with traffic congestion to calculate the fuel consumption of the vehicle during normal driving and idling. This improves the shortcoming of most models that fuel consumption is only related to the transportation distance, so that the vehicle transportation cost and carbon emission cost can be calculated more accurately.

[0014] (7) An improved whale optimization algorithm is proposed. The improved algorithm classifies whale populations through the DBSCAN clustering algorithm, which effectively improves the performance of the algorithm. The clustering whale optimization algorithm (CWOA) is used to solve the model.

[0015] (8) Provide guidance for vehicle scheduling and route planning based on the cold chain logistics vehicle distribution route optimization results output after the optimization is completed.

[0016] Furthermore, the method for obtaining the initial information in step (1) is as follows: constructing a supply-demand relationship matrix based on the storage of the distribution center and the demand of the demand points, and establishing a location information matrix based on the location information of the distribution center and each demand point; and calculating the following data based on the initial information: the transportation distance between each demand point, the transportation distance between each demand point and the distribution center, and the time required for unloading at each demand point; and pre-estimating the number of vehicles required to complete the delivery task based on the total demand of each demand point and the load weight, cargo type and volume of a single cold chain vehicle of the distribution center and the maximum loading volume of the cold chain vehicle;

[0017] Among them, the number of vehicles required to complete the delivery task is estimated in advance as follows

[0018]

[0019] m=max(m1,m2) (1)

[0020] Where m1 is the number of vehicles estimated based on the weight of the goods, m2 is the number of vehicles estimated based on the volume of the goods, and m represents the number of vehicles required to complete the delivery task. represents the rounding function; n represents the total number of demand points; g i represents the demand at demand point i; f represents the volume occupied by goods per unit weight, and its value is determined by the type of goods; Q V It represents the volume of cargo that can be loaded in a single cold chain vehicle compartment; λ1 and λ2 represent vehicle loading coefficients, which are related to the complexity of loading and vehicle constraints, including cargo type, loading method, weight constraint and volume constraint.

[0021] Furthermore, in order to improve the shortcomings of many previous models that do not consider traffic conditions, a traffic condition detection mechanism is introduced. In step (2), the traffic conditions around each demand point and distribution center over a period of time are predicted, including:

[0022] First, according to the size of the overall solution space, three radius ranges are determined: 0 to r1, r1 to r2, and greater than r2; then a traffic congestion detection system is established and a congestion determination function is designed. Based on the historical traffic conditions around each demand point and distribution center, the traffic congestion function values ​​z1, z2, and z3 corresponding to different radius ranges of the demand point and distribution center are obtained. The worse the road conditions, the larger the traffic congestion function value. This traffic congestion function value will affect the time of cold chain delivery and thus affect the total cost. The calculation method is as follows: First, obtain historical road condition information around each distribution center and demand point, and obtain the travel time data of the area corresponding to each time period. Use this data to establish a proxy model, and continuously optimize the proxy model through Bayesian optimization to make it close to the real model. Finally, the optimized model outputs the travel time relationship of the area in the next period of time. Substitute the travel time of each area into the congestion calculation function to obtain its corresponding traffic congestion function value.

[0023] Furthermore, the objective function of the cold chain logistics distribution path optimization mathematical model in step (3) is composed of five costs, including:

[0024] (1) Vehicle fixed costs

[0025] The distribution center uses refrigerated delivery vehicles to provide services to various demand points, which incurs certain fixed costs. These costs mainly include driver wages, vehicle depreciation costs, and maintenance costs. They are related to the number of cold chain vehicles used, but not the distance traveled. The number of delivery vehicles used by the distribution center should be arranged reasonably to avoid waste and excessive costs. If the number of demand points is n, and the number of vehicles predicted by the method described in formula (1) is m, then the vehicle fixed cost is:

[0026]

[0027] Where F is the sum of the driver's salary, vehicle depreciation, and maintenance costs for one vehicle. In this application, this cost is only related to the number of trips and the number of vehicles, and is not related to the distance and time of each trip. The matrix y is used to record which vehicle delivers to each demand point, and n+1 represents the distribution center. The values ​​of the matrix y are as follows:

[0028]

[0029] (2) Transportation costs

[0030] Transportation cost generally refers to the fuel consumption cost during the transportation process. The fuel consumption during transportation is determined by the actual load rate of the cold chain vehicle and the transportation distance. Equation (3) is the calculation process of the actual load rate of vehicle k on the transportation path from demand point i to demand point j.

[0031] U=p ijk / Q (3)

[0032] Where: p ijk represents the actual load of vehicle k on the transport path from demand point i to demand point j; Q represents the maximum load of the vehicle; the specific calculation of the load matrix p will be given below.

[0033] C'=(1+bT idle )d ij [e1+p ijk (e2-e1) / Q] (4)

[0034] Where: d ij represents the transport distance; e1 represents the fuel consumption per unit distance when the vehicle is unloaded; e2 represents the fuel consumption per unit distance when the vehicle is fully loaded; b represents the fuel consumption when the vehicle is idling; T idle Indicates the vehicle idling time. Equation (4) is the total fuel consumption on the transportation path from demand point i to demand point j.

[0035] Formula (4) has given the total fuel consumption on a route. Assuming the unit oil price is a, the total transportation cost is

[0036]

[0037] Where s ijk Used to record vehicle driving trajectory, its specific values ​​are as follows

[0038]

[0039] s is the vehicle path matrix, which is converted from the search agent address composed of random variables. The specific assignment is given below.

[0040] (3) Cargo damage costs

[0041] Cargo damage cost refers to the loss cost of cold chain products caused by the passage of time and temperature changes during transportation and unloading. Its expression is as follows

[0042]

[0043] Where p3 is the unit price of the product, q1 is the deterioration rate of the product during transportation, q2 is the deterioration rate of the product during unloading, 1 to n represent the demand points, n+1 is the distribution center, t ij is the transportation time of the vehicle from point i to point j, T j is the time taken by the vehicle to unload at demand point j. The specific calculation of transportation time t is given below.

[0044] (4) Refrigeration cost

[0045] Refrigeration cost refers to the cost incurred by the cold chain vehicle due to the consumption of refrigerant during the vehicle pre-cooling, transportation and unloading process. pre The assignment method is as follows:

[0046] Frozen goods: When the set temperature is -20℃ and the ambient temperature is >30℃, T pre =2h; 30℃>ambient temperature>20℃, T pre =1.5h; when the ambient temperature is less than 20℃, T pre =1h.

[0047] Refrigerated goods: When the set temperature is 4°C and the ambient temperature is >30°C, T pre =1.5h; 30℃>ambient temperature>20℃, T pre =1h; when the ambient temperature is less than 20℃, T pre =2 / 3h.

[0048] Assuming that the temperature is 25 degrees, during the vehicle pre-cooling process, the refrigeration cost generated by the cold chain vehicle per unit time is p 41 During the transportation process, the refrigeration cost of the cold chain vehicle in transporting products per unit time is p 42During the unloading process, the refrigeration cost incurred by the cold chain vehicle in unloading the goods per unit time is p 43 ; α represents the impact of temperature on cooling energy consumption, so the cooling cost is

[0049]

[0050] (5) Carbon emission costs

[0051] Carbon emission cost refers to the cost of carbon dioxide emissions caused by vehicle energy consumption and refrigerant consumption during transportation. The fuel consumption of a vehicle is related to factors such as the vehicle's driving distance, load capacity, and speed. The more fuel consumed, the more carbon dioxide emissions. Assume p c is the unit carbon tax price, e is the carbon dioxide emission coefficient, E is the energy consumption per unit time of the refrigeration equipment, and the fuel consumption of each path is given by formula (4). Therefore, the total carbon emission cost of vehicle operation and refrigeration equipment is

[0052]

[0053] In summary, the cold chain distribution path optimization model of this scheme aims to minimize the total cost, including vehicle fixed cost, transportation cost, cargo damage cost, refrigeration cost and carbon emission cost. Since most previous researchers did not consider the impact of traffic conditions on costs, we introduced the traffic congestion function value to evaluate the traffic conditions around each demand point and distribution center, so that the vehicle transportation time t is closer to the actual situation and the calculation of cargo damage cost and refrigeration cost is more accurate; in addition, most models only consider the transportation distance but not the cargo capacity for fuel consumption, so we improved the calculation method of fuel consumption, so that we can more accurately calculate vehicle transportation cost and carbon emission cost.

[0054] In addition, the objective function must also satisfy the vehicle load constraint, that is, for any p ijk , there is p ijk <=Q. To facilitate algorithm programming, this application integrates the constraint that vehicles cannot be overloaded into the objective function of the shortest path, and the overload penalty cost is:

[0055]

[0056] In addition, each demand point is delivered by only one vehicle, with the following constraints:

[0057]

[0058] The number of vehicles returning to the distribution center is equal to the m vehicles dispatched for distribution, subject to the following constraints:

[0059]

[0060] The matrix y can be calculated from the matrix s, and the relationship between the two is as follows:

[0061]

[0062] The variables in the matrix x all satisfy the 0-1 constraint, and we have:

[0063] x ijk (x ijk -1)=0 (14)

[0064] The variables in the matrix y all satisfy the 0-1 constraint, and we have:

[0065] y ijk (y ijk -1)=0 (15)

[0066] To sum up, the cold chain distribution path optimization model is:

[0067] minC=C1+C2+C3+C4+C5+C6 (16)

[0068] Constraints:

[0069]

[0070] x ijk (x ijk -1)=0 (14)

[0071] y ijk (y ijk -1)=0 (15)

[0072] Furthermore, in order to make the transportation time in step (4) closer to the actual situation, it is necessary to establish a traffic congestion detection system and design a congestion determination function. By analyzing the traffic conditions around each demand point and distribution center, the traffic congestion function values ​​corresponding to different radius ranges of the demand point and distribution center can be obtained. Introducing the traffic congestion function value can make the vehicle transportation time t closer to the actual situation and make the cost calculation more accurate. The transportation time calculation process based on traffic conditions is as follows: if d is the distance from the starting point to the end point and all transportation vehicles travel at a constant speed, the transportation time t in the mathematical model can be expressed as:

[0073]

[0074] Among them, r1 and r1 represent different radius ranges, and z1, z2, and z3 represent the traffic congestion function values ​​for each radius range.

[0075] Furthermore, in order to simplify the construction of the model and the calculation of the algorithm, in step (5), the search agent address composed of random variables in the optimization algorithm is converted into a vehicle path matrix. In the past, when solving path optimization problems, people mostly used ant colony algorithms and genetic algorithms. This is because the position updates of such algorithms are easy to correspond to the path access sequence. This patent proposes a method that can convert a set of irregular one-dimensional arrays with arbitrary values ​​into a matrix representing the path access sequence, so that algorithms such as whale optimization can more easily solve path optimization problems. The steps are as follows:

[0076] a. If you need to solve the problem of delivering m vehicles to n demand points, then input a one-dimensional array x containing n + m - 1 numbers.

[0077] b. Sort x and store the result in array A. Array x remains unchanged.

[0078] c. Traverse array x and compare it with array A, output the position of the elements in A and store them in array B.

[0079] d. Define an array C of n+m-1 with all zeros, and let B i For the i-th element in B, let h = B i， Traverse array B and do the following:

[0080] B i =B i +C h

[0081] C h =C h +1 (18)

[0082] This removes duplicate elements from B, making it strictly arranged from 1 to n+m-1.

[0083] e. Change the values ​​of elements in array B that are greater than n+1 to n+1 to separate the front and back elements, representing the next transport vehicle. Elements that are less than n+1 are demand points, thus dividing B into m parts, each representing a vehicle. The order from front to back is the order in which the demand points are visited.

[0084] f. Traverse matrix B and enter its information into the three-dimensional path access sequence matrix s.

[0085] Furthermore, step (6) dynamically analyzes the actual load rate of each route and calculates the fuel consumption of the vehicle when idling, taking into account the traffic congestion situation. This improves the shortcoming of most models that fuel consumption is only related to the transportation distance, thereby being able to more accurately calculate the vehicle transportation cost and carbon emission cost. Assume that the vehicle's idling fuel consumption per unit time is b, and the idling time is T idle It is related to the congestion coefficient r, and its specific value is:

[0086]

[0087] The calculation steps of load matrix p, actual load rate U and fuel consumption are as follows:

[0088] a. Traverse array B in reverse order and refer to the access order matrix s, add the demand of each demand point to the corresponding path in turn and record it in the three-dimensional matrix p.

[0089] b. Calculate the actual load rate according to formula (3):

[0090] U=p ijk / Q (3)

[0091] c. Calculate fuel consumption according to formula (4):

[0092] C'=(1+bT idle )d ij [e1+p ijk (e2-e1) / Q] (4)

[0093] Substituting traffic conditions, actual load rate and fuel consumption into the constructed mathematical model for cold chain logistics distribution path optimization can more accurately calculate vehicle transportation costs and carbon emission costs, making the model closer to actual conditions.

[0094] Furthermore, step (7) uses the DBSCAN clustering algorithm to classify the whale population. As the iteration proceeds, the size of the clustered population will increase, so that the algorithm gradually shifts from global search to local utilization. At the same time, the cluster limit maxpopulation is set to ensure that there are at least m individuals for global search, thus solving the problem that the algorithm is prone to falling into local optimality in the late iteration. The main strategies for improving the whale optimization algorithm are as follows:

[0095] Population partitioning strategy: DBSCAN clustering

[0096] i = dbscan(X(t), e, min)

[0097] k=i(X * (t)) (19)

[0098] Where X(t) is the position matrix of the population members at the tth iteration, X * (t) represents the position of the whale with the best fitness at the tth iteration. dbscan is a pre-set function that clusters the matrix X(t) according to the parameters e and min. The parameter e represents the neighborhood radius and min represents the minimum number of neighborhood points. After clustering, the classification index of each whale individual will be recorded in the one-dimensional array i. The parameter k is used to record the category of the best individual. In this algorithm, the parameter settings are as follows:

[0099]

[0100] min=n / 30

[0101] Where ub(j) is the upper limit of the value given for the current dimension, lb(j) is the lower limit of the value given for the current dimension, and n is the number of whales.

[0102] Clustering restrictions:

[0103] As the algorithm iterates, whale individuals will cluster under the influence of the movement strategy described in equation (2). If no clustering limit is set, all individuals will become clustered individuals in the middle and late stages of the algorithm, causing the algorithm to almost lose its global search ability and easily fall into local optimality. Therefore, the clustering limit maxpopulation and parameter m are set. If the individual satisfies: i(X(t)) = k && m <= maxpopulation, the individual will update its position according to the spiral approach in the whale optimization algorithm. Otherwise, the individual will randomly approach other individuals. The parameter m is initially 1, and its value increases by 1 each time the clustered individual updates its position.

[0104] Furthermore, the specific results of step (8) cold chain logistics vehicle delivery route optimization are as follows:

[0105] a. First, obtain the initial information matrix and then establish a traffic congestion detection system;

[0106] b. Input the traffic conditions around each demand point and distribution center to obtain the corresponding traffic congestion function value;

[0107] c. Substitute the traffic congestion function value into formula (17) to calculate the transportation time;

[0108] d. Generate the path access sequence through step (5);

[0109] e. Calculate the load matrix and fuel consumption of each path by step (6) and by accessing the path in sequence;

[0110] f. Substitute the above data into the cold chain logistics distribution path optimization mathematical model;

[0111] g. Set this cold chain logistics distribution path optimization mathematical model as the objective function and initialize the optimization algorithm using the relevant information provided at the beginning, including the scope of the solution space, the number of search agents, the initial population size, and the number of iterations. The fitness of each individual is then calculated, and the whale population is clustered. The initial cluster population is formed with the optimal individual as the center. Non-clustered members and clustered members are then updated according to different movement strategies, and the clusters are re-clustered after each update. The fitness of the parent and offspring individuals is compared, and the individuals with better fitness are selected as the new offspring. The fitness of the individuals in the new population is then compared, and the optimal individual is selected as the new leader to proceed to the next generation of iterations.

[0112] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0113] 1. The cold chain logistics distribution path optimization method based on the improved whale optimization algorithm involved in the present invention is based on the transportation conditions of the goods, and takes into account the vehicle load capacity and the traffic conditions around the demand point, and constructs a site selection model with the goal of minimizing the total logistics cost, making the selection and optimization of the cold chain logistics distribution path more realistic and reasonable, and helping to select the optimal path for the distribution vehicle, thereby reducing the cost of the entire logistics process.

[0114] 2. The present invention involves a cold chain logistics distribution path optimization method based on an improved whale optimization algorithm, which proposes an improved whale optimization algorithm. Compared with traditional optimization algorithms, the improved whale optimization algorithm has a certain improvement in high-dimensional accuracy. For complex functions, the algorithm can effectively avoid falling into local optimality, and its convergence speed is faster than other optimization algorithms. Applying it to the cold chain logistics distribution path optimization problem can effectively avoid the shortcomings of slow convergence speed and low convergence accuracy, thereby optimizing the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 This is a schematic diagram of the overall process of the cold chain distribution path optimization method based on the improved whale optimization algorithm;

[0116] Figure 2 This is a flowchart of the algorithm for cold chain distribution path optimization based on the improved whale optimization algorithm;

[0117] Figure 3 This is the test result of the improved whale optimization algorithm proposed in this invention;

[0118] Figure 4 Iteration diagrams of various algorithms used to test examples of the present invention;

[0119] Figure 5 The distribution path optimization solution for the cold chain logistics instance obtained by the traditional whale optimization algorithm.

[0120] Figure 6 This is the cold chain logistics distribution path optimization solution obtained by the present invention based on the improved whale optimization algorithm. DETAILED DESCRIPTION

[0121] In order to make the purpose, technical solutions and advantages of the present invention more clear and distinct, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0122] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is in no way intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0123] Refer to the attached Figures 1-6 As shown, the present invention relates to a cold chain distribution path optimization method based on an improved whale optimization algorithm, which comprises the following steps:

[0124] (1) Obtain initial information on demand points and distribution centers and construct an information matrix;

[0125] (2) Obtain historical traffic information around the distribution center and each demand point in the traffic management system. Based on this information, a data-driven approach is used to predict the traffic conditions around each demand point and distribution center over a period of time. Based on the relationship between distance and travel time, a corresponding traffic congestion function is designed.

[0126] (3) Based on the information matrix mentioned above, the objective function with the minimum total cost as the goal is to set relevant constraints and build a mathematical model for cold chain logistics distribution path optimization;

[0127] (4) Calculate the corresponding transportation time by using the traffic congestion function value and the Euclidean distance between the distribution center and each demand point;

[0128] (5) The search agent address composed of random variables is converted into a vehicle path matrix through modeling methods, making the construction of the model and the calculation of the algorithm easier.

[0129] (6) Dynamically analyze the actual load rate of each route, and combine it with traffic congestion to calculate the fuel consumption of the vehicle during normal driving and idling. This improves the shortcoming of most models that fuel consumption is only related to the transportation distance, so that the vehicle transportation cost and carbon emission cost can be calculated more accurately.

[0130] (7) An improved whale optimization algorithm is proposed. The improved algorithm classifies whale populations through the DBSCAN clustering algorithm, which effectively improves the performance of the algorithm. The clustering whale optimization algorithm (CWOA) is used to solve the model.

[0131] (8) Provide guidance for vehicle scheduling and route planning based on the cold chain logistics vehicle distribution route optimization results output after the optimization is completed.

[0132] First, for the mathematical modeling of this cold chain logistics distribution path optimization, the following assumptions are made:

[0133] a. The demand volume, geographical location, and time window of each demand point are known, and the demanded goods must be delivered in one go;

[0134] b. The fixed cost of vehicle dispatch is known and is only related to the number of delivery vehicles;

[0135] c. The distance from the distribution center to each demand point is known;

[0136] d. The types of cold chain products are limited;

[0137] e. The unit price of the commodity and the price of oil are known;

[0138] f. All vehicles are of the same model and have the same load capacity;

[0139] The initial information obtained in step (1) includes a supply-demand relationship matrix constructed based on the storage capacity of the distribution center and the demand of the demand points, and a location information matrix constructed based on the location information of the distribution center and each demand point. Based on the initial information, the following data are calculated: the transportation distance between each demand point, the transportation distance between each demand point and the distribution center, and the time required for unloading at each demand point. In this example, a distribution center provides distribution services to 20 demand points. The distribution center is located at (50, 50). The fixed cost of a vehicle is 1500 per trip. The maximum load capacity Q of the distribution vehicle is 7, and the maximum loading volume Q is 1. V=8, the fuel cost per unit distance traveled when empty is e1 = 0.56, the fuel cost per unit distance traveled when fully loaded is e2 = 1.85, and the idling fuel consumption b = 0.15. Suppose m vehicles depart from the selected distribution center at a speed of 50 km / h. The unloading time is proportional to the amount of goods unloaded at the distribution center, with a proportionality coefficient of 1 / 15. The unit price of cold chain products is 4,000 yuan / ton, the deterioration rate during transportation is 0.02, and the deterioration rate during unloading is 0.05. The refrigeration cost for pre-cooling the cold chain vehicle is 10 yuan / hour, the refrigeration cost for transporting goods is 20 yuan / hour, and the refrigeration cost for unloading goods is 25 yuan / hour. The vehicle fuel consumption is 0.25 L / ton, the energy consumption of the refrigeration equipment is 0.003 L / t.km, the unit carbon tax price is 25 yuan / kg, and the carbon dioxide emission coefficient is 2.6 kg / L. The relevant data of the demand point are shown in Table 1.

[0140] Table 1 Demand point related data

[0141]

[0142]

[0143] Estimate the number of vehicles needed to complete the delivery task based on the total demand at each demand point, the load weight of a single cold chain vehicle at the distribution center, the type and volume of goods, and the maximum loading volume of the cold chain vehicle. The method for estimating the number of vehicles is as follows:

[0144]

[0145] m=max(m1,m2) (1)

[0146] Where m1 is the number of vehicles estimated based on the weight of the goods, m2 is the number of vehicles estimated based on the volume of the goods, and m represents the number of vehicles required to complete the delivery task. represents the rounding function; n represents the total number of demand points; g i represents the demand at demand point i; f represents the volume occupied by goods per unit weight, and its value is determined by the type of goods; Q V It represents the volume of cargo that can be loaded in a single cold chain vehicle compartment; λ1 and λ2 represent vehicle loading coefficients, which are related to the complexity of loading and vehicle constraints, including cargo type, loading method, weight constraint and volume constraint.

[0147] In this example, let the vehicle loading coefficient λ1=1, λ2=1, and the unit cargo volume f=0.5, then Number of delivery vehicles m = 6.

[0148] In step (2), historical traffic information around each distribution center and demand point is obtained from the traffic management system, and the data-driven optimization method is used to predict the traffic conditions around it for a long period of time based on historical data. Since the location of the distribution center is not given in advance in the center location problem and the path is also uncertain, it is difficult to determine the congestion situation of each path;

[0149] This solution first determines three radius ranges (0 to r1, r1 to r2, and greater than r2) based on the size of the overall solution space. Then, a traffic congestion detection system is established. The traffic conditions around each demand point and distribution center are input and substituted into the corresponding congestion calculation function. Finally, the traffic congestion function values ​​z1, z2, and z3 for the three circular areas corresponding to the three radius ranges of each point are obtained (the worse the road conditions, the greater the traffic congestion function value).

[0150] This traffic congestion function value affects the time of cold chain delivery, and thus the total cost. Its calculation method is as follows: First, obtain historical road condition information around each distribution center and demand point, and obtain data on the area-travel time corresponding to each time period. Use this data to establish a proxy model. Through Bayesian optimization, the proxy model is continuously optimized to make it closer to the true model. Finally, the optimized model outputs the area-travel time relationship for the next period of time. Substituting the travel time of each area within the three radius into the congestion calculation function, the corresponding traffic congestion function value can be obtained. In this example, set r1 = 5km and r2 = 15km. The traffic congestion function values ​​of each demand point and distribution point are shown in Table 2.

[0151] Table 2 Traffic congestion function values ​​at each point

[0152]

[0153] The method for constructing the mathematical model for cold chain logistics distribution path optimization in step (3) is as follows. The objective function of the cold chain logistics distribution path optimization solution consists of five costs, specifically including:

[0154] (1) Vehicle fixed costs

[0155] The distribution center uses refrigerated delivery vehicles to provide services to various demand points, which incurs certain fixed costs. These costs mainly include driver wages, vehicle depreciation costs, and maintenance costs. They are related to the number of cold chain vehicles used, but not to the driving distance. The number of delivery vehicles used by the distribution center should be arranged reasonably to avoid waste and excessive costs. If the number of demand points is n, and the number of vehicles predicted by the method described in formula (1) is m, then the vehicle fixed cost is

[0156]

[0157] Where F is the sum of the driver's salary, vehicle depreciation, and maintenance costs for one vehicle. In this patent, this cost is only related to the number of trips and the number of vehicles, and is not related to the distance and time of each trip. The matrix y is used to record which vehicle delivers to each demand point, and n+1 represents the distribution center. The matrix y values ​​are as follows

[0158]

[0159] (2) Transportation costs

[0160] Transportation cost generally refers to the fuel consumption cost during the transportation process. The fuel consumption during transportation is determined by the actual load rate of the cold chain vehicle and the transportation distance. Equation (3) is the calculation process of the actual load rate of vehicle k on the transportation path from demand point i to demand point j.

[0161] U=p ijk / Q (3)

[0162] Where: p ijk represents the actual load of vehicle k on the transport path from demand point i to demand point j; Q represents the maximum load of the vehicle; the specific calculation of the load matrix p will be given later.

[0163] C'=(1+bT idle )d ij [e1+p ijk (e2-e1) / Q] (4)

[0164] Where: d ij represents the transport distance; e1 represents the fuel consumption per unit distance when the vehicle is unloaded; e2 represents the fuel consumption per unit distance when the vehicle is fully loaded; b represents the fuel consumption when the vehicle is idling; T idle Indicates the vehicle idling time. Equation (4) is the total fuel consumption on the transportation path from demand point i to demand point j.

[0165] Formula (4) has given the total fuel consumption on a route. Assuming the unit oil price is a, the total transportation cost is

[0166]

[0167] Where s ijk Used to record vehicle driving trajectory, its specific values ​​are as follows

[0168]

[0169] s is the vehicle path matrix, which is converted from the search agent address composed of random variables. The specific assignment will be given later.

[0170] (3) Cargo damage costs

[0171] Cargo damage cost refers to the loss cost of cold chain products caused by the passage of time and temperature changes during transportation and unloading. Its expression is as follows

[0172]

[0173] Where p3 is the unit price of the product, q1 is the deterioration rate of the product during transportation, q2 is the deterioration rate of the product during unloading, 1 to n represent the demand points, n+1 is the distribution center, t ij is the transportation time of the vehicle from point i to point j, T j is the time taken by the vehicle to unload at demand point j. The specific calculation of transportation time t will be given later.

[0174] (4) Refrigeration cost

[0175] Refrigeration cost refers to the cost incurred by the cold chain vehicle due to the consumption of refrigerant during the vehicle pre-cooling, transportation and unloading process. pre The assignment method is as follows:

[0176] Frozen goods: When the set temperature is -20℃ and the ambient temperature is >30℃, T pre =2h; 30℃>ambient temperature>20℃, T pre =1.5h; when the ambient temperature is less than 20℃, T pre =1h.

[0177] Refrigerated goods: When the set temperature is 4°C and the ambient temperature is >30°C, T pre =1.5h; 30℃>ambient temperature>20℃, T pre =1h; when the ambient temperature is less than 20℃, T pre =2 / 3h.

[0178] Assuming that the temperature is 25 degrees, during the vehicle pre-cooling process, the refrigeration cost generated by the cold chain vehicle per unit time is p 41 During the transportation process, the refrigeration cost of the cold chain vehicle in transporting products per unit time is p 42 During the unloading process, the refrigeration cost incurred by the cold chain vehicle in unloading the goods per unit time is p 43 ; α represents the impact of temperature on cooling energy consumption, so the cooling cost is

[0179]

[0180] (5) Carbon emission costs

[0181] Carbon emission cost refers to the cost of carbon dioxide emissions caused by vehicle energy consumption and refrigerant consumption during transportation. The fuel consumption of a vehicle is related to factors such as the vehicle's driving distance, load capacity, and speed. The more fuel consumed, the more carbon dioxide emissions. Assume p c is the unit carbon tax price, e is the carbon dioxide emission coefficient, E is the energy consumption per unit time of the refrigeration equipment, and the fuel consumption of each path is given by formula (4). Therefore, the total carbon emission cost of vehicle operation and refrigeration equipment is

[0182]

[0183] In summary, the cold chain distribution path optimization model of this scheme aims to minimize the total cost, including vehicle fixed cost, transportation cost, cargo damage cost, refrigeration cost and carbon emission cost. Since most previous researchers did not consider the impact of traffic conditions on costs, we introduced the traffic congestion function value to evaluate the traffic conditions around each demand point and distribution center, so that the vehicle transportation time t is closer to the actual situation and the calculation of cargo damage cost and refrigeration cost is more accurate; in addition, most models only consider the transportation distance but not the cargo capacity for fuel consumption, so we improved the calculation method of fuel consumption, so that we can more accurately calculate vehicle transportation cost and carbon emission cost.

[0184] In addition, the objective function must also satisfy the vehicle load constraint, that is, for any p ijk , there is p ijk <=Q. To facilitate algorithm programming, this application integrates the constraint that the vehicle cannot be overloaded into the objective function of the shortest path, and the overload penalty cost is

[0185]

[0186] In addition, each demand point is delivered by only one vehicle, with the following constraints:

[0187]

[0188] The number of vehicles returning to the distribution center is equal to the m vehicles dispatched for distribution, subject to the following constraints:

[0189]

[0190] The matrix y can be calculated from the matrix s, and the relationship between the two is as follows:

[0191]

[0192] The variables in the matrix x all satisfy the 0-1 constraint, and we have:

[0193] x ijk (x ijk -1)=0 (14)

[0194] The variables in the matrix y all satisfy the 0-1 constraint, and we have:

[0195] y ijk (y ijk -1)=0 (15)

[0196] To sum up, the cold chain distribution path optimization model is:

[0197] minC=C1+C2+C3+C4+C5+C6 (16)

[0198] Constraints:

[0199]

[0200] x ijk (x ijk -1)=0 (14)

[0201] y ijk (y ijk -1)=0 (15)

[0202] Among them, in order to make the transportation time in the step (4) closer to the actual situation, it is necessary to establish a traffic congestion detection system and design a congestion determination function. Through the traffic conditions around each demand point and distribution center, the traffic congestion function values ​​corresponding to different radius ranges of the demand point and distribution center are obtained. Introducing the traffic congestion function value can make the vehicle transportation time t closer to the actual situation and make the cost calculation more accurate.

[0203] The transportation time calculation process based on traffic conditions is as follows. If d is the distance from the starting point to the end point and all transport vehicles travel at a constant speed, the transportation time t in the mathematical model can be expressed as:

[0204]

[0205] In order to simplify the construction of the model and the calculation of the algorithm, the search agent address composed of random variables in the optimization algorithm is converted into a vehicle path matrix in step (5). In the past, when solving path optimization problems, people mostly used ant colony algorithms and genetic algorithms. This is because the position updates of such algorithms are easy to correspond to the path access sequence. This patent proposes a method that can convert a set of irregular one-dimensional arrays with arbitrary values ​​into a matrix representing the path access sequence, so that algorithms such as whale optimization can more easily solve path optimization problems. The steps are as follows:

[0206] a. To solve a delivery problem involving m vehicles to n demand points, input a one-dimensional array x containing n+m-1 numbers. In this example, n = 20 and m = 6, so assume the random whales x = (25, 43, 8, 66, 5, 99, 25, 34, 67, 66, 12, 83, 91, 69, 78, 43, 25, 57, 34, 82, 53, 49, 21, 5, 93).

[0207] b. Sort x and store the result in array A. Array x remains unchanged.

[0208] Then A=(5, 5, 8, 12, 21, 25, 25, 25, 34, 34, 43, 43, 49, 53, 57, 66, 66, 67, 69, 78, 82, 83, 91, 93, 99).

[0209] c. Traverse array x and compare it with array A, output the position of the elements in A and store them in array B.

[0210] Then B = (6, 11, 3, 16, 1, 25, 6, 9, 18, 16, 4, 22, 23, 19, 20, 11, 6, 15, 9, 21, 14, 13, 5, 1, 24).

[0211] d. Define an array C of n+m-1 with all zeros, and let B i For the i-th element in B, let h = B i， Traverse array B and do the following:

[0212] B i =B i +C h

[0213] C h =C h +1 (18)

[0214] This removes duplicate elements from B, making it strictly arranged from 1 to n+m-1.

[0215] At this time, B = (6, 11, 3, 16, 1, 25, 7, 9, 18, 17, 4, 22, 23, 19, 20, 12, 8, 15, 10, 21, 14, 13, 5, 2, 24).

[0216] e. Change the values ​​of elements in array B greater than n+1 to n+1 to separate the front and back elements, representing the next delivery vehicle. Elements less than n+1 represent demand points, thus dividing B into m sections, each representing a vehicle. The order of demand point access from front to back is the order in which the demand points are accessed. Thus, array B represents the access order for the six vehicles as follows: Vehicle 1: 0-6-11-3-16-1-0; Vehicle 2: 0-7-9-18-17-4-0; Vehicle 3: 0-0; Vehicle 4: 0-19-20-12-8-15-10-0; Vehicle 5: 14-13-5-2-0; Vehicle 6: 0-0. Vehicles 3 and 6 are not being delivered.

[0217] f. Traverse matrix B and enter its information into the three-dimensional path access sequence matrix s.

[0218] Among them, step (6) dynamically analyzes the actual load rate of each path and calculates the fuel consumption of the vehicle when idling in combination with the traffic congestion situation. This improves the shortcoming of most models that fuel consumption is only related to the transportation distance, thereby being able to more accurately calculate the vehicle transportation cost and carbon emission cost. Assume that the vehicle's idling fuel consumption per unit time is b, and the idling time is T idle It is related to the congestion coefficient r, and its specific value is:

[0219]

[0220] The calculation steps of load matrix p, actual load rate U and fuel consumption are as follows:

[0221] a. Traverse array B in reverse order and refer to the access order matrix s, add the demand of each demand point to the corresponding path in turn and record it in the three-dimensional matrix p.

[0222] b. Calculate the actual load rate according to formula (3):

[0223] U=p ijk / Q (3)

[0224] c. Calculate fuel consumption according to formula (4):

[0225] C'=(1+bT idle )d ij [e1+p ijk (e2-e1) / Q] (4)

[0226] Substituting traffic conditions, actual load rate and fuel consumption into the constructed mathematical model for cold chain logistics distribution path optimization can more accurately calculate vehicle transportation costs and carbon emission costs, making the model closer to actual conditions.

[0227] Among them, step (7) uses the DBSCAN clustering algorithm to classify the whale population. As the iteration proceeds, the size of the clustered population will increase, so that the algorithm gradually shifts from global search to local utilization; at the same time, the cluster limit maxpopulation is set to ensure that there are at least m individuals for global search, thus solving the problem that the algorithm is prone to falling into local optimality in the late iteration. The main strategies for improving the whale optimization algorithm are as follows:

[0228] (1) Population division strategy: DBSCAN clustering

[0229] i = dbscan(X(t), e, min)

[0230] k=i(X * (t)) (19)

[0231] Where X(t) is the position matrix of the population members at the tth iteration, X * (t) represents the position of the whale with the best fitness at the tth iteration. dbscan is a pre-set function that clusters the matrix X(t) according to the parameters e and min. The parameter e represents the neighborhood radius and min represents the minimum number of neighborhood points. After clustering, the classification index of each whale individual will be recorded in the one-dimensional array i. The parameter k is used to record the category of the best individual. In this algorithm, the parameter settings are as follows:

[0232]

[0233] min=n / 30

[0234] Where ub(j) is the upper limit of the value given for the current dimension, lb(j) is the lower limit of the value given for the current dimension, and n is the number of whales.

[0235] (2) Mobile Strategy

[0236] In the improved whale optimization algorithm described in this application, cluster members use a spiral approach strategy, and cluster members update their positions according to this movement strategy. The characteristic of this movement strategy is that as the number of iterations increases, the trajectory of the spiral gradually approaches the leader, which enhances the algorithm's local utilization ability and helps the algorithm converge to the optimal value. The specific movement strategy is as follows:

[0237] When the probability p<0.5

[0238] D L =|C*X * (t)-X(t)|

[0239] X(t+1)=X * (t)-A*D L (twenty one)

[0240] When the probability p>=0.5

[0241] D L2 =|C*X * (t)-X(t)|

[0242] X(t+1)=D L2 *e bl *cos(2πl)+X * (t) (22)

[0243] Where X * (t) is the position of the leader in the jth dimension at the tth iteration, X(t) is the position of the population member in the jth dimension at the tth iteration, X(t+1) is the position of the population member in the jth dimension at the t+1th iteration, C, A, b, l are related parameters, and the function f(x,y) is assumed to take values ​​in the interval [x,y], where:

[0244] A=2*a*r1-a;

[0245] C = 2*r2;

[0246] b=1;

[0247] l=(a2-1)*f(0,1)+1;

[0248] The non-cluster member movement strategy is to move closer to any individual:

[0249] D L =|C*X'(t)-X(t)|

[0250] X(t+1)=X'(t)-A*D L (twenty three)

[0251] Where X'(t) is the position of a randomly selected individual in the jth dimension at the tth iteration.

[0252] (3) Clustering restriction strategy

[0253] As the algorithm iterates, whale individuals will cluster under the action of the movement strategy described in formula (2). If no clustering limit is set, all individuals will become clustered individuals in the middle and late stages of the algorithm, causing the algorithm to almost lose its global search ability and easily fall into local optimality. Therefore, the clustering limit maxpopulation and parameter m are set. If the individual satisfies the following formula:

[0254] i(X(t))=k && m<=maxpopulation (24)

[0255] Then the individual updates its position according to equations (21) and (22), otherwise the individual updates its position according to equation (23). In the equation, the initial value of parameter m is 1, and its value increases by 1 each time the clustered individual updates its position.

[0256] The algorithm was tested. The test function is shown in Table 3, the test results are shown in Table 4, and the test iteration process is shown in the appendix. Figure 3 As shown in the figure, compared with traditional optimization algorithms, the improved whale optimization algorithm (CWOA) has improved the accuracy in high dimensions. For complex functions, the algorithm can effectively avoid falling into local optimality, and its convergence speed is faster than other optimization algorithms. Applying it to the location selection problem of cold chain logistics distribution centers can effectively avoid the shortcomings of slow convergence speed and low convergence accuracy.

[0257] Table 3 Algorithm test function

[0258]

[0259] Table 4 Algorithm test results

[0260]

[0261] The specific results of step (8) cold chain logistics vehicle delivery route optimization are as follows:

[0262] a. First, obtain the initial information matrix and then establish a traffic congestion detection system;

[0263] b. Input the traffic conditions around each demand point and distribution center to obtain the corresponding traffic congestion function value;

[0264] c. Substitute the traffic congestion function value into formula (17) to calculate the transportation time;

[0265] d. Generate the path access sequence through step (5);

[0266] e. Calculate the load matrix and fuel consumption of each path by step (6) and by accessing the path in sequence;

[0267] f. Substitute the above data into the cold chain logistics distribution path optimization mathematical model;

[0268] g. Set the cold chain logistics distribution path optimization mathematical model as the objective function, and use the relevant information given at the beginning to initialize the optimization algorithm, including the solution space range, the number of search agents, the initial population size and the number of iterations. Then calculate the fitness of each individual, cluster the whale population, and form an initial cluster population with the best individual as the center. After that, the non-cluster members and cluster members update their positions according to different movement strategies, and re-cluster after each update. Compare the fitness of the parent individual and the offspring individual, select the individual with better fitness as the new offspring, compare the fitness of the new population individuals, select the best individual as the new leader and iterate for the next generation. The iteration of each algorithm in this example is as follows: Figure 4 The example output data is shown in Table 5. CWOA not only converges faster than other traditional algorithms, but also has higher accuracy than other algorithms. It can be well applied in the location selection problem of cold chain logistics distribution centers.

[0269] Table 5 Example test results

[0270]

[0271] In this example, the example data is substituted into the improved algorithm, and the optimal path is finally obtained as shown in Table 5. The minimum total cost obtained by CWOA is 17139.2243, which is better than the 18054.5785 of the traditional whale optimization algorithm WOA. A total of 6 vehicles are required to complete the delivery. The paths of each vehicle are shown in the table above. The schematic diagrams of the solutions obtained by WOA and CWOA are shown in Figure 5. Figure 5 、 Figure 6 shown.

[0272] From the above examples, it can be seen that the application of the improved whale optimization algorithm in the present invention to the cold chain logistics distribution path optimization problem is scientific. Through this algorithm, the company's expenses in cold chain logistics distribution path optimization can be reduced, the logistics service quality can be improved, and the company's transportation costs can be reduced.

[0273] The present invention has been described in detail above with reference to the embodiments. However, the contents described are only preferred embodiments of the present invention and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A cold chain distribution path optimization method based on an improved whale optimization algorithm, characterized by: The steps include: (1) Obtain initial information on demand points and distribution centers and construct an information matrix; (2) Obtain historical traffic information around the distribution center and each demand point in the traffic management system, use data-driven methods to predict the traffic conditions around each demand point and distribution center over a period of time, and design corresponding traffic congestion functions based on the relationship between the distance and travel time; (3) Based on the information matrix and the objective function with the goal of minimizing the total cost, relevant constraints are set to construct a mathematical model for optimizing the cold chain logistics distribution path; (4) Calculate the corresponding transportation time by using the traffic congestion function value and the Euclidean distance between the distribution center and each demand point; (5) Convert the search agent address composed of random variables into a vehicle path matrix through modeling methods; (6) Dynamically analyze the actual load rate of each route and, taking into account traffic congestion conditions, calculate the fuel consumption of vehicles during normal driving and idling; (7) Classify whale populations using the DBSCAN clustering algorithm and solve the model using the clustered whale optimization algorithm; (8) Provide guidance for vehicle scheduling and route planning based on the cold chain logistics vehicle distribution route optimization results output after the optimization is completed.

2. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 1 is characterized by: The method for obtaining the initial information in step (1) is: Construct a supply-demand relationship matrix based on the distribution center's reserves and the demand at the demand points; Establish a location information matrix based on the location information of the distribution center and each demand point; Based on the initial information, the following data are calculated: the transportation distance between each demand point, the transportation distance between each demand point and the distribution center, and the time required for unloading at each demand point; Estimate the number of vehicles needed to complete the delivery task in advance based on the total demand of each demand point, the load weight of a single cold chain vehicle in the distribution center, the type and volume of goods, and the maximum loading volume of the cold chain vehicle; Among them, the estimated number of vehicles required to complete the delivery task is as follows: m=max(m1,m2) (1) Where m1 is the number of vehicles estimated based on the weight of the goods, m2 is the number of vehicles estimated based on the volume of the goods, and m represents the number of vehicles required to complete the delivery task. represents the rounding function; n represents the total number of demand points; g i represents the demand at demand point i; f represents the volume occupied by goods per unit weight, and its value is determined by the type of goods; Q V It represents the volume of cargo that can be loaded in a single cold chain vehicle compartment; λ1 and λ2 represent vehicle loading coefficients, which are related to the complexity of loading and vehicle constraints, including cargo type, loading method, weight constraint and volume constraint.

3. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 2 is characterized by: The traffic conditions around each demand point and distribution center within a period of time predicted in step (2) include: First, according to the size of the overall solution space, three radius ranges are determined: 0 to r1, r1 to r2, and greater than r2; Then, a traffic congestion detection system is established and a congestion determination function is designed. Based on the historical traffic conditions around each demand point and supply point, the traffic congestion function values ​​z1, z2, and z3 corresponding to different radius ranges of the demand point and distribution center are obtained. The worse the road conditions, the greater the traffic congestion function value. The specific calculation method is as follows: First, obtain the historical traffic information around each distribution center and demand point, and obtain the travel time data of the area corresponding to each time period. Use this data to establish a proxy model, and continuously optimize the proxy model through Bayesian optimization to make it close to the real model. Finally, use the optimized model to output the travel time relationship of the area in the next period of time, substitute the travel time of each area into the congestion calculation function, and thus obtain the corresponding traffic congestion function value.

4. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 2 or 3 is characterized in that: The objective function of the mathematical model for cold chain logistics distribution path optimization in step (3) consists of five costs, including: (1) Vehicle fixed costs If the number of demand points is n, and the number of vehicles calculated in formula (1) is m, then the fixed cost of vehicles is: Where F is the sum of the driver's salary, vehicle depreciation, and maintenance costs for one vehicle. Matrix y is used to record which vehicle delivers each demand point. n+1 represents the distribution center. The values ​​of matrix y are as follows: (2) Transportation costs The transportation cost is the fuel consumption cost during the transportation process. The fuel consumption during transportation is determined by the actual load rate of the cold chain vehicle and the transportation distance, as follows: U=p ijk / Q (3) Formula (3) is the calculation process of the actual load rate of vehicle k on the transportation path from demand point i to demand point j: ijk represents the actual load of vehicle k on the transport path from demand point i to demand point j; Q represents the maximum load of the vehicle; C’=(1+bT idle )d ij [e1+p ijk (e2-e1) / Q] (4) Formula (4) is the total fuel consumption on the transportation path from demand point i to demand point j, where: ij represents the transport distance; e1 represents the fuel consumption per unit distance when the vehicle is unloaded; e2 represents the fuel consumption per unit distance when the vehicle is fully loaded; b represents the fuel consumption when the vehicle is idling; T idle Indicates the vehicle idling time; Formula (4) has given the total fuel consumption on a route. Assuming the unit oil price is a, the total transportation cost is: Where s ijk Used to record vehicle driving trajectory. Its specific values ​​are as follows: s is the vehicle path matrix, which is transformed from the search agent address composed of random variables; (3) Cargo damage costs Cargo damage cost refers to the loss cost of cold chain products caused by the passage of time and temperature changes during transportation and unloading. Its expression is as follows: Where p3 is the unit price of the product, q1 is the deterioration rate of the product during transportation, q2 is the deterioration rate of the product during unloading, 1 to n represent the demand points, n+1 is the distribution center, t ij is the transportation time of the vehicle from point i to point j, T j is the time taken by the vehicle to unload at demand point j; (4) Refrigeration cost The refrigeration cost is the cost incurred by the cold chain vehicle due to the consumption of refrigerant during the vehicle pre-cooling, transportation and unloading process. The pre-cooling time T pre The assignment method is as follows: Frozen goods: When the set temperature is -20℃ and the ambient temperature is >30℃, T pre =2h; 30℃>ambient temperature>20℃, T pre =1.5h; when the ambient temperature is less than 20℃, T pre =1h; Refrigerated goods: When the set temperature is 4°C and the ambient temperature is >30°C, T pre =1.5h; 30℃>ambient temperature>20℃, T pre =1h; when the ambient temperature is less than 20℃, T pre =2 / 3h; Assuming that the temperature is 25 degrees, during the vehicle pre-cooling process, the refrigeration cost generated by the cold chain vehicle per unit time is p 41 During the transportation process, the refrigeration cost of the cold chain vehicle in transporting products per unit time is p 42 During the unloading process, the refrigeration cost incurred by the cold chain vehicle in unloading the goods per unit time is p 43 ; α represents the impact of temperature on cooling energy consumption, so the cooling cost is (5) Carbon emission costs Carbon emission cost is the cost of carbon dioxide emissions caused by vehicle energy consumption and refrigerant consumption during transportation. The fuel consumption of a vehicle is related to factors such as the vehicle's driving distance, load capacity, and speed. The more fuel consumed, the more carbon dioxide emissions will be. Let p c is the unit carbon tax price, e is the carbon dioxide emission coefficient, and E is the energy consumption per unit time of the refrigeration equipment. The fuel consumption of each path is given according to formula (4). Therefore, the total carbon emission cost of vehicle operation and refrigeration equipment is: In addition, the objective function must also satisfy the vehicle load constraint, that is, for any p ijk , there is p ijk <=Q, integrating the constraint that the vehicle cannot be overloaded into the objective function of the shortest path, the overload penalty cost is obtained as: In addition, each demand point is delivered by only one vehicle, with the following constraints: The number of vehicles returning to the distribution center is equal to the m vehicles dispatched for distribution, subject to the following constraints: The matrix y can be calculated from the matrix s, and the relationship between the two is as follows: The variables in the matrix x all satisfy the 0-1 constraint, and we have: x ijk (x ijk -1)=0 (14) The variables in the matrix y all satisfy the 0-1 constraint, and we have: and ijk (and ijk -1)=0 (15) To sum up, the cold chain distribution path optimization model is: minC=C1+C2+C3+C4+C5+C6 (16) Constraints: x ijk (x ijk -1)=0 (14) and ijk (and ijk -1)=0 (15)。 5. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 4 is characterized in that: In step (4), the transportation time calculation process based on traffic conditions is as follows: If d is the distance from the starting point to the end point, and all transport vehicles travel at a constant speed, the transport time t in the mathematical model can be expressed as: t=d*z1 / v,d <r1 Among them, r1 and r1 represent different radius ranges, and z1, z2, and z3 represent the traffic congestion function values ​​for each radius range.

6. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 4 is characterized by: In step (5), the search agent address composed of random variables is converted into a vehicle path matrix through a modeling method, including: a. If you need to solve the problem of delivering m vehicles to n demand points, then input a one-dimensional array x containing n+m-1 numbers; b. Sort x and store the result in array A, while array x remains unchanged; c. Traverse array x and compare it with array A, output the position of the element in A and store it in array B; d. Define an array C of n+m-1 with all zeros, and let B i For the i-th element in B, let h = B i， Traverse array B and do the following: B i =B i +C h C h =C h +1 (18) Thus, duplicate elements in B are removed, making it strictly arranged from 1 to n+m-1; e. Change the values ​​of elements in array B that are greater than n+1 to n+1 to separate the front and back elements, representing the next transport vehicle. Elements less than n+1 are demand points, thus dividing B into m parts, each representing a vehicle. The order of visiting demand points from front to back is the order; f. Traverse matrix B and enter its information into the three-dimensional path access sequence matrix s.

7. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 4 is characterized by: Step (6) dynamically analyzes the actual load rate of each path and, combined with traffic congestion, calculates the fuel consumption of the vehicle at idle speed, including: Assume that the vehicle's idling fuel consumption per unit time is b, and the idling time is T idle It is related to the congestion coefficient r, and its specific value is: The calculation steps of load matrix p, actual load rate U and fuel consumption are as follows: a. Traverse array B in reverse order and refer to the access order matrix s, add the demand of each demand point to the corresponding path in sequence and record it in the three-dimensional matrix p; b. Calculate the actual load rate according to formula (3); c. Calculate fuel consumption according to formula (4).

8. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 1 is characterized by: Step (7) classifies the whale population using the DBSCAN clustering algorithm. When solving the model using the clustered whale optimization algorithm, at least m individuals are searched globally, specifically including: Population partitioning strategy: DBSCAN clustering i = dbscan(X(t), e, min) k=i(X * (t)) (19) Where X(t) is the position matrix of the population members at the tth iteration, X * (t) represents the position of the whale with the best fitness at the tth iteration. dbscan is a pre-set function that clusters the matrix X(t) according to the parameters e and min. The parameter e represents the neighborhood radius and min represents the minimum number of neighborhood points. After clustering, the classification index of each whale individual will be recorded in the one-dimensional array i. The parameter k is used to record the category of the best individual. In this algorithm, the parameter settings are as follows: min=n / 30 Where ub(j) is the upper limit of the value given by the current dimension, lb(j) is the lower limit of the value given by the current dimension, and n is the number of whales; Clustering restrictions: As the algorithm iterates, whale individuals will cluster under the action of the movement strategy in formula (2). Set the clustering limit maxpopulation and parameter m. If the individual satisfies: i(X(t)) = k && m < = maxpopulation, the individual will update its position according to the spiral approach in the whale optimization algorithm. Otherwise, the individual will randomly approach other individuals. The initial value of parameter m is 1, and its value increases by 1 each time the clustered individual updates its position.

9. The cold chain distribution path optimization method based on the improved whale optimization algorithm according to claim 7 is characterized in that: The specific results of step (8) cold chain logistics vehicle delivery route optimization are as follows: a. First, obtain the initial information matrix and then establish a traffic congestion detection system; b. Input the traffic conditions around each demand point and distribution center to obtain the corresponding traffic congestion function value; c. Substitute the traffic congestion function value into formula (17) to calculate the transportation time; d. Generate the path access sequence through step (5); e. Calculate the load matrix and fuel consumption of each path by step (6) and by accessing the path in sequence; f. Substitute the above data into the cold chain logistics distribution path optimization mathematical model; g. Set the mathematical model for cold chain logistics distribution path optimization as the objective function, calculate the fitness of each individual, cluster the whale population, and form an initial cluster population centered on the optimal individual. Afterwards, non-clustered members and cluster members update their positions according to different movement strategies and re-cluster after each update. Compare the fitness of parent and offspring individuals, select individuals with better fitness as the new offspring, compare the fitness of the new population individuals, select the optimal individual as the new leader, and iterate to the next generation.

Images