CN115879657A - Electric vehicle power station changing location path optimization method considering multi-station capacity design - Google Patents
Electric vehicle power station changing location path optimization method considering multi-station capacity design Download PDFInfo
- Publication number
- CN115879657A CN115879657A CN202211217714.9A CN202211217714A CN115879657A CN 115879657 A CN115879657 A CN 115879657A CN 202211217714 A CN202211217714 A CN 202211217714A CN 115879657 A CN115879657 A CN 115879657A
- Authority
- CN
- China
- Prior art keywords
- station
- operator
- power
- vehicle
- constraint
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Electric Propulsion And Braking For Vehicles (AREA)
Abstract
The invention discloses an optimization method of an electric vehicle power changing station site selection path considering multi-station capacity design, which comprises the following steps of S10, determining the relation between the power changing station capacity design and a site selection path strategy; s21, determining a problem target and a constraint condition; s22, determining symbolic representation of the parameters and the variables; s23, establishing a mathematical model; s31, encoding and evaluating a solution; s32, designing in an initialization stage; s33, breaking the operator design; s34, repairing operator design; s35, heuristic power station selection strategy design is carried out; and S36, designing a local search strategy. The method constructs an optimization model of the site selection path of the electric vehicle battery replacement station by taking the minimum total cost as a target, and solves the optimization model by adopting an improved self-adaptive large neighborhood search algorithm.
Description
Technical Field
The invention belongs to the technical field of logistics optimization, and relates to an optimization method for site selection paths of electric vehicle battery replacement stations by considering multi-station capacity design.
Background
In recent years, with the rapid development of mobile internet and the change of living habits, people are increasingly interested in obtaining take-out delivery services. "contactless services" represented by take-out have been developed at an accelerated speed. In the face of increasingly severe environmental issues and stricter emission reduction targets, governments of various countries encourage logistics operators to use alternative vehicle technologies, such as plug-in electric vehicles, hybrid electric vehicles, battery electric vehicles, and the like. The introduction of alternative vehicle technologies is beneficial to the development of sustainable urban logistics. Take-away delivery is a part of urban logistics and typically entails some short-distance delivery tasks. Most take-away distribution enterprises equip distributors with electric vehicles to complete the work. However, electric vehicles face many limitations in real applications, such as limited driving range, too long charging time, etc. From 2019, companies such as iron tower energy, hazaro trip and the like continuously provide services for the power changing cabinet and corresponding supporting systems, and the related market of the power changing cabinet rapidly expands. The full charge of the battery usually requires several hours, and the distributor only needs several minutes to replace the battery, and the replaced battery is charged in the battery replacement cabinet. Trade the required area of electric cabinet far less than fill electric pile, and more have the security. Therefore, the power conversion mode well meets the requirements of the take-out distribution industry. However, due to rapid market expansion, the electric vehicle battery replacement station is lack of planning in site selection, and certain resources are wasted.
By the present time, most researchers neglect the influence of site capacity when deciding the optimal electric vehicle distribution path and the power station changing site selection strategy. They generally assume that the capacity of a charging station or a battery replacement station is infinite, allowing unrestricted access by electric vehicles. Meanwhile, the construction cost of one intermediate site is assumed to be a large fixed cost, and the influence of the capacity change of the site on the construction cost is not considered. Therefore, how an enterprise effectively optimizes a delivery scheme and a site selection strategy while considering site capacity design, so as to provide better delivery service for customers and save more resources has become a problem to be solved urgently.
Delivery optimization problems such as express delivery, takeaway, freshness and the like in urban logistics can be collectively referred to as vehicle path problems. The vehicle path problem is a typical NP-hard problem, and at present, the research on solving methods mainly focuses on metaheuristic methods. The self-adaptive large neighborhood search algorithm is taken as a classic meta-heuristic algorithm and is formally proposed by Ropke and Pisinger in 2006 at the earliest. On the basis of the large neighborhood search algorithm, the self-adaptive large neighborhood search algorithm changes the selection strategy of the damage and repair operators, and the damage and repair operators are dynamically selected according to historical performance. Thus, each operator is assigned an initial score, each time it takes an action, the corresponding score is incremented according to the quality of the new solution, and then the selection probability of each operator is updated using the roulette mechanism. The self-adaptive large neighborhood search algorithm is suitable for solving the discreteness problems such as the vehicle path problem and the like. But the original self-adaptive large neighborhood search algorithm cannot optimize the distribution scheme and the site selection strategy at the same time. Therefore, it is necessary to design and improve an adaptive large neighborhood search algorithm so that the adaptive large neighborhood search algorithm is suitable for solving the problem of address selection paths of the electric vehicle power station.
Disclosure of Invention
The invention specifically considers the influence of the power station capacity design on the total cost aiming at the defects of the prior art, solves the problem by adopting an improved self-adaptive large neighborhood search algorithm, and specifically provides an optimization method of an electric vehicle power station site selection path considering the multi-station capacity design, which comprises the following steps:
s10, determining the relation between the power station capacity design and the address selection path strategy;
s20, establishing an electric bicycle battery replacement station site selection path optimization model;
s30, solving algorithm of design model
Wherein, S20 specifically comprises the following steps:
s21, determining a problem target and a constraint condition;
s22, determining symbolic representation of the parameters and the variables;
s23, establishing a mathematical model;
s30 specifically comprises the following steps:
s31, encoding and evaluating the solution;
s32, designing in an initialization stage;
s33, breaking the operator design;
s34, repairing operator design;
s35, heuristic power station selection strategy design is carried out;
and S36, designing a local search strategy.
Preferably, in S10, a relationship between the power conversion station capacity design and the site selection path policy is determined, specifically, the construction cost of a unit power conversion station includes two parts, one part is fixed site lease cost and operation cost, the other part is equipment cost of a power distribution and replacement cabinet and a corresponding number of batteries, and the specific construction cost is as follows:
wherein, C f The construction cost of the candidate power change station located at the node f is calculated; lambda [ alpha ] f Fixed cost for unit BSS construction; lambda b The purchase and operation costs for a unit cell; n is a radical of hydrogen f Representing the number of available batteries of the candidate charging station configuration at node f; y is f And if the binary decision variable is a binary decision variable, if the power change station at the node f is started, the value is 1, and if not, the value is 0.
Preferably, in S21, a problem target and a constraint condition are determined; specifically, the optimization target is that the total cost is minimum, the problem is defined in a single distribution center and directed network G = (V, A), wherein V represents a logistics node set in the network and comprises the distribution center, a group of customer nodes and a group of candidate power station nodes, and A represents an arc set of the distribution network; the requirements and positions of the customers and the positions of the candidate battery replacement stations are known, and meanwhile, the distribution center is provided with a takeout distribution fleet which is provided with electric vehicles of the same model and can meet the requirements of the customers.
Preferably, in S22, determining symbolic representations of the parameters and the variables specifically includes:
preferably, in S23, a mathematical model is established, specifically: and finally establishing a mathematical model for optimizing the vehicle path according to the problem target, the constraint condition and the symbolic representation:
the objective function (1) aims at calculating the total cost, and finds a path with the lowest cost, and the location and configuration of the power conversion station by taking the sum of the construction cost of the power conversion station and the transportation cost of the electric motorcade as a target; constraint (2) ensures that a vehicle visits the client node to be served in each delivery process, and the vehicle leaves after the service is completed; constraint (3) ensures that the traffic of each vehicle at all nodes is balanced, i.e. a vehicle must leave after visiting a node; constraint (4) indicates that the vehicle that departed during each delivery must return to the delivery centre; the constraint (5) indicates that the electric vehicle leaving the distribution center in each distribution process is distributed with a route at most once to serve the corresponding customer; the restraint (6) ensures that the battery of the electric vehicle is only replaced at the position where the power replacement station is built; the constraint (7) limits the number of batteries which should be configured in the power conversion station, and the constraint needs to meet the power conversion requirements of all electric vehicles accessing the power conversion station; constraints (8) ensure that only the established swapping stations will configure the battery; constraint (9) represents a vehicle number limit, meaning that the number of vehicles involved in delivery during each delivery should be less than a predetermined set fleet size, as some vehicles may be scheduled for use; the constraint (10) represents that the sum of the residual load of the electric vehicles entering the power swapping station is equal to the sum of the residual load of the electric vehicles leaving the power swapping station, so that the balance of the residual load after the vehicles visit the power swapping station is ensured, and a vehicle can repeatedly visit the same power swapping station in each distribution process; the method comprises the steps that the constraint (11) sequentially tracks the residual load quantity of the electric vehicle at each node based on the running path of the electric vehicle; the constraints (12) limit the remaining load range of the electric vehicle at each node; the method comprises the steps that the constraint (13) sequentially tracks the residual electric quantity of the electric vehicle at each node based on the running path of the electric vehicle; the constraint (14) indicates that the electric vehicles are all started from the distribution center in a full-power state; constraint (15) indicates that the battery power level is reset to Q after the electric vehicle visits the battery replacement station; constraints (16) ensure that the electric vehicle does not consume battery power while serving the customer site; the constraint (17) ensures that the electric quantity of the battery of the electric vehicle is kept above a safety line in the distribution process, the limitation accords with the practical situation, the service life of the battery can be effectively prolonged, and the special situation can be dealt with; constraints (18), (19) represent the binary and positive integer nature of the decision variables.
Preferably, in S31, encoding and evaluating the solution, the solution of the problem is represented by using a natural number encoding method, and according to an encoding rule, a dimension Dim = N + K +1 of each solution, where N represents the number of customers and K represents the number of vehicles; and adding penalty cost for exceeding vehicle capacity and vehicle mileage on the basis of the objective function.
Preferably, the step S32 of initializing the phase design includes determining the vehicle path by a greedy heuristic method and positioning and allocating the swapping stations by a greedy heuristic swapping station selection method.
Preferably, in S33, the destructive operator is designed, specifically, the destructive operator is divided into two types: the first damage operator is an operator only related to the removal of the customer point, and comprises a random removal operator, a worst separation removal operator, a worst removal operator and a similar removal operator; the second damage operator is an operator for removing the client point and the power station simultaneously, and comprises a path removal operator, a random site removal operator and a site-based removal operator;
and S34, designing a repair operator, wherein the repair operator comprises a distance greedy insertion operator, a cost greedy insertion operator, a regret value insertion operator, a distance greedy insertion operator with noise disturbance and a cost greedy insertion operator with noise disturbance.
Preferably, in step S35, the heuristic power swapping station selection strategy design includes removing unreasonable power swapping station nodes in the solution and inserting power swapping station nodes into the solution, so that all sub paths meet the battery power constraint, and a complete power swapping station selection and distribution decision is obtained.
Preferably, the S36, local search strategy design, includes using an inversion operator and an exchange operator.
The invention has the following beneficial effects:
compared with the prior art, the method provided by the invention has the advantages that when the site selection path of the electric vehicle battery changing station is optimized, the influence of the battery changing station capacity design on the site selection path strategy is considered from the practical application perspective, the site selection path optimization model of the electric vehicle battery changing station is constructed by taking the minimum total cost as a target, and the optimization model is solved by adopting an improved self-adaptive large neighborhood search algorithm.
The invention not only can make logistics enterprises plan the delivery path of the driver more reasonably under the condition of limited resources, help the enterprises to improve the delivery efficiency and save the delivery cost, but also can make the path planning process take account of the site selection and capacity allocation strategy of the power station, and reasonably reduce the investment cost of the enterprises in the aspect of infrastructure. Both aspects have important significance for improving the competitiveness of enterprises.
Drawings
Fig. 1 is a flow chart of an improved adaptive large neighborhood search algorithm in an electric vehicle power change station location selection path optimization method considering multi-station capacity design according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a greedy heuristic power station selection strategy in the optimization method of the site selection path of the electric vehicle power station considering the multi-site capacity design according to the embodiment of the invention;
fig. 3 is a comparison diagram of convergence curves of an improved adaptive large neighborhood search algorithm and other meta-heuristic algorithms of the optimization method for the electric vehicle power change station selection path considering the multi-station capacity design in the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
On the contrary, the invention is intended to cover alternatives, modifications, equivalents and alternatives which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of the present invention, certain specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent to one skilled in the art that the present invention may be practiced without these specific details.
The specific technical scheme of the invention comprises the following steps:
s10, determining the relation between the power station capacity design and the address selection path strategy;
in the takeaway distribution scene, the power change station operates in the form of a power change cabinet. Therefore, the construction cost of the unit power conversion station comprises two parts, one part is fixed site renting cost and operation cost, the other part is equipment cost for distributing and replacing the power cabinet and the corresponding number of batteries, and the specific construction cost is as follows:
wherein C is f For construction cost, λ, of candidate power stations located at node f b Cost per unit cell acquisition and operation, N f Representing the number of available batteries of the candidate charging station configuration at node f.
S20, establishing an optimization model;
s21, determining a problem target and a constraint condition;
the optimization goal of the invention is to minimize the total cost, which is defined in a single distribution center and directed network G = (V, a), where V represents the set of logistics nodes in the network, including the distribution center, a set of customer nodes and a set of candidate swap station nodes, and a represents the arc set in the network. The requirements and positions of the customers and the positions of the candidate battery replacement stations are known, and meanwhile, the distribution center is provided with a takeout distribution fleet which is provided with electric vehicles of the same model and can meet the requirements of the customers. According to the reality, assuming that the take-away delivery clerk works eight hours a day, every hour there will be new customers to be serviced, and the take-away delivery clerk completes one round of order delivery every hour. Assuming that the distribution center is provided with a power changing station and stores a sufficient number of batteries, in each distribution stage, the electric vehicle starts from the distribution center in a full power state and returns to the distribution center after a group of customers are served. Each electric vehicle has a maximum loading capacity, so that the total demand of a vehicle to service a customer must not exceed its maximum loading capacity. During distribution, each electric vehicle needs to keep the residual electric quantity larger than a warning value, and the fully charged battery can be replaced in any power changing station. The replaced battery is assumed to be charged in the battery swapping cabinet and cannot be used continuously on the same day, so that in order to meet the battery swapping requirements of all takeaway distribution staff in one day, the scale of the battery swapping station, that is, the number of batteries configured to the battery swapping cabinet, should be greater than the number of battery swapping times of the electric vehicle in the battery swapping station. The total cost comprises two parts, namely the cost generated by building the battery replacement station at the candidate battery replacement station node and the transportation cost generated by all electric vehicles. The problem is directed to making decisions on vehicle path, site selection of the swapping station and battery storage quantity simultaneously to minimize the total cost and thus meet the needs of all customers.
S22, determining symbolic representation of the parameters and the variables;
in order to facilitate the establishment of the subsequent mathematical model, it is necessary to first explain the parameters and variables involved in the model.
And S23, establishing a mathematical model.
According to the problem targets, constraints and symbolic representations, the invention finally establishes a mathematical model for vehicle path optimization:
the objective function (1) aims to calculate the total cost, and find the path with the lowest cost, the power station location and the configuration by taking the minimum sum of the construction cost of the power station and the transportation cost of the electric motorcade as the target. Constraint (2) ensures that a customer node that needs to be serviced has a vehicle visit during each delivery and the vehicle leaves after the service is completed. Constraint (3) ensures that the traffic per vehicle is balanced at all nodes, i.e. a vehicle must leave after visiting a node. Constraint (4) indicates that the vehicle that was started in each delivery process must return to the delivery center. The constraint (5) indicates that the electric vehicle leaving the distribution center is distributed with a route at most once to serve the corresponding customer in each distribution process. The constraint (6) ensures that the electric vehicle only replaces batteries at the position where a power replacement station is built. The constraint (7) limits the number of batteries to be configured in the battery replacement station, and the battery replacement requirement of all electric vehicles visiting the battery replacement station needs to be met. Constraints (8) ensure that only established swapping stations will configure the battery. Constraint (9) represents a vehicle number limit, meaning that the number of vehicles involved in delivery during each delivery should be less than a predetermined set fleet size, as some vehicles may be scheduled for backup. The constraint (10) indicates that the sum of the residual load of the electric vehicles entering the power swapping station is equal to the sum of the residual load of the electric vehicles leaving the power swapping station, so that the balance of the residual load after the vehicles visit the power swapping station is guaranteed, and a vehicle can repeatedly visit the same power swapping station in each distribution process. The constraint (11) sequentially tracks the remaining load amount of the electric vehicle at each node based on the travel path of the electric vehicle. Constraints (12) limit the range of remaining loads of the electric vehicle at each node. The constraint (13) sequentially tracks the remaining capacity of the electric vehicle at each node based on the travel path of the electric vehicle. The constraint (14) indicates that the electric vehicles are all started from the distribution center in a full-power state. Constraint (15) indicates that the battery power level is reset to Q after the electric vehicle visits the battery swapping station. The constraints (16) ensure that the electric vehicle does not consume battery power while servicing the customer site. The constraint (17) ensures that the battery capacity of the electric vehicle is kept above a safety line in the distribution process, the limitation accords with the practical situation, the service life of the battery can be effectively prolonged, and the special situation can be dealt with. Constraints (18) (19) represent the binary and positive integer nature of the decision variables.
And S30, designing a solving algorithm of the model.
The improved adaptive large neighborhood search algorithm mainly comprises the following five processes:
(1) And in the initial stage, generating an initial feasible solution according to heuristic rules, and initializing all parameters of the algorithm.
(2) And in the damage and repair stage, a group of damage and repair operators are selected according to the self-adaptive rule to update the current solution.
(3) And a greedy heuristic power station selection strategy is used for perfecting the power station selection and capacity allocation decision in the solution.
(4) And an operator weight updating stage, namely adding points to the operators according to the quality of the new solution.
(5) And in the local search stage, a local search operator is introduced to improve the solution.
The flow chart of the algorithm is shown in fig. 1. The method comprises the following specific steps:
s31, encoding and evaluating a solution;
in the present invention, the solution to the siting path optimization problem consists of the travel paths of all vehicles. The invention adopts a natural number coding mode to express the solution of the problem. Assuming that there are 6 customers needing delivery services, there are 2 candidate power stations, and the delivery center has 3 vehicles, the solution to the problem in this case can be represented by the sequence shown in fig. 2. Wherein the value 0 represents the distribution center, 1-6 are the numbers of the customers served by the vehicle, and 7-8 are the numbers of the candidate power stations. The sequence between two values 0 represents the delivery path of one vehicle, the delivery path of 3 vehicles constituting a complete solution to the problem. Thus, vehicle 1 takes path 0-1-3-7-0, vehicle 2 takes path 0-2-4-8-0, and vehicle 3 takes path 0-5-7-6-0. According to the coding rule, dimension Dim = N + K +1 for each solution, where N represents the number of customers and K represents the number of vehicles.
Since the search for a solution is random, the algorithm is likely to generate an infeasible solution. To appreciate the diversity, the present invention allows the existence of infeasible solutions. Therefore, the objective function value cannot be directly used for the evaluation of the solution. The invention chooses to use the formula shown below as the evaluation function, i.e. the penalty cost for exceeding the vehicle capacity and the vehicle mileage is additionally added on the basis of the original objective function. Wherein t is ij Denotes the distance, x, between positions i and j ijk Is a binary variable for indicating whether the vehicle k is onThe over-path (i, j), alpha is the penalty coefficient, y ik Is a binary variable indicating whether the vehicle k visits the client i, q i Indicating the demand of customer i and U indicates the maximum load capacity of the vehicle.
S32, designing in an initialization stage;
the generation of the initial solution is divided into two phases. The first stage determines the vehicle path using a distance greedy heuristic, which is very fast and easy to implement. Starting a vehicle from a distribution center, and sequentially inserting an unserviceable customer point which is closest to the current vehicle point without considering the limit of the driving mileage of the electric vehicle and a power station selection strategy. When the load constraint is not satisfied, the vehicle returns to the distribution center. Another vehicle is then enabled to repeat the above operations until all customer points are inserted into the vehicle path. And in the second stage, based on the path generated in the first stage and the travel mileage constraint of the electric vehicle, a greedy heuristic power station swapping selection method is used for positioning and distributing the power stations, and the method is described in detail in the following steps.
The improved self-adaptive large neighborhood search algorithm adopts a simulated annealing criterion as a standard for judging whether to accept a new solution. Under the standard, the algorithm always accepts the ratio S current More preferred S new But for the ratio S current Worse S new The algorithm also has a certain probability of acceptance. Let f (S) be the objective function of the solution, T 0 Is the initial temperature, T is the temperature in the iterative process, and the acceptance probability of inferior solutions is
After each iteration, the temperature T will gradually decrease, T = ∈ T, ∈ being the rate of decrease in temperature, 0<ε<1. After all N iterations, the method is terminated and S is output best 。
S33-S34, designing a damage and repair operator;
the destroy operator is divided into two types. First kind of destructionOperators are operators only involved in customer point removal, including random removal operators, worst distance removal operators, worst removal operators, and similar removal operators. The second damage operator is an operator for removing the client point and the power swapping station simultaneously, and comprises a path removal operator, a random site removal operator and a site-based removal operator. Destroy operator removes customer number n r Number of clients n c Correlation, n r ＝[λ*n c ]Where λ represents the remove customer coefficient. L is r To deposit the list of customer points removed by the destroy operator, each iteration is initially an empty list.
(1) And (3) random removal: this operator randomly removes n from the current solution S r A customer order and put it in L r . The random selection of the client points is helpful for improving the diversity of algorithm search and expanding the search space of the solution.
(2) Most distant removal: this operator aims to remove the customer points from the current solution S that are the most costly away. The distance cost is defined as the sum of the distances of the current client point i from the previous node and the next node, i.e. discost = d i+1,i +d i-1,i . The worst distance removal operator needs to calculate the discosts of all the customer points in the current solution and delete the customer node i with the highest distance cost * ＝argmax i∈S discost, placing it into L r Repeating this step until n is removed r And (4) each customer point.
(3) Worst removal: this operator removes the most costly customer points from the current solution. Defining the cost of the customer point i in the current solution as Δ cost (S, i) = f (S) -f -i (S), where f (S) is the target value of the current solution, and f -i And (S) is the target value after the customer point i is removed. The worst removal operator needs to compute f (S) first, and then f for each customer point -i (S), so as to obtain the Δ cost (S, i) of each customer point, finally, select the customer point i with the highest cost * ＝argmax i∈S Δ cost (S, i) is removed and placed in L r Repeating this step until n r Individual customer points are removed.
(4) Similar removal: this operator was first proposed by Shaw (1998) to remove a set of predefined basesSimilar customer points are founded. Similarity between clients i and j is expressed as sim (i, j) = α 1 d ij +α 2 |q i -q j |+ α 3 η ij The smaller the sim (i, j) value, the more similar the two points are. Wherein alpha is 1 、α 2 And alpha 3 Is a weight coefficient between 0 and 1, alpha 1 +α 2 +α 3 ＝1。d ij Is the distance, q, between customer point i and customer point j i For the demand of a customer point i, η ij Indicating whether customer point i and customer point j are in the same sub-path, if yes, η ij =1, otherwise η ij And =0. The similarity removal operator firstly randomly removes a client point in the current solution and puts the client point into L r . Then calculating the customer point and L in the current solution r The similarity of the last client point in the list is selected, and the client point with the maximum similarity is selected for removal, i * ＝argmax i∈S sim (i, j), placing it in L r Repeating this step until n r 1 customer site is removed.
(5) Path removal: the operator removes a complete sub-path in the current solution. The path removal operator firstly judges the number of sub paths in the current solution, if the number of the sub paths is larger than 1, one path is randomly selected to be removed, the path comprises client nodes and power station changing nodes, and the removed client nodes can be placed into the L r . This operator is advantageous for deriving solutions that use fewer vehicles.
(6) Random site removal: this operator is an advanced version of the random removal operator. After the random removal operator is executed, a random number of power station swapping nodes are removed. The current solution power station selection strategy is destroyed to a certain extent, so that the solution can be repaired conveniently by a greedy heuristic power station selection strategy, and the understood search space is enlarged.
(7) Based on site removal: the operator removes one swap station and all connected customer points. The method comprises the steps of randomly selecting a built power change station, and removing all nodes of the power change station in the current solution. Then sequentially removing the customer points connected with the power exchange station node and putting the customer points into the L r 。
The repair operators are respectively a distance greedy insertion operator, a cost greedy insertion operator, a regret value insertion operator, a distance greedy insertion operator with noise disturbance and a cost greedy insertion operator with noise disturbance. It will be removed from list L r The client points in (a) are re-inserted into the solution generated by the destruction stage, thereby generating a new solution.
(1) Distance greedy insertion: the operator randomly selects L r And calculates it at solution S destroy Insertion cost for all positions: delta C ij ＝d j,i +d j-1,i -d j,j-1 Wherein j-1 and j are the last node and the next node after the insertion solution of the client i respectively to obtain the optimal insertion cost
Then insert client i into the optimal insertion location and from L r And repeats this step until all deleted clients are inserted into the solution.
(2) Cost greedy insertion: the distance greedy insertion operator accounts for the increase in path cost after the insertion of a customer point. The insertion cost of the operator client point is defined as: delta C ij ＝f +i,j (S) -f (S), where f (S) is the target value of the current solution, f +i,j (S) inserting a target value of the solution after the corresponding position is inserted for the client i. The step of this operator inserting the client point is the same as the distance greedy insertion operator.
(3) Regret-k insertion: this operator is intended to insert the deleted client with the maximum regret value in turn into the optimal location. The insertion cost definition of the customer point is the same as the distance greedy insertion operator, Δ C ij ＝d j,i +d j-1,i - d j,j-1 . The regret value of a client point is defined as:
wherein->
For optimum insertion cost, <' > based on the number of frames>
The operator calculates the remorse value of all the removed client points, and selects the client i with the maximum remorse value * The insertion is carried out to the optimal position,
and from L r In remove customer i * This step is repeated until all deleted clients are inserted into the solution.
(4) Greedy insertion of distance with noise perturbation: the operator is improved on the basis of a distance greedy interpolation operator. The distance greedy insertion operators select optimal positions to insert the clients, and the insertion mode lacks certain diversity and is likely to cause the algorithm to be trapped in local optimization. The operator increases the influence of noise disturbance on the insertion cost, and the insertion cost after the noise disturbance is defined as:
wherein->
u is a noise parameter and r is [ -1,1]The random number in (c). The remaining steps are the same as the distance greedy interpolation operator.
(5) Cost greedy insertion with noise perturbation: the operator is improved on the basis of a cost greedy insertion operator. The operator increases the influence of noise disturbance on the insertion cost of the considered target value, and the calculation formula of the insertion cost after the noise disturbance is the same as the distance greedy insertion operator with the noise disturbance.
S35, designing a greedy heuristic power station selection strategy;
after the break phase and the repair phase, the vehicle path of the solution is changed. However, the battery power constraint is not considered in the repairing stage, so that part of the vehicle paths may not meet the battery power constraint, and the power station selection and distribution strategy for the solution is incomplete. The greedy heuristic algorithm for swapping station selection proposed in this section improves the swapping station selection and allocation decision of the solution, so that the solution meets the constraint of the battery power, and the current solution S is generated current Correspond to each otherNew solution S of new 。
The first stage of the greedy heuristic algorithm for power swap station selection is to remove unreasonable power swap station nodes in the solution. Because the customer site is removed and inserted in the destruction phase and the repair phase, the repaired solutions may have the condition that the same power conversion station is adjacent to each other. Therefore, the operation can repeatedly delete the same adjacent nodes of the power swapping station node until no redundant power swapping station node exists, so as to improve the quality of the solution.
And in the second stage, a power conversion station node is inserted into the solution, so that all sub paths meet the constraint of the electric quantity of the battery, and a complete power conversion station selection and distribution decision is obtained. Defining the set of sub-paths in the solution as R = { R = 1 ,r 2 ,r 3 ,…,r |K| Denoted as r, sub-path k ＝{o,v 1 ,…,v n O' }. For each non-null sub-path r k Each node in the path is required to sequentially judge whether the battery capacity constraint is violated when the vehicle visits. Creating a list
This list is used to store nodes that potentially affect the power station insertion decision at the present time. If a node v violating the constraint is encountered * V. will be * Nodes before and until a swapping station or a distribution center are sequentially joined into->
Then, slave->
Starting with the first node in the system, generating a set of feasible battery replacement stations F v Namely, the vehicle meets the battery power constraint when arriving at the power change station in any set. And if the set of the available battery stations is empty, calculating the next node until the set of the available battery stations of one node is not empty. Definition F v The insertion cost of the medium battery replacement station i after being inserted into the node v is as follows: delta C iv ＝β 1 (d v,i +d v+1,i -d v,v+1 )+β 2 d v+1,i +β 3 ρλ f . Wherein beta is 1 、β 2 And beta 3 Is a weight coefficient between 0 and 1, beta 1 +β 2 +β 3 And =1.v +1 is the node after node v is in the solution. Rho is a binary variable, and is 1 when the power conversion station i is in the state of being addressed in the solution, otherwise is 0. Taking d into account of the insertion cost of the switching station v+1,i The influence of (2) is to avoid that the vehicle goes to a built power change station far away from the node v +1 as far as possible, so that the residual electric quantity of the vehicle reaching the node v +1 is higher, the utilization efficiency of the battery is improved, and the subsequent power change requirement is possibly reduced. From F v Selecting the power change station i with the lowest insertion cost * Inserted into a solution node v and then emptied>
And then, whether the subsequent nodes violate the battery capacity constraint is continuously judged from the inserted power change station node. The above operations are repeated until the nodes of all sub-paths satisfy the battery power constraint.
Designing an adaptive mechanism:
the adaptive mechanism comprises adaptive weight adjustment of the destroy and repair operators and an adaptive selection mechanism of the operators. Define the destroy and repair operator set as A d And A r The operator a ∈ A d ∪A r Fraction of (a) is pi a . At the beginning of the algorithm run, the initial fraction of all operators is π 0 . In one iteration, if a new solution S is generated new Better than global optimal solution S best Then pi of the destruction and repair operator used a Increase of sigma 1 . If S is new Is superior to the current solution S current Then pi of two operators a Increase of sigma 2 . If S is new Inferior to S current But S is new Accepted then σ of two operators a Increase of sigma 3 Otherwise increase σ 4 . Wherein sigma 1 ,σ 2 ,σ 3 ,σ 4 >0. Let
Representative operatora weight on the Nth iteration, <' > based on>
Representing the number of uses of operator a during N iterations. After the adaptive scoring, the algorithm can adaptively adjust the weight of the used operator. The adaptive weight adjustment formula is:
wherein θ ∈ [0,1 ]]The self-adaptive weight adjustment response factor is used for controlling the response speed of the operator effectiveness change to the self-adaptive weight. The operator weights not used in the Nth iteration remain unchanged, i.e.
In each iteration, the algorithm independently selects a destruction operator and a repair operator based on adaptive weights through a roulette mechanism. Given m 1 ＝|A d L destruction operators, with m 2 ＝|A r | repair operators, let |>
Represents the probability of selecting operator a in the Nth iteration, <' >>
The calculation formula of (a) is as follows:
s36, designing in a local search stage;
in the stage, the updated solution is subjected to local search, so that the solution is improved, the convergence speed of the algorithm is increased, and the depth of a search space is enlarged. This stage uses two classical local search operators, the inversion operator and the exchange operator. At the beginning, the algorithm will randomly select one of the two operators to use, and select the overall best improvement solution in the neighborhood.
(1) And (3) inversion operator: the operator operates within the path. Firstly, a non-null sub-path r is randomly selected k ＝{o,v 1 ,v 2 ,…,v n O' then selects two nodes i and j within the path, where i e { v ∈ } 1 ,…,v n-1 J ∈ { i +1, \ 8230;, v ∈ { i +1 } n And then flip all nodes between them. And finally, selecting the solution of the optimal target value under all the value conditions.
(2) And (3) exchanging operators: the operator operates from path to path. First, two non-null sub-paths are randomly selected, r k ＝{o,v 1 ,v 2 ,…,v n O 'and r' k ＝{o,v′ 1 ,v′ 2 ,…,v′ n O', then selects node i and node j, respectively, from the two paths, where i e { v ∈ } 1 ,…,v n J ∈ { v' 1 ,…,v′ n And then exchange the positions of the two nodes. And finally, selecting a solution of the optimal target value under all the value taking conditions.
Results and analysis of the experiments
The experiment content is mainly divided into two parts: the improved adaptive large neighborhood search algorithm is used for verifying the performance of small examples and medium and large examples, and the contents of the two parts correspond to the first embodiment and the second embodiment respectively.
The first embodiment;
in order to verify the performance of the improved self-adaptive large neighborhood search algorithm (IALNS), a commercial solver Gurobi, a large neighborhood search algorithm (LNS) and a variable neighborhood search algorithm (VNS) are selected for comparison experiments. And both the LNS and the VNS adopt operators in the IALNS and a greedy heuristic power station swapping selection strategy to ensure fairness.
The number of customers for the small instances in this embodiment is 5, 8, and 12. We randomly choose six instances from the 25 customer point Solomon instances of different customer distributions, and then extract the corresponding number of customer points from these middle-sized instances, thereby generating 3 × 2 × 3=18 small test cases. The small test cases were solved using our proposed IALNS and the results compared to the optimal solution obtained by the commercial solver Gurobi 9.1.2 using the given mixed integer programming formula. The results show that the IALNS can solve the optimal target value as Gurobi on 18 small test cases. For LNS and VNS, the average difference from the optimal target value solved by Gurobi is 1.51% and 1.58%, and it cannot be guaranteed that the optimal solution is solved for each instance. In addition, gurobi's computation time can increase dramatically with increasing instance size, and in some instances, while optimal solutions can be obtained, the run time exceeds the upper limit. All heuristics can find a solution that is the same as or close to the optimal solution obtained by Gurobi in a short time. The average calculated time for Gurobi, IALNS, LNS and VNS was 436.5s, 2.34s, 2.84s and 3.18s for all small test cases, respectively. In general, the ability of three heuristic algorithms to solve BSS-MD-CLRP in a short time is better than Gurobi, wherein the performance of IALNS is better than that of LNS and VNS. Therefore, these three heuristics will be experimented with on a larger scale example. The comparative figure is shown in fig. 3.
Example two
Since Gurobi takes an extremely long time to solve a larger scale instance, three heuristics IALNS, LNS, and VNS are used to compare to evaluate the performance of IALNS. In this embodiment, three instances are randomly selected from the Solomon instance sets of 25 clients, 50 clients, and 100 clients, respectively, according to the distribution of each client, and 3 × 3=27 instances are generated after modification. Each experiment was repeated ten times, for a total of 27 × 10 × 3=810 experiments. The delivery mode was set to multiple deliveries in the experiment, each delivery round picked 80% of the customers as the target of the required service. In order to ensure the effectiveness of comparison, when different heuristic algorithms are used for performing calculation experiments on the same example, the sets of clients needing to be served in multiple deliveries are the same. The results in the table show that the optimum target values obtained by the IALNS are in most cases better than LNS and VNS. In C101, C109, and C104 of 25 × 5, VNS achieved better target values. For the three sets of test examples, the average difference between the optimal target values for IALNS and LNS were 3.15%, 3.64%, and 2.62%, respectively, and the average difference between the optimal target values for IALNS and VNS was 1.28%, 1.40%, and 2.02%, respectively. The average quality of the solution obtained by the IALNS is better than LNS and VNS, while the IALNS performs better in terms of computation time. Therefore, it can be seen that the IALNS has better solution stability and solution efficiency. Taken together, the performance of the IALNS is superior to both LNS and VNS.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. An electric vehicle power change station site selection path optimization method considering multi-station capacity design is characterized by comprising the following steps:
s10, determining the relation between the power station capacity design and the address selection path strategy;
s20, establishing an electric bicycle battery replacement station site selection path optimization model;
s30, solving algorithm of design model
Wherein, S20 specifically comprises the following steps:
s21, determining a problem target and a constraint condition;
s22, determining symbolic representation of the parameters and the variables;
s23, establishing a mathematical model;
s30 specifically comprises the following steps:
s31, encoding and evaluating a solution;
s32, designing in an initialization stage;
s33, breaking the operator design;
s34, repairing operator design;
s35, heuristic power station selection strategy design;
and S36, designing a local search strategy.
2. The method according to claim 1, wherein, in step S10, a relationship between a power conversion station capacity design and an address selection path policy is determined, specifically, a construction cost of a unit power conversion station includes two parts, one part is a fixed site lease cost and an operation cost, the other part is an equipment cost for distributing and replacing a power cabinet and a corresponding number of batteries, and the specific construction cost is as follows:
wherein, C f The construction cost of the candidate power change station located at the node f is calculated; lambda [ alpha ] f Fixed cost for unit BSS construction; lambda [ alpha ] b The purchase and operation costs for a unit cell; n is a radical of hydrogen f Representing the number of available batteries of the candidate charging station configuration at node f; y is f And if the binary decision variable is a binary decision variable, if the power swapping station at the node f is started, the value is 1, otherwise, the value is 0.
3. The method according to claim 2, wherein, S21, a problem target and a constraint condition are determined; specifically, the optimization goal is to minimize the total cost, the problem is defined in a single distribution center and directed network G = (V, a), where V denotes a set of logistics nodes in the network, including the distribution center, a set of customer nodes, and a set of candidate power station nodes, and a denotes an arc set of the distribution network; the requirements, the positions and the positions of the candidate power stations of the customers are known, and meanwhile, the distribution center is provided with a takeout distribution fleet which is provided with electric vehicles of the same model and can meet the requirements of the customers.
4. The method according to claim 3, wherein the step S22 of determining symbolic representations of the parameters and variables specifically comprises:
5. the method according to claim 4, wherein, in the step S23, a mathematical model is established, specifically: and finally establishing a mathematical model for optimizing the vehicle path according to the problem target, the constraint condition and the symbolic representation:
minimize
subject to
the objective function (1) aims to calculate the total cost, and finds a path with the lowest cost, power station selection and configuration by taking the sum of the construction cost of the power station and the transportation cost of the electric motorcade as the objective; the constraint (2) ensures that a vehicle visits a client node needing to be served in each delivery process, and the vehicle leaves after the service is completed; constraint (3) ensures that the traffic of each vehicle at all nodes is balanced, i.e. a vehicle must leave after visiting a node; constraint (4) indicates that the vehicle that departed in each delivery process must return to the delivery centre; constraint (5) indicates that the electric vehicle leaving the distribution center in each distribution process is distributed with a route at most once to serve the corresponding customer; the restraint (6) ensures that the battery of the electric vehicle is only replaced at the position where the power replacement station is built; the constraint (7) limits the number of batteries which should be configured in the battery swapping station, and the battery swapping requirement of all electric vehicles visiting the battery swapping station needs to be met; constraints (8) ensure that only the established power change stations will configure the battery; constraint (9) represents a vehicle number limit, meaning that the number of vehicles involved in delivery during each delivery should be less than a predetermined set fleet size, as some vehicles may be scheduled for use; constraint (10) indicates that the sum of the residual load of the electric vehicles entering the power swapping station is equal to the sum of the residual load of the electric vehicles leaving the power swapping station, so that the balance of the residual load after the vehicles visit the power swapping station is ensured, and a vehicle can repeatedly visit the same power swapping station in each distribution process; the method comprises the steps that constraint (11) sequentially tracks the residual load amount of the electric vehicle at each node based on the traveling path of the electric vehicle; the constraints (12) limit the remaining load range of the electric vehicle at each node; constraining (13) to sequentially track the remaining capacity of the electric vehicle at each node based on the travel path of the electric vehicle; the constraint (14) indicates that the electric vehicles are started from the distribution center in a full-electricity state; constraint (15) indicates that the battery power level is reset to Q after the electric vehicle visits the battery swapping station; constraints (16) ensure that the electric vehicle does not consume battery power while servicing the customer site; the constraint (17) ensures that the electric quantity of the battery of the electric vehicle is kept above a safety line in the distribution process, the limitation accords with the practical situation, the service life of the battery can be effectively prolonged, and the special situation can be met; constraints (18), (19) represent the binary and positive integer nature of the decision variables.
6. The method according to claim 5, wherein the S31, encoding and evaluating the solution, represents the solution of the problem by using a natural number encoding method, and dimension Dim = N + K +1 of each solution according to the encoding rule, where N represents the number of customers and K represents the number of vehicles; and adding penalty cost for exceeding vehicle capacity and vehicle mileage on the basis of the objective function.
7. The method of claim 6, wherein S32, initializing phase design, comprises determining vehicle paths using a distance greedy heuristic and using a greedy heuristic power station selection method to locate and assign power stations.
8. The method according to claim 7, wherein, in S33, the damage operator design, specifically, the damage operator is divided into two types: the first damage operator is an operator only related to the removal of the customer point, and comprises a random removal operator, a worst separation removal operator, a worst removal operator and a similar removal operator; the second damage operator is an operator for removing the client point and the power station at the same time, and comprises a path removal operator, a random site removal operator and a site-based removal operator;
and S34, designing a repair operator, wherein the repair operator comprises a distance greedy insertion operator, a cost greedy insertion operator, a regret value insertion operator, a distance greedy insertion operator with noise disturbance and a cost greedy insertion operator with noise disturbance.
9. The method of claim 8, wherein, in the step S35, a heuristic power swapping station selection strategy design includes removing unreasonable power swapping station nodes in the solution and inserting power swapping station nodes into the solution, so that all sub paths satisfy the battery power constraint, and a complete power swapping station selection and distribution decision is obtained.
10. The method of claim 9, wherein S36, local search strategy design, comprises using inversion operators and swap operators.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211217714.9A CN115879657A (en) | 2022-09-30 | 2022-09-30 | Electric vehicle power station changing location path optimization method considering multi-station capacity design |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211217714.9A CN115879657A (en) | 2022-09-30 | 2022-09-30 | Electric vehicle power station changing location path optimization method considering multi-station capacity design |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115879657A true CN115879657A (en) | 2023-03-31 |
Family
ID=85770272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211217714.9A Pending CN115879657A (en) | 2022-09-30 | 2022-09-30 | Electric vehicle power station changing location path optimization method considering multi-station capacity design |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115879657A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117132010A (en) * | 2023-09-13 | 2023-11-28 | 东北农业大学 | Vehicle distribution path optimization method based on genetic algorithm |
| CN117522071A (en) * | 2023-12-01 | 2024-02-06 | 谷斗科技(上海)有限公司 | LLM-guided ALNS algorithm-based production scheduling and resource allocation cooperative system |
-
2022
- 2022-09-30 CN CN202211217714.9A patent/CN115879657A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117132010A (en) * | 2023-09-13 | 2023-11-28 | 东北农业大学 | Vehicle distribution path optimization method based on genetic algorithm |
| CN117522071A (en) * | 2023-12-01 | 2024-02-06 | 谷斗科技(上海)有限公司 | LLM-guided ALNS algorithm-based production scheduling and resource allocation cooperative system |
| CN117522071B (en) * | 2023-12-01 | 2024-04-26 | 谷斗科技(上海)有限公司 | LLM (logical Link management) guided ALNS algorithm-based production scheduling and resource allocation cooperative system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764777B (en) | Electric logistics vehicle scheduling method and system with time window | |
| Mehta et al. | Double-layered intelligent energy management for optimal integration of plug-in electric vehicles into distribution systems | |
| CN115879657A (en) | Electric vehicle power station changing location path optimization method considering multi-station capacity design | |
| CN107180274B (en) | Typical scene selection and optimization method for electric vehicle charging facility planning | |
| Li-ying et al. | Multiple charging station location-routing problem with time window of electric vehicle. | |
| Dönmez et al. | The mixed fleet vehicle routing problem with partial recharging by multiple chargers: Mathematical model and adaptive large neighborhood search | |
| Chekired et al. | Fog-computing-based energy storage in smart grid: A cut-off priority queuing model for plug-in electrified vehicle charging | |
| Ma et al. | Rebalancing stochastic demands for bike-sharing networks with multi-scenario characteristics | |
| CN108269008B (en) | Charging facility optimization planning method considering user satisfaction and distribution network reliability | |
| Kumar et al. | A multiobjective planning framework for EV charging stations assisted by solar photovoltaic and battery energy storage system in coupled power and transportation network | |
| CN110309436B (en) | Automobile service network collaborative site selection method and system considering owner selection behavior | |
| CN110796286B (en) | Flexible planning method of power distribution system suitable for electric automobile large-scale application | |
| CN111222705B (en) | Nonlinear charging vehicle path optimization method | |
| Qureshi et al. | Scheduling and routing of mobile charging stations with stochastic travel times to service heterogeneous spatiotemporal electric vehicle charging requests with time windows | |
| CN110189025A (en) | Consider the electric automobile charging station programme acquisition methods that different load increases | |
| CN113553769B (en) | Electric vehicle charging facility optimal configuration method considering charging efficiency | |
| CN111626493B (en) | Charging network planning method considering charging service capacity and operation efficiency | |
| CN111582581B (en) | Electric automobile battery replacement station robust site selection method considering random traffic flow offset | |
| CN113344267A (en) | Logistics network resource allocation optimization method based on cooperation | |
| CN111626492A (en) | Fuzzy multi-target opportunity constraint planning method for electric vehicle charging network | |
| Shen et al. | Integrated optimization of electric vehicles charging location and allocation for valet charging service | |
| Ni et al. | Mobility and energy management in electric vehicle based mobility-on-demand systems: Models and solutions | |
| Liu et al. | Providing levelized cost and waiting time inputs for hdv hydrogen refueling station planning: A case study of us i-75 corridor | |
| CN115018206B (en) | New energy vehicle battery pack charging decision method and device | |
| Al-Alwash et al. | Optimal charging scheme for electric vehicles (EVs) in SDN-based vehicular network, scheduling, and routing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |