Shared electric vehicle operation optimization method based on column generation and ant colony algorithm fusion

Abstract

The present invention relates to a shared electric vehicle operation optimization method and system based on the fusion of column generation and ant colony algorithm, and the method includes: S1: obtaining shared electric vehicle reservation orders and forecast information; S2: determining the objectives and constraints of shared electric vehicle fleet configuration and path optimization according to the shared electric vehicle reservation orders and forecast information; S3: establishing a mixed integer linear programming model according to the objectives and constraints to maximize the total profit of the shared electric vehicle system; S4: using D-W decomposition to decompose and reconstruct the mixed integer linear programming model to obtain a reconstructed planning model; S5: using a hybrid algorithm composed of column generation and ant colony algorithm to solve the reconstructed planning model, and calculate the fleet composition and path planning scheme. The present invention solves the operation optimization problem in the shared electric vehicle system, improves the profitability of the operating enterprise, and has high decision reliability and strong calculation stability.

Description

Technical Field

[0001] The present invention relates to the field of urban shared transportation, and in particular to a shared electric vehicle operation optimization method and system based on the fusion of column generation and ant colony algorithm. Background Art

[0002] With the development of new technologies such as the Internet, cloud computing and artificial intelligence, people's daily travel methods are undergoing profound changes. Private cars are convenient to use, but the problems of high ownership costs and insufficient parking spaces are very prominent; public transportation such as buses and subways face the problems of frequent parking, inconvenient transfers and insufficient capacity. With the development of mobile Internet, more and more people are beginning to use intelligent and personalized shared transportation methods. Shared transportation is a new type of transportation between private cars and public transportation. It is a subdivision of the car rental market, namely time-sharing rental. The essence of this travel method is to arrange the same car for multiple customers whose travel times do not overlap, which makes up for the shortcomings of the traditional transportation system.

[0003] Shared car companies use electric vehicles as their main products. On the one hand, they respond to the national new energy policy and can achieve green travel; on the other hand, electric vehicles are easy to manage and have low operating costs. Electric vehicles are easy to operate and easy to connect to the Internet, allowing companies to more conveniently dispatch and monitor battery and other operating data. Although electric vehicles have a higher purchase cost than similar fuel vehicles, pure electric vehicles can enjoy subsidy policies such as exemption from purchase tax when purchased. In addition, in the long run, electric vehicles have lower operating costs. In order to meet the diverse travel needs of citizens and enhance the competitiveness of shared services, shared electric vehicle companies have begun to support off-site vehicle returns. However, after a period of operation, the sharing model that supports off-site vehicle returns will have a situation where the distribution of vehicles cannot match the needs of customers. Compared with traditional fuel vehicles, the main problem of electric vehicles is short driving range and long charging time. For companies that provide shared car services, in order to provide more satisfactory shared electric vehicle rental services, they must consider the rebalancing and charging operations of the vehicle when arranging sharing tasks to ensure that the vehicle can complete the sharing tasks to be performed. In addition, with the development of the sharing industry and the popularization of new electric vehicles, the electric vehicle fleets of shared electric vehicle companies have gradually changed from a single model to multiple models. Small vehicles have low costs but small battery capacity and require frequent energy replenishment; vehicles with long mileage can complete long-distance orders and have the ability to execute multiple orders continuously, but the acquisition cost has also increased significantly. The existence of this reality of multiple models has further increased the complexity of shared electric vehicle fleet management and route planning in the operation of shared electric vehicle companies.

[0004] At present, the sharing systems of shared electric vehicle companies mostly manage their fleets based on experience during operation. The random selection of shared electric vehicles to execute order tasks lacks systematic management and global coordination, which seriously affects the utilization rate of shared electric vehicles and the profitability of shared electric vehicle systems, and is not conducive to the long-term healthy development of shared electric vehicle companies. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a shared electric vehicle operation optimization method and system based on the fusion of column generation and ant colony algorithm.

[0006] The technical solution of the present invention is: a shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm, comprising:

[0007] Step S1: Collecting the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, preprocessing it, and obtaining the shared electric vehicle reservation order and prediction information; wherein the preprocessing includes: predicting the actual driving distance, car rental fee and penalty fee of the user based on empirical data; adding virtual start and end tasks;

[0008] Step S2: According to the shared electric vehicle reservation order and the forecast information, determine the objectives and constraints of the shared electric vehicle fleet configuration and route optimization, wherein the objective is to maximize the total profit of the shared electric vehicle system, and the constraints include: the user cannot be visited repeatedly, and the shared electric vehicle is limited in mileage, time and space;

[0009] Step S3: establishing a mixed integer linear programming model according to the objective and constraint conditions; wherein the mixed integer linear programming model includes: an objective function and its constraint conditions;

[0010] Step S4: Decomposing and reconstructing the mixed integer linear programming model using DW decomposition, introducing new variables, and obtaining a reconstructed programming model, including: a main problem model and corresponding sub-problem models;

[0011] Step S5: adopting a hybrid algorithm consisting of column generation and ant colony algorithm to solve the reconstructed planning model, obtain the fleet charging plan of the shared electric vehicles and the path planning results of the shared electric vehicles, and calculate the fleet composition and path planning plan.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] 1. The present invention discloses a shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm. Based on the actual situation of multiple types of electric vehicles in the current shared electric vehicle system operating enterprises, the previous research on shared electric vehicle fleet management and route planning for a single model is no longer applicable. The present invention introduces the situation of multiple types of electric vehicles into the model, allowing shared electric vehicles to partially charge and selectively provide services to customers, and comprehensively considers fleet composition, order assignment, route planning, rebalancing and charging, which has practical use value.

[0014] 2. The shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm disclosed in the present invention aims to maximize the total profit of the system, comprehensively considers car rental income, rebalancing costs, fleet fixed costs and penalties for rejected orders, and regards the current fleet management and path optimization problems as a further extension of traditional path planning problems. It can provide decision support and system planning for shared electric vehicle systems and improve the total profit of shared electric vehicle companies.

[0015] 3. The shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm disclosed in the present invention combines the DW decomposition technology and column generation technology in the exact algorithm with the ant colony optimization algorithm in the heuristic algorithm when solving the model, which can improve the calculation speed while ensuring the accuracy of the solution. Through random case comparison, it was found that in small-scale instances, the average difference between the fusion algorithm of the present invention and the exact solution found by CPLEX was less than 5%. For large instances with more than 100 orders, the planning results are better than the optimal solution found by CPLEX within 1 hour, and are also significantly better than the solution obtained by the pure ant colony optimization algorithm in a similar time. The method of the present invention is applicable to instances of all scales and has achieved a balance between solution quality and solution speed, and has obvious advantages in efficiency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a method for optimizing the operation of a shared electric vehicle based on the fusion of column generation and ant colony algorithm in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of the fusion process of the column generation ant colony algorithm in an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a spatiotemporal network diagram of a path planning result in an embodiment of the present invention;

[0019] Figure 4 It is a structural block diagram of a shared electric vehicle operation optimization system based on the fusion of column generation and ant colony algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention provides a shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm, which solves the operation optimization problem in the shared electric vehicle system by integrating precise and intelligent heuristic algorithms, improves the profitability of shared electric vehicle operating companies, and has the advantages of high decision-making reliability and strong computing stability.

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below through specific implementations and in conjunction with the accompanying drawings.

[0022] Embodiment 1

[0023] like Figure 1 As shown, an embodiment of the present invention provides a shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm, comprising the following steps:

[0024] Step S1: Collecting the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, preprocessing it, and obtaining the shared electric vehicle reservation order and prediction information; wherein the preprocessing includes: predicting the user's actual driving distance, car rental fee and penalty fee based on empirical data; adding virtual start and end tasks;

[0025] Step S2: Determine the objectives and constraints of the fleet configuration and route optimization of the shared electric vehicles based on the shared electric vehicle reservation orders and forecast information; wherein the objective is to maximize the total profit of the shared electric vehicle system, and the constraints include: users cannot be visited repeatedly, and the shared electric vehicles are limited in mileage, time and space;

[0026] Step S3: Establishing a mixed integer linear programming model according to the objective and constraints; wherein the mixed integer linear programming model includes: an objective function and its constraints;

[0027] Step S4: Decompose and reconstruct the mixed integer linear programming model using DW decomposition, introduce new variables, and obtain a reconstructed programming model, including: a main problem model and corresponding sub-problem models;

[0028] Step S5: A hybrid algorithm consisting of column generation and ant colony algorithm is used to solve the reconstructed planning model, obtain the fleet charging plan of shared electric vehicles and the path planning results of shared electric vehicles, and calculate the fleet composition and path planning plan.

[0029] In one embodiment, the above step S1: collects the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, pre-processes it, and obtains the shared electric vehicle reservation order and prediction information; wherein the pre-processing includes: predicting the user's actual driving distance, car rental fee and penalty fee based on empirical data; adding virtual start and end tasks, specifically including:

[0030] The reservation information uploaded from the user terminal of the shared electric vehicle system includes the time and location of picking up and returning the vehicle, as shown in Table 1. In order to facilitate model input, the exact time is converted into an integer. In the embodiment of the present invention, the operating time is 8:00-10:00, 8:00 is regarded as the 0 moment, and each minute is regarded as a time node. The vehicle usage time can be converted into an integer in the interval [0,120]. The conversion results of the embodiment of the present invention are shown in the 4th and 5th columns of Table 2. According to the traffic conditions in Beijing and the charging standards for shared vehicles, it is assumed that the speed of shared electric vehicles is 30km / h. The charging standards for electric vehicles are divided into two parts: usage time and actual mileage. 2 yuan / km+0.2 yuan / minute, thus predicting the actual mileage and car charges for each scheduled order. The loss (penalty) caused to the operator after being abandoned is 1 / 2 of the income; adding virtual placement positions 4 and 5, moving from any position to positions 4 and 5 does not consume time or electricity, and the distance between other stations is represented by vector distance; the virtual starting order (order 0) and ending order (order 8) are parked at positions 4 and 5 respectively, and the pick-up / return time is the minimum time 0 and the maximum time 120 respectively, which will not consume actual car charges. In order to ensure that they are served, the penalty value is set to a maximum value, which is 1000000 in this case. The processed reservation information is shown in Table 2;

[0031] Table 1 Reservation information uploaded by the user terminal

[0032]

[0033] Table 2 Processed reservation order information

[0034]

[0035] In one embodiment, the above step S2: according to the shared electric vehicle reservation order and forecast information, determine the objectives and constraints of the shared electric vehicle fleet configuration and route optimization, wherein the objective is to maximize the total profit of the shared electric vehicle system, and the constraints include: users cannot be visited repeatedly, and the shared electric vehicle is limited in mileage, time and space, specifically including:

[0036] Determine the objective function and constraints of the heterogeneous fleet configuration and path optimization model. In the mixed integer linear programming model with the goal of maximizing the total profit of the shared electric vehicle system, the income in the objective function is the service fee obtained by the shared electric vehicle for executing orders, and the expenditure is the cost of rebalancing the shared electric vehicle, the loss to the company caused by the negative impact of unexecuted orders, and the fixed cost of the fleet. Shared electric vehicles must also meet the following constraints during operation: (1) Each order is served at most once, or it may not be served; (2) The start time of subsequent orders in the task chain is always later than the end time of the previous order; (3) There is sufficient time between two tasks to adjust the position of the shared electric vehicle; (4) The available mileage of the shared electric vehicle is always greater than the next task (rebalancing or service order); (5) The data type of the decision variable meets the requirements.

[0037] In one embodiment, the above step S3: establishes a mixed integer linear programming model according to the objective and the constraints; wherein the mixed integer linear programming model includes: the objective function and its constraints, specifically including:

[0038] The objective function is shown in the following formula (1):

[0039]

[0040] Among them, the first part of formula (1) represents the total revenue obtained by shared electric vehicles from executing orders, the second part represents the total cost of the shared electric vehicle rebalancing process, the third part represents the loss to the company caused by the negative impact of not executing orders, and the fourth part represents the fixed cost of the fleet;

[0041] The constraints are shown in the following formulas (2) to (13):

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] The meanings of the mathematical symbols used in formulas (1) to (13) are as follows:

[0055] H is the set of all shared electric vehicles, h∈H;

[0056] A is the set of all orders, a,b∈A;

[0057] A0 is the set of all orders and virtual start tasks, and there exists A0=A∪a0;

[0058] A n+1 For the set of all orders and virtual termination tasks, there exists A n+1 =A∪a n+1 ;

[0059] is the maximum mileage of the shared electric vehicle h;

[0060] The safe driving range of the shared electric vehicle h;

[0061] FC h is the fixed cost of the shared electric vehicle h;

[0062] EC h is the charging rate of the shared electric vehicle h;

[0063] ST a is the pickup time for order a;

[0064] ET a is the return time of order a;

[0065] L a is the actual mileage of order a;

[0066] G a The income obtained after serving order a;

[0067] P a The penalty for losing order a;

[0068] The remaining mileage after shared electric vehicle h completes order a;

[0069] is the cost of moving the shared electric vehicle h from the return location of order a to the pick-up location of order b;

[0070] T ab is the travel time from the return location of order a to the pick-up location of order b;

[0071] D ab The distance between the return location of order a and the pick-up location of order b;

[0072] is a decision variable, which means that if the shared electric vehicle h provides service for order a, it takes 1, otherwise it takes 0;

[0073] is a decision variable, which means that if the shared electric vehicle h serves order a before serving order b, it takes 1, otherwise it takes 0;

[0074] is a decision variable, which represents the charging time of the shared electric vehicle h at the pick-up location of order a;

[0075] is a decision variable, which represents the charging time of the shared electric vehicle h at the return location of order a;

[0076] is an intermediate variable, which indicates the additional mileage of the shared electric vehicle h after charging at the pick-up and return locations of order a.

[0077] Among them, constraint (2) means that each order can be accessed at most once, and it is allowed not to be accessed;

[0078] Constraint (3) represents two decision variables and The connection between them, if one order is executed it must be the predecessor task of some other order;

[0079] Constraint (4) is the flow balance constraint in the network, which means that each serviced order has both a predecessor task and a successor task;

[0080] Constraint (5) is a time constraint, which can limit the shared electric vehicle to move to the pick-up location of the order before the start time of the next order;

[0081] Constraints (6) to (11) are constraints on mileage: Constraint (6) indicates that the shared electric vehicle has sufficient available mileage for rebalancing before rebalancing;

[0082] Constraints (7) and (8) indicate that the available driving range of a shared electric vehicle after each charge does not exceed its maximum driving range;

[0083] Constraint (9) indicates that after completing each order, the remaining mileage of the shared electric vehicle is no less than the safe mileage;

[0084] Constraints (10) and (11) are state transition equations about mileage, which together indicate that the mileage of a shared electric vehicle after charging at the pick-up location of the mission is the same as the mileage before starting the mission;

[0085] Constraints (12) and (13) define the data types of the decision variables.

[0086] In one embodiment, the above step S4: using DW decomposition to decompose and reconstruct the mixed integer linear programming model, introducing new variables, and obtaining a reconstructed programming model, including: a main problem model and corresponding sub-problem models, specifically including:

[0087] Step S41: Introduce a new set Ω h is the set of all feasible paths for shared electric vehicle h, r∈Ω h ; New variables is a 0-1 variable, which takes the value 1 if path r contains order a, and takes the value 0 otherwise; is a decision variable, which takes 1 if the driving path of the shared electric vehicle h is r, otherwise it takes 0;

[0088] Step S42: convert the mixed integer linear programming model into the following main problem model:

[0089] The objective function of the main problem is:

[0090] The main problem constraints are:

[0091]

[0092]

[0093]

[0094] in, is a 0-1 variable, which takes the value 1 if path r contains order a, and takes the value 0 otherwise;

[0095] is a decision variable, which takes 1 if the driving path of the shared electric vehicle h is r, otherwise it takes 0;

[0096] P a The penalty for losing order a;

[0097] FC h is the fixed cost of the shared electric vehicle h;

[0098] is the total revenue obtained after serving all orders in path r, there exists

[0099] is the total cost of rebalancing operations in path r, there exists in, is a 0-1 variable. If order a and order b are consecutive in path r, it takes 1, otherwise it takes 0;

[0100] Step S43: Decompose the mixed integer linear programming model to obtain a sub-problem model:

[0101] The sub-problem objective function is shown in the following formula (18):

[0102]

[0103] The constraints of the sub-problem are: Formulas (3) to (13)

[0104] Where: π a ,π0 represent the dual variables that restrict the first and second constraints of the master problem respectively;

[0105] It is a 0-1 variable, which takes the value 1 if the path r contains order a, and takes the value 0 otherwise.

[0106] In one embodiment, the above step S5: using a hybrid algorithm consisting of column generation and ant colony algorithm to solve the reconstructed planning model, obtaining a fleet charging plan for shared electric vehicles and a path planning result for shared electric vehicles, and calculating a fleet composition and a rebalancing plan for shared electric vehicles, specifically includes:

[0107] In this step, the processed reservation order information (as shown in Table 2), the locations of three delivery sites (as shown in Table 3) and three electric vehicle parameters (as shown in Table 4) are used as input, and the column generation ant colony fusion method is used to solve the reconstructed model in step S4. First, only some columns of the main problem are saved to obtain the restricted main problem; then the column generation method is used to combine the restricted main problem and the pricing subproblem, and the pricing subproblem is solved using the ant colony algorithm to obtain a legal path that meets the constraints, that is, a column in the column generation method, and determine whether to add columns to the restricted main problem until no columns can be found to be added, and the current restricted main problem is solved. The solution obtained at this time is also the optimal solution to the original mixed integer linear programming model, and the configuration and path planning plan of the fleet are obtained. The flow chart of the column generation and ant colony fusion method is shown as follows. Figure 2 shown.

[0108] Table 3 Information on the possible locations for shared electric vehicles

[0109] Delivery site number longitude latitude 1 13 6 2 15 11 3 14 18

[0110] Table 4 Key parameter information of shared electric vehicles

[0111]

[0112] The specific steps are as follows:

[0113] Step S51: Use the set Ω' of partially feasible paths containing the shared electric vehicle h h Alternative set Ω h The restricted master problem is obtained; where the initial Ω' h It only contains one path that goes directly from the virtual starting point to the virtual end point;

[0114] In this step, a set Ω′ is first used that only includes a path for the shared electric vehicle h to move from the virtual placement position 4 to the virtual placement position 5. h Alternative set Ω h Get the restricted master problem.

[0115] Step S52: calling the solver CPLEX to solve the restricted main problem and calculate the dual variables of the decision variables, and using the dual variables to update the sub-problem objective function; wherein the value corresponding to the objective function of the sub-problem is called the test number;

[0116] Step S53: Use the ant colony algorithm to solve the pricing sub-problem, and obtain the test number and the legal path including the rebalancing solution and the charging solution. The path is a potential column in the column generation method, specifically including:

[0117] Step S531: Data initialization: input all order data, shared electric vehicle placement location data, and shared electric vehicle parameters into the algorithm and set the ant colony algorithm parameters, where:

[0118] The data of the placement locations of shared electric vehicles include: the longitude and latitude of the placement locations, as shown in Table 3;

[0119] The order data includes: pick-up time, pick-up location, return time, return location, estimated actual driving distance, order fee, and penalty value when not served, as shown in Table 2;

[0120] The parameters of shared electric vehicles include: the maximum driving range of electric vehicles, charging rate, cost per kilometer of movement, and daily fixed cost, as shown in Table 4;

[0121] The parameters involved in the ant colony algorithm in the embodiment of the present invention are as follows:

[0122] K is the number of ants, which is the same as the number of orders, i.e. 7;

[0123] I is the maximum number of times the same target value is returned, and its value is 20;

[0124] ρ is the volatility of pheromone, which takes the value of 0.2;

[0125] τ ab represents group wisdom, that is, the pheromone concentration between order a and order b, with an initial value of 10;

[0126] η ab Represents individual intelligence, that is, the tendency of an ant individual to choose order b after executing order a is 1≤η ab ≤10;

[0127] α is the collective wisdom inspiration factor, which takes the value of 1;

[0128] β is the individual wisdom inspiration factor, which takes the value of 1;

[0129] Q is a constant, taking the value of 1;

[0130] Step S532: Construct a path. The shared electric vehicle always starts from the virtual starting order 0, visits the scheduled orders one by one until the virtual ending order 8 is visited, and a complete task chain is formed, which represents a feasible solution of the shared electric vehicle path. The path taken by the ant represents the feasible solution of the shared electric vehicle path; ant k will decide its direction of travel, that is, the next order to visit, according to individual preferences and group wisdom, and score each accessible order according to the target value increment that the order can bring. The score given by ant k for order b can be recorded as In the nth iteration, the probability calculation formula (19) that ant k takes order b as the next service object after visiting the current order a is as follows:

[0131]

[0132] Where: Set N k represents the set of orders that ant k can access, η ab (n) represents the preference of the ant individual in the nth iteration, and its value is During the iteration process, the higher the score of an order, the higher the probability of being selected;

[0133] In this step, the shared electric vehicle always starts from the virtual starting order 0, and visits the scheduled orders one by one until the virtual ending order 8 is visited, forming a complete task chain, which represents a feasible solution for the shared electric vehicle path. After ant k completes the current task, it first selects the set N of orders that can be the next service object according to constraints (5)-(13). k . Score each accessible order based on the target value increment that the order can bring. Ant k is a set N k The order score of order b is recorded as According to individual preferences and group wisdom, the direction of its progress is determined, that is, the next order to be visited. In the nth iteration, after ant k visits the current order a, it will collect N k The probability that order b in is the next service object is calculated by formula (19);

[0134] Among them, the score The calculation formula (20) is as follows:

[0135]

[0136] in: is the score of the next service object of ant k, which is order b;

[0137] is the target value corresponding to the pricing subproblem of ant k’s next service object being order b;

[0138] is the maximum target value of the corresponding pricing subproblem among all next-point accessible orders of ant k;

[0139] is the minimum objective value of the corresponding pricing subproblem among all next-point accessible orders of ant k;

[0140] Step S533: using a roulette wheel method to select the next service object from all accessible orders according to the probability calculated in step S532;

[0141] Step S534: Determine whether all ants have reached the end point. If all have reached the end point, go to step S535; otherwise, return to step S532;

[0142] Step S535: Select the best path found in this round from the paths found by all ants according to the target value, compare the best path in this round with the best historical path of the ant colony up to this point, update the current best historical path, and record the path planning result and the corresponding rebalancing and charging scheme, i.e., the column in the column generation;

[0143] Step S536: Determine whether the termination condition is met. If the number of times the optimal target value of the found path remains unchanged continuously reaches a preset value, the optimal solution is obtained, and the path planning result including the rebalancing and charging scheme found by the ant colony is output; otherwise, all ants are located at the virtual starting point, and the score of accessible points is updated based on the pheromone concentration. Then, return to step S532 to start a new round of iteration. The greater the contribution of continuous orders to the inspection number, the greater the increase in pheromone concentration corresponding to the order;

[0144] Among them, the updated calculation formulas (21) to (22) of the pheromone concentration are as follows:

[0145]

[0146]

[0147] Where: Δ(n) is the growth coefficient in the nth iteration;

[0148] RC max (n) is the number of tests corresponding to the best path obtained in the nth iteration;

[0149] RC max is the best path in history found in the first n rounds of iterations;

[0150] τ ab (n) is the concentration of pheromone between order a and order b after the nth iteration;

[0151] Q is a constant.

[0152] Step S54: Based on the solution of the pricing subproblem, determine whether the test number meets the test criteria for column generation: If it meets the criteria, add the column to the restricted main problem, that is, add the obtained legal path as an element to the set Ω' h and returns to step S52; if it does not meet the standard, it indicates that there is no valuable column, then the current restricted main problem is solved, and the solution obtained is the optimal solution to the original main problem. The path planning results including the charging plan and the rebalancing plan are output, and the fleet composition and the initial deployment position are deduced based on the results.

[0153] According to the above steps, the heterogeneous fleet determination and path planning model is solved to obtain the fleet size of the sharing system, the response to orders, and the path planning solution. Figure 3 As shown in the figure, two Type 1 vehicles were used without charging in between. The two vehicles were placed at Position 3 and Position 1 respectively. The task chains were 0-1-3-6-8 and 0-5-7-8 respectively. The tasks of accepting Order 2 and Order 4 were abandoned. The specific location and remaining mileage of the shared electric vehicle at each moment are shown in Figure 3 Winning bid.

[0154] The present invention discloses a shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm. Based on the actual situation of multiple types of electric vehicles in the current shared electric vehicle system operating enterprises, the previous research on shared electric vehicle fleet management and path planning for a single model is no longer applicable. The present invention introduces the situation of multiple types of electric vehicles into the model, allowing shared electric vehicles to partially charge and selectively provide services to customers, and comprehensively considers fleet composition, order assignment, path planning, rebalancing and charging, which has practical use value.

[0155] The shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm disclosed in the present invention aims to maximize the total profit of the system, comprehensively considers car rental income, rebalancing cost, fleet fixed cost and penalty for rejected orders, regards the current fleet management and path optimization problems as a further extension of traditional path planning problems, and can provide decision support and system planning for the shared electric vehicle system, thereby improving the total profit of shared electric vehicle enterprises.

[0156] The shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm disclosed in the present invention combines the DW decomposition technology and column generation technology in the exact algorithm with the ant colony optimization algorithm in the heuristic algorithm when solving the model, which can improve the calculation speed while ensuring the accuracy of the solution. Through random case comparison, it was found that in small-scale instances, the average difference between the fusion algorithm of the present invention and the exact solution found by CPLEX was less than 5%. For large instances with more than 100 orders, the planning results are better than the optimal solution found by CPLEX within 1 hour, and are also significantly better than the solution obtained by the pure ant colony optimization algorithm in a similar time. The method of the present invention is applicable to instances of all scales and has achieved a balance between solution quality and solution speed, and has obvious advantages in efficiency and reliability.

[0157] Embodiment 2

[0158] like Figure 4 As shown, an embodiment of the present invention provides a shared electric vehicle operation optimization system based on the fusion of column generation and ant colony algorithm, including the following modules:

[0159] The preprocessing module 61 is used to collect the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, preprocess it, and obtain the shared electric vehicle reservation order and prediction information; wherein the preprocessing includes: predicting the user's actual driving distance, car rental fee and penalty fee based on empirical data; adding virtual start and end tasks;

[0160] A target and constraint determination module 62 is used to determine the target and constraint conditions of the fleet configuration and route optimization of the shared electric vehicles according to the shared electric vehicle reservation order and forecast information, wherein the target is to maximize the total profit of the shared electric vehicle system, and the constraint conditions include: the user cannot be repeatedly visited, and the shared electric vehicles are limited in mileage, time and space;

[0161] The mixed integer linear programming model establishment module 63 is used to establish a mixed integer linear programming model according to the objective and constraint conditions; wherein the mixed integer linear programming model includes: an objective function and its constraint conditions;

[0162] The reconstructing planning model module 64 is used to decompose and reconstruct the mixed integer linear programming model using DW decomposition, introduce new variables, and obtain a reconstructed planning model, including: a main problem model and corresponding sub-problem models;

[0163] The module 65 for solving the reconstructed planning model is used to solve the reconstructed planning model by adopting a hybrid algorithm consisting of column generation and ant colony algorithm, obtain the fleet charging plan of shared electric vehicles and the path planning results of shared electric vehicles, and calculate the fleet composition and path planning plan.

[0164] The above embodiments are provided only for the purpose of describing the present invention, and are not intended to limit the scope of the present invention. The scope of the present invention is defined by the appended claims. Various equivalent substitutions and modifications made without departing from the spirit and principles of the present invention should all be included within the scope of the present invention.

Claims

1. A shared electric vehicle operation optimization method based on the fusion of column generation and ant colony algorithm, characterized in that: include: Step S1: Collecting the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, preprocessing it, and obtaining the shared electric vehicle reservation order and prediction information; wherein the preprocessing includes: predicting the actual driving distance, car rental fee and penalty fee of the user based on empirical data; adding virtual start and end tasks; Step S2: According to the shared electric vehicle reservation order and the forecast information, determine the objectives and constraints of the shared electric vehicle fleet configuration and route optimization; wherein the objective is to maximize the total profit of the shared electric vehicle system, and the constraints include: the user cannot be visited repeatedly, and the shared electric vehicle is limited in mileage, time and space; Step S3: Establish a mixed integer linear programming model according to the objective and constraints; wherein the mixed integer linear programming model includes: an objective function and its constraints, specifically including: The objective function is shown in the following formula (1): （1） Among them, the first part of formula (1) represents the total revenue obtained by shared electric vehicles from executing orders, the second part represents the total cost of the shared electric vehicle rebalancing process, the third part represents the loss to the company caused by the negative impact of not executing orders, and the fourth part represents the fixed cost of the fleet; The constraint conditions are shown in the following formulas (2) to (13): （2） （3） （4） （5） （6） （7） （8） （9） （10） （11） （12） （13） The meanings of the mathematical symbols used in formula (1) to formula (13) are as follows: is the collection of all shared electric vehicles, ; is the set of all orders, ; For the set of all orders and virtual start tasks, there exists ; For the set of all orders and virtual termination tasks, there exists ; For shared electric vehicles Maximum mileage; For shared electric vehicles safe driving mileage; For shared electric vehicles fixed costs; For shared electric vehicles The charging rate; For orders Pick-up time; For orders The return time of the vehicle; For orders Actual mileage; For service orders income obtained after For lost orders punishment; For shared electric vehicles Complete order The remaining mileage after To share electric vehicles From the order Move the return location to the order The cost of the pickup location; For orders from Move the return location to the order The driving time to the pickup location; For orders The return location to the order The distance between the pickup locations; is a decision variable, indicating that if the shared electric vehicle For orders 1 if service is provided, 0 otherwise; is a decision variable, indicating that if the shared electric vehicle Complete order Later service order Take 1, otherwise take 0; is the decision variable, representing the shared electric vehicle In order Charging time at the pickup location; is the decision variable, representing the shared electric vehicle In order Charging time at the return location; is an intermediate variable, representing shared electric vehicles In order The additional mileage after charging at the pick-up and return locations. ; Among them, constraint (2) means that each order can be accessed at most once, and it is allowed not to be accessed; Constraint (3) represents two decision variables and The connection between them, if one order is executed, then this order is a predecessor task of some other order; Constraint (4) is the flow balance constraint in the network, which means that each serviced order has both a predecessor task and a successor task; Constraint (5) is a time constraint, which can limit the shared electric vehicle to move to the pick-up location of the order before the start time of the next order; Constraints (6) to (11) are constraints on mileage: Constraint (6) indicates that shared electric vehicles must have sufficient available mileage for rebalancing operations before rebalancing; Constraints (7) and (8) indicate that the available driving range of a shared electric vehicle after each charge does not exceed its maximum driving range; Constraint (9) indicates that after completing each order, the remaining mileage of the shared electric vehicle is no less than the safe mileage; Constraints (10) and (11) are state transfer equations about the driving mileage, and the constraints (10) and (11) together indicate that the driving mileage of the shared electric vehicle after charging at the pick-up location of the task is the same as the driving mileage before starting the task; Constraints (12) and (13) define the data types of the decision variables; Step S4: Decompose and reconstruct the mixed integer linear programming model using DW decomposition, introduce new variables, and obtain a reconstructed programming model, including: a main problem model and corresponding sub-problem models, specifically including: Step S41: Introducing a new set For shared electric vehicles The set of all feasible paths, ; New variables is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; is a decision variable, if shared electric vehicles The driving path is Take 1, otherwise take 0; Step S42: convert the mixed integer linear programming model into the following main problem model: The objective function of the main problem is: (14) The main problem constraints are: （15） （16） （17） in, is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; is a decision variable, if shared electric vehicles The driving path is Take 1, otherwise take 0; For lost orders punishment; For shared electric vehicles fixed costs; To complete the service path The total revenue obtained from all orders in ; For path The total expenditure of the rebalancing operation in ,in, is a 0-1 variable. If the path Order With order Continuously take 1, otherwise take 0; Step S43: Decomposing the mixed integer linear programming model to obtain a sub-problem model: The sub-problem objective function is shown in the following formula (18): （18） The constraints of the sub-problem are: Formula (3)~(13) in: denote the dual variables that constrain the first and second constraints of the master problem, respectively; is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; Step S5: adopting a hybrid algorithm consisting of column generation and ant colony algorithm to solve the reconstructed planning model, obtain the fleet charging plan of the shared electric vehicles and the path planning results of the shared electric vehicles, and calculate the fleet composition and path planning plan.

2. The shared electric vehicle operation optimization method based on column generation and ant colony algorithm fusion according to claim 1 is characterized in that: The step S5: using a hybrid algorithm consisting of column generation and ant colony algorithm to solve the reconstructed planning model, and obtaining the fleet charging plan and the path planning results of the shared electric vehicles. Calculating the fleet composition and path planning plan specifically includes: Step S51: Use a shared electric vehicle The set of partially feasible paths Alternative Collection Get the restricted master problem; the initial It only contains one path that goes directly from the virtual starting point to the virtual end point; Step S52: calling the solver CPLEX to solve the restricted main problem and calculate the dual variables of the decision variables, and using the dual variables to update the sub-problem objective function; wherein the value corresponding to the objective function of the sub-problem is called a test number; Step S53: using an ant colony algorithm to solve the pricing sub-problem, obtaining the test number and a legal path including a rebalancing solution and a charging solution, where the path is a potential column in the column generation method; Step S54: Based on the solution of the pricing sub-problem, determine whether the test number meets the test criteria for column generation: if it meets the criteria, add a column to the restriction main problem, that is, add the obtained legal path as an element to the set and returns to step S52; if it does not meet the standard, it indicates that there is no valuable column, then the current restricted main problem is solved, and the solution obtained is the optimal solution to the original main problem, and the path planning results including the charging plan and the rebalancing plan are output, and the fleet composition and the initial deployment position are deduced based on the results.

3. The shared electric vehicle operation optimization method based on column generation and ant colony algorithm fusion according to claim 2 is characterized in that: The step S53: using an ant colony algorithm to solve the pricing sub-problem, obtaining the test number and a legal path including a rebalancing solution and a charging solution, the path being a potential column in the column generation method, specifically comprising: Step S531: Data initialization: input all order data, shared electric vehicle placement location data, and shared electric vehicle parameters into the algorithm and set the ant colony algorithm parameters, where: The data of the placement location of shared electric vehicles includes: the longitude and latitude of the placement location; Order data includes: pick-up time, pick-up location, return time, return location, estimated actual mileage, order fee, and penalty value for non-service; Shared electric vehicle parameters include: the maximum driving range of electric vehicles, charging rate, cost per kilometer of movement and daily fixed cost; The parameters involved in the ant colony algorithm are as follows: is the number of ants; The maximum number of times the same target value is returned; is the volatilization rate of pheromone; Represents collective wisdom, that is, order With order pheromone concentration between; Represents individual intelligence, that is, the ant individual completes the task Then select the order tendency; It is the inspiration factor of group wisdom; It is the inspiration factor of individual wisdom; Step S532: construct a path, using the path taken by the ants to represent the feasible solution of the shared electric vehicle path; The direction of the next order to be visited will be determined based on individual preferences and group wisdom, and each accessible order will be scored based on the target value increment that the order can bring. For orders The score can be recorded as , in the nth iteration, the ants Complete the current order After the order The probability calculation formula (19) for the next service object is as follows: （19） Among them: Collection Represents ants The collection of orders that can be accessed, Indicates the preference of ant individuals in the nth iteration, and its value is ,In the iteration process, the higher the score of an order, the higher the probability of being selected; Fraction The calculation formula (20) is as follows: （20） in: For Ants The next service object is order score; For Ants The next service object is order The target value corresponding to the pricing sub-problem of ; For Ants The maximum objective value of the corresponding pricing subproblem among all next-point accessible orders; For Ants The minimum objective value of the corresponding pricing subproblem among all next-point accessible orders; Step S533: using a roulette wheel method to select the next service object from all accessible orders according to the probability calculated in step S532; Step S534: Determine whether all ants have reached the end point. If all have reached the end point, go to step S535; otherwise, return to step S532; Step S535: Select the best path found in this round from the paths found by all ants according to the target value, compare the best path in this round with the best historical path of the ant colony up to this point, update the current best historical path, and record the path planning result and the corresponding rebalancing and charging scheme, i.e., the column in the column generation; Step S536: Determine whether the termination condition is met. If the number of times the optimal target value of the found path remains unchanged continuously reaches a preset value, the optimal solution is obtained, and the path planning result including the rebalancing and charging scheme found by the ant colony is output; otherwise, all ants are located at the virtual starting point, and the score of accessible points is updated based on the pheromone concentration. Then, return to step S532 to start a new round of iteration. The greater the continuous order contribution to the inspection number, the greater the increase in the pheromone concentration corresponding to the continuous order; Wherein, the updated calculation formulas (21) to (22) of the pheromone concentration are as follows: （21） （22） in: is the growth coefficient in the nth iteration; For the The number of tests corresponding to the best path obtained in the round iteration; For the front The best path in history found in rounds of iterations; For the After the round of iterations, the order With order The concentration of pheromones between is a constant.

4. A shared electric vehicle operation optimization system based on the fusion of column generation and ant colony algorithm, characterized in that: Includes the following modules: A preprocessing module is used to collect the shared electric vehicle reservation information uploaded by the user terminal of the shared electric vehicle system, preprocess it, and obtain the shared electric vehicle reservation order and prediction information; wherein the preprocessing includes: predicting the actual driving distance, car rental fee and penalty fee of the user based on empirical data; adding virtual start and end tasks; A module for determining objectives and constraints, for determining objectives and constraints for fleet configuration and route optimization of the shared electric vehicles according to the shared electric vehicle reservation order and the forecast information, wherein the objective is to maximize the total profit of the shared electric vehicle system, and the constraints include: the user cannot be visited repeatedly, and the shared electric vehicles are limited in mileage, time and space; A mixed integer linear programming model module is established, which is used to establish a mixed integer linear programming model according to the objectives and constraints; wherein the mixed integer linear programming model includes: an objective function and its constraints, specifically including: The objective function is shown in the following formula (1): （1） Among them, the first part of formula (1) represents the total revenue obtained by shared electric vehicles from executing orders, the second part represents the total cost of the shared electric vehicle rebalancing process, the third part represents the loss to the company caused by the negative impact of not executing orders, and the fourth part represents the fixed cost of the fleet; The constraint conditions are shown in the following formulas (2) to (13): （2） （3） （4） （5） （6） （7） （8） （9） （10） （11） （12） （13） The meanings of the mathematical symbols used in formula (1) to formula (13) are as follows: is the collection of all shared electric vehicles, ; is the set of all orders, ; For the set of all orders and virtual start tasks, there exists ; For the set of all orders and virtual termination tasks, there exists ; For shared electric vehicles Maximum mileage; For shared electric vehicles safe driving mileage; For shared electric vehicles fixed costs; For shared electric vehicles The charging rate; For orders Pick-up time; For orders The return time of the vehicle; For orders Actual mileage; For service orders income obtained after For lost orders punishment; For shared electric vehicles Complete order The remaining mileage after To share electric vehicles From the order Move the return location to the order The cost of the pickup location; For orders from Move the return location to the order The driving time to the pickup location; For orders The return location to the order The distance between the pickup locations; is a decision variable, indicating that if the shared electric vehicle For orders 1 if service is provided, 0 otherwise; is a decision variable, indicating that if the shared electric vehicle Complete order Later service order Take 1, otherwise take 0; is the decision variable, representing the shared electric vehicle In order Charging time at the pickup location; is the decision variable, representing the shared electric vehicle In order Charging time at the return location; is an intermediate variable, representing shared electric vehicles In order The additional mileage after charging at the pick-up and return locations. ; Among them, constraint (2) means that each order can be accessed at most once, and it is allowed not to be accessed; Constraint (3) represents two decision variables and The connection between them, if one order is executed, then this order is a predecessor task of some other order; Constraint (4) is the flow balance constraint in the network, which means that each serviced order has both a predecessor task and a successor task; Constraint (5) is a time constraint, which can limit the shared electric vehicle to move to the pick-up location of the order before the start time of the next order; Constraints (6) to (11) are constraints on mileage: Constraint (6) indicates that shared electric vehicles must have sufficient available mileage for rebalancing operations before rebalancing; Constraints (7) and (8) indicate that the available driving range of a shared electric vehicle after each charge does not exceed its maximum driving range; Constraint (9) indicates that after completing each order, the remaining mileage of the shared electric vehicle is no less than the safe mileage; Constraints (10) and (11) are state transfer equations about the driving mileage, and the constraints (10) and (11) together indicate that the driving mileage of the shared electric vehicle after charging at the pick-up location of the task is the same as the driving mileage before starting the task; Constraints (12) and (13) define the data types of the decision variables; The reconstructing planning model module is used to decompose and reconstruct the mixed integer linear programming model using DW decomposition, introduce new variables, and obtain a reconstructed planning model, including: a main problem model and corresponding sub-problem models, specifically including: Step S41: Introducing a new set For shared electric vehicles The set of all feasible paths, ; New variables is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; is a decision variable, if shared electric vehicles The driving path is Take 1, otherwise take 0; Step S42: convert the mixed integer linear programming model into the following main problem model: The objective function of the main problem is: (14) The main problem constraints are: （15） （16） （17） in, is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; is a decision variable, if shared electric vehicles The driving path is Take 1, otherwise take 0; For lost orders punishment; For shared electric vehicles fixed costs; To complete the service path The total revenue obtained from all orders in ; For path The total expenditure of the rebalancing operation in ,in, is a 0-1 variable. If the path Order With order Continuously take 1, otherwise take 0; Step S43: Decomposing the mixed integer linear programming model to obtain a sub-problem model: The sub-problem objective function is shown in the following formula (18): （18） The constraints of the sub-problem are: Formula (3)~(13) in: denote the dual variables that constrain the first and second constraints of the master problem, respectively; is a 0-1 variable. If the path Included in the order Take 1, otherwise take 0; The module for solving the reconstructed planning model is used to solve the reconstructed planning model by adopting a hybrid algorithm consisting of column generation and ant colony algorithm, obtain the fleet charging plan of the shared electric vehicles and the path planning results of the shared electric vehicles, and calculate the fleet composition and path planning plan.

Images