A collaborative method for battery swap station site selection and electric vehicle path planning

Abstract

The present invention discloses a collaborative method for battery swap station site selection and electric vehicle route planning. First, based on the relationship between detour probability and the electric vehicle's state of charge, a relationship expression is established between detour probability and accumulated mileage anxiety. Then, an optimization model is established that minimizes the battery swap station construction cost and the electric vehicle's driving cost, as well as the additional cost incurred by the driver's accumulated mileage anxiety throughout the entire journey. Finally, the solution of the optimization model is represented by a battery swap station site selection sequence and an electric vehicle route sequence, and the optimization model is iteratively solved to obtain the optimal battery swap station site selection and electric vehicle route planning scheme. The present invention simultaneously considers the battery swap station site selection issue and the collaborative behavior of both supply and demand parties, and can effectively reduce supply chain costs and improve supply chain efficiency.

Description

Technical Field

[0001] The present application belongs to the technical field of battery swap station site selection and electric vehicle path planning, and in particular relates to a collaborative method for battery swap station site selection and electric vehicle path planning. Background Art

[0002] The rapid development of road transport and urbanization has exacerbated global greenhouse gas emissions, posing significant environmental and economic challenges. Consequently, electric vehicles (EVs), owing to their lower carbon emissions and maintenance costs, have gained widespread adoption in the supply chain, particularly for last-mile delivery. The EV routing problem, an extension of the green car routing problem, aims to design an optimal routing solution for EVs. In recent years, the EV routing problem has been expanded to encompass the battery swap station location-EV routing problem, which simultaneously considers both EV routing and the location of charging facilities.

[0003] Uncertain real-world factors, such as traffic jams and varying driving speeds, can lead to excessive battery consumption in electric vehicles. Drivers may experience anxiety about their electric vehicle's insufficient charge to reach their destination or a battery swap station, a phenomenon known as "range anxiety." Range anxiety reduces the reliability of route planning and incurs additional costs. To alleviate this, drivers may deviate from their customer service routes to reach battery swap stations. This phenomenon, known as "detour behavior," increases transportation costs and reduces logistics efficiency.

[0004] Logistics companies can achieve cost reduction and improved service quality through collaboration between supply and demand during the transportation process. Previous research has considered collaboration across various aspects of the supply chain, such as collaboration between customers and between warehouses. However, the limited range of electric vehicles and the irrational layout of charging facilities make the collaborative problem of battery swap station site selection and electric vehicle routing more complex than traditional vehicle routing. Previous research has not simultaneously considered the uncertainty caused by range anxiety and collaborative behavior. Summary of the Invention

[0005] The purpose of this application is to provide a collaborative method for battery swap station site selection and electric vehicle route planning to overcome the uncertainty caused by drivers' mileage anxiety and reduce the construction cost of battery swap stations and the additional cost caused by mileage anxiety.

[0006] In order to achieve the above objectives, the technical solutions of this application are as follows:

[0007] A collaborative method for battery swap station site selection and electric vehicle route planning, comprising:

[0008] Based on the relationship between the detour probability and the battery status of the electric vehicle, the relationship between the detour probability and the accumulated mileage anxiety is established;

[0009] Establish an optimization model that minimizes the construction costs of battery swap stations and the driving costs of electric vehicles, as well as the additional costs caused by the driver's accumulated mileage anxiety throughout the entire journey;

[0010] The solution of the optimization model is represented by the battery swap station site selection sequence and the electric vehicle path sequence. The optimization model is solved iteratively to obtain the optimal battery swap station site selection and electric vehicle path planning scheme.

[0011] Furthermore, the expression of the relationship between the detour probability and the accumulated mileage anxiety may include:

[0012] A rectangular coordinate system is constructed with the detour probability as the Y-axis and the driving distance as the X-axis, and the area enclosed by the detour probability curve and the X-axis is used as the value of accumulated mileage anxiety.

[0013] Furthermore, the optimization model is expressed by the following formula:

[0014]

[0015] Among them, c B represents the construction cost of the battery swap station, V B represents the set of available battery swap stations; y i represents a binary variable. If the i-th battery swap station is built, it is equal to 1, otherwise it is 0; ε represents the driving cost per unit distance of the electric vehicle, V represents the set of all nodes, V E represents the set of electric vehicles; x ijk represents a binary variable, which is equal to 1 if arc (i, j) is traveled by vehicle k, and 0 otherwise; s ij represents the distance between nodes i and j; A ijk (ζ) represents the cumulative range anxiety of the driver of vehicle k when traveling in arc (i, j); σ represents the additional cost per unit of cumulative range anxiety.

[0016] Furthermore, the optimization model is iteratively solved to obtain the optimal battery swap station location and electric vehicle path planning solution:

[0017] Step F1, initializing the battery swap station site selection sequence and electric vehicle path sequence;

[0018] Step F2: Determine whether the stagnation criterion is met. If so, apply a local search strategy to the currently recorded optimal battery swap station location sequence. Otherwise, use a location update mechanism to update the battery swap station location sequence to obtain a new battery swap station location sequence.

[0019] Step F3: Use the destruction operator and the insertion operator to destroy and repair the electric vehicle path sequence respectively, and generate a new electric vehicle path sequence;

[0020] Step F4: determine whether the new electric vehicle path sequence is better than the currently recorded optimal electric vehicle path sequence. If so, proceed to step F5; otherwise, return to step F3.

[0021] Step F5: Update the weights and scores of the destroy operator and the insert operator;

[0022] Step F6: Using the new electric vehicle route sequence as the currently recorded optimal electric vehicle route sequence, and using the new battery swap station location sequence as the currently recorded optimal battery swap station location sequence;

[0023] Step F7: Determine whether the maximum number of iterations has been reached. If so, stop the iteration and output the optimal battery swap station site selection sequence and electric vehicle path sequence. Otherwise, return to step F2 and continue the iteration.

[0024] Furthermore, the position update mechanism uses the following formula to update the position of the battery swap station site selection sequence:

[0025]

[0026]

[0027] in, represents the speed of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t-th iteration, and rand() is a random number between 0 and 1 that obeys a uniform distribution.

[0028] Furthermore, the initializing the electric vehicle path sequence further includes:

[0029] Based on the distance from the customer to each warehouse, the initial EV route sequence is divided into subsequences corresponding to each warehouse;

[0030] According to the preset rules, nodes are selected from each subsequence and / or battery swap station location sequence and added to the feasible solution of the corresponding warehouse to obtain the feasible solution of each warehouse and form the final initial electric vehicle path sequence.

[0031] Furthermore, according to the preset rules, the following formula is satisfied:

[0032]

[0033] Among them, S n Represents the nth warehouse D n The feasible solution of Indicates the nearest neighbor of the j-th customer node, as the (j+1)-th customer node, B j+1 represents the nearest battery swap station to the (j+1)th customer node, s (j+1)r Indicates the distance from the (j+1)th customer node to the nearest battery swap station B j+1 Distance, s jD Indicates the jth customer node to warehouse D n Distance, s j(j+1) represents the distance between nodes j and j+1, q j represents the electric vehicle power at node j, p j+1 represents the demand of the (j+1)th customer, w j represents the electric vehicle load at node j, represents a binary variable, and e represents the energy consumption rate.

[0034] This application proposes a collaborative method for battery swap station site selection and electric vehicle path planning, which takes into account the site selection problem of battery swap stations and the collaborative behavior of supply and demand parties. The constructed CELRP-BSS-RA model can effectively reduce supply chain costs and improve supply chain efficiency by searching for the optimal site selection of battery swap stations and electric vehicle paths. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of the collaborative method for battery swap station site selection and electric vehicle route planning for this application;

[0036] Figure 2 This is a diagram showing the relationship between the detour probability of this application and the battery status of the electric vehicle and the accumulated mileage anxiety;

[0037] Figure 3 Solve the optimization model flow chart for this application;

[0038] Figure 4 A schematic diagram illustrating an embodiment of the present application;

[0039] Figure 5 To compare the performance of the proposed method with existing methods under different population sizes;

[0040] Figure 6 Comparison of iterative trajectory experimental results between the proposed method and existing methods. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0042] The facility location-vehicle routing problem consists of the classic facility location problem and the vehicle routing problem. Current research on the battery swap station location-electric vehicle routing problem has not simultaneously considered the uncertainty and collaborative behavior brought about by range anxiety. This application aims to propose a collaborative electric location routing problem with battery-swappingstations while considering range anxiety (CELRP-BSS-RA for short). It is worth noting that the goal of this application is not only to optimize the location of battery swap stations and reduce the driving cost of electric vehicles, but also to minimize the additional costs caused by mileage anxiety during travel. Taking into account the collaborative behavior of the supply and demand sides in the supply chain, the detour probability function is used to deal with the uncertainty of mileage anxiety, and a new mileage anxiety function is proposed to evaluate the accumulated mileage anxiety during driving.

[0043] In one embodiment, Figure 1 As shown in the figure, a collaborative method for battery swap station site selection and electric vehicle path planning is proposed, including:

[0044] Step S1: Based on the relationship between the detour probability and the state of charge of the electric vehicle, a relationship expression between the detour probability and the accumulated mileage anxiety is established.

[0045] In order to study the relationship between drivers' detour behavior and accumulated mileage anxiety, this application uses a detour probability function and proposes a new mileage anxiety function.

[0046] The detour probability function takes into account the impact of the electric vehicle's power state on the driver's detour behavior and can be used to quantify the relationship between the probability of detour behavior and range anxiety, as shown in formula (1).

[0047]

[0048] Among them, P ijk (ζ) is the detour probability of vehicle k when traveling on arc (i, j), ζ is a sensitivity coefficient between 0 and 1, which represents the driver's psychological tolerance, τ is the mileage anxiety threshold, s ij represents the distance between nodes i and j, q i and q j Represents the power of the electric vehicle at nodes i and j, e is a fixed energy consumption rate, and Q is the maximum battery capacity of the electric vehicle. When the power of the electric vehicle at the end node j is not less than the threshold τ, that is, q j =q i -es ij When ≥τ, Pijk (ζ) is equal to 0; when the electric vehicle's power is lower than τ during driving, the driver will have P ijk (ζ) The probability of taking a detour to the battery swap station.

[0049] from Figure 2 As can be seen from the upper part of , as the driving distance increases, the electric vehicle's power decreases linearly with a slope of constant e. Figure 2 As can be seen from the lower part, when the travel distance reaches s τ , that is, when the electric vehicle's power begins to drop below the threshold τ, as the distance increases, P ijk (ζ) begins to increase convexly when the electric vehicle's driving distance approaches the electric vehicle's maximum driving distance s max , when P ijk The value of (ζ) is close to 1.

[0050] In addition, considering that the accumulated mileage anxiety can better reflect the accumulation of mileage anxiety during driving than the mileage anxiety at a certain point, this application focuses on the relationship between detour probability and accumulated mileage anxiety, and proposes a new mileage anxiety function through formula (2):

[0051]

[0052] Among them A ijk (ζ) is the cumulative mileage anxiety of the driver of vehicle k in arc (i, j). When the electric vehicle’s battery at the end node j is not less than the threshold τ, A ijk (ζ) is equal to 0; when the power during driving is lower than τ, A ijk (ζ) begins to increase as the distance traveled by the electric vehicle increases. Figure 2 As shown in the lower part, A ijk The value of (ζ) is defined as the area enclosed by the detour probability function and the X-axis.

[0053] Step S2: Establish an optimization model that minimizes the construction cost of the battery swap station and the driving cost of the electric vehicle, as well as minimizing the additional cost caused by the driver's accumulated mileage anxiety during the entire journey.

[0054] The optimization model of this embodiment is shown in formula (3). The objective function consists of three parts. The first and second parts minimize the construction cost of the battery swap station and the driving cost of the electric vehicle, respectively. This is a common objective function in the battery swap station site selection and electric vehicle route planning problem. The last part minimizes the additional cost incurred by the driver's accumulated mileage anxiety throughout the entire trip.

[0055]

[0056] The constraints of the model are as follows:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] Constraint (4) states that each customer is visited only once by an electric vehicle. Constraint (5) ensures that each battery swap station is visited at most once by an electric vehicle. Constraint (6) ensures that the flow at each node is conserved. Constraints (7) and (8) state that customer demand is a non-negative constant and should not exceed the maximum load capacity of the electric vehicle. Constraints (9) and (10) track the state of the electric vehicle's battery charge and ensure that the electric vehicle's charge never falls below zero. Constraint (11) is used to eliminate sub-loops in the path. Finally, constraints (12)-(14) define the range of the decision variables.

[0069] The symbols used in this application are as follows:

[0070] V C Represents a customer collection;

[0071] V D Represents a collection of warehouses;

[0072] V B Represents the set of available battery swap stations;

[0073] V' represents the set of dummy nodes, which are virtual nodes used to establish virtual battery swap station nodes when building the model;

[0074] V represents the set of all nodes, and node refers to all nodes, namely warehouse nodes, customer nodes, battery swap station nodes, and dumb nodes.

[0075] V E Represents a collection of electric vehicles;

[0076] W represents the maximum load capacity of the electric vehicle;

[0077] Q represents the maximum battery capacity of the electric vehicle;

[0078] w i represents the electric vehicle load at node i;

[0079] p i represents the demand of customer i;

[0080] e represents the energy consumption rate (per unit distance);

[0081] q i represents the electric vehicle power at node i;

[0082] s ij represents the distance between nodes i and j;

[0083] ε represents the driving cost per unit distance of electric vehicles;

[0084] c B represents the construction cost of the battery swap station;

[0085] P ijk (ζ) represents the detour probability of vehicle k when traveling on arc (i, j) under range anxiety, and ζ is a non-negative sensitivity coefficient;

[0086] A ijk (ζ) represents the cumulative range anxiety of the driver of vehicle k when traveling in arc (i, j);

[0087] τ represents the range anxiety threshold;

[0088] σ represents the additional cost of anxiety per unit of accumulated mileage;

[0089] x ijk represents a binary variable, which is equal to 1 if arc (i, j) is traveled by vehicle k, and 0 otherwise;

[0090] y i represents a binary variable, which is equal to 1 if the i-th battery swap station is built, otherwise it is 0;

[0091] It represents a binary variable. If vehicle k initially plans to travel along arc (i, j) and eventually travels to the rth battery swap station under the influence of range anxiety, then it is 1, otherwise it is 0.

[0092] Step S3: The solution of the optimization model is represented by a battery swap station site selection sequence and an electric vehicle path sequence, and the optimization model is iteratively solved to obtain the optimal battery swap station site selection and electric vehicle path planning scheme.

[0093] This application aims to search for the optimal location of battery swap stations and the optimal path for electric vehicles based on the optimization model, effectively reducing supply chain costs and improving supply chain efficiency. The process of searching for the optimal location of battery swap stations and the optimal path for electric vehicles is the process of iteratively solving the optimization model. The solution of this application's optimization model is represented by a sequence of battery swap station locations and a sequence of electric vehicle paths.

[0094] This application iteratively solves the optimization model, improves the discrete binary particle swarm algorithm (BPSO), and proposes the EBPSO method to update the battery swap station location sequence. Then, the large neighborhood search algorithm (ALNS) is used to update the electric vehicle path sequence. That is, the EBPSO and ALNS algorithms are combined to propose a new solution method, which is called the HALNS algorithm in this application. Figure 3 As shown, the following steps are included:

[0095] Step F1: Initialize the battery swap station site selection sequence and electric vehicle path sequence.

[0096] In the EBPSO method proposed in this application, like BPSO, a particle population is first initialized, and then an initial particle is randomly selected from it to start subsequent iterative updates. Figure 4 As shown in Figure 1, the particles represent the solution of the battery swap station location sequence. For the battery swap station location sequence, 1 indicates that the corresponding battery swap station needs to be built, and 0 indicates that the corresponding battery swap station does not need to be built. The initial particles selected from the initial particle population are the initial battery swap station location sequence.

[0097] In a specific embodiment, in order to obtain a better initial feasible solution, the present application initializes the electric vehicle path sequence, further comprising:

[0098] Step 1) Divide the initial electric vehicle path sequence into subsequences corresponding to each warehouse based on the distance from the customer to each warehouse.

[0099] The site selection sequence Y of the battery swap station in this embodiment B ={y1,y2,...,y B},y i =1(i∈{1,2,...,B}) indicates that the i-th battery swap station is built, otherwise it is 0. For the path sequence of electric vehicles, X C ={x1,x2,...,x C},x j =c(j∈{1,2,...,C}) means serving the cth customer. Figure 4 As shown in (a), five battery swap stations are built, the other four are not built, and the path sequence of electric vehicles is randomly generated.

[0100] In this application optimization model, the collaborative behavior of the supply and demand sides ensures that the warehouse and electric vehicles can serve different types of customers. Therefore, in order to obtain a better initial feasible solution, in the path sequence X of the electric vehicle C The customer clustering method is used in the model to assign customers to the nearest warehouse. Assuming that there are N warehouses in the model, after using the customer clustering method, X C is divided into several subsequences, namely like Figure 4 As shown in (b), the path sequence of the electric vehicle is divided into three subsequences according to the distance from the customer to each warehouse.

[0101] Step 2) According to the preset rules, select nodes from each subsequence and / or the battery swap station location sequence and add them to the feasible solution of the corresponding warehouse to obtain the feasible solution of each warehouse and form the final initial electric vehicle path sequence.

[0102] The nearest neighbor method (NN) can generate a good initial feasible solution, so it has been widely used in traditional vehicle path planning problems. However, in the optimization model of this application, due to the battery capacity constraints, load capacity constraints and the driver's detour behavior of electric vehicles, the generation of the initial feasible solution is more complicated than the traditional vehicle path planning problem, making the original NN method no longer applicable. Therefore, this application proposes an ENN method, which uses the preset rule formula (15) to update the nth warehouse D n The initial feasible solution S n .

[0103]

[0104] Among them, S n Represents the nth warehouse D n The feasible solution of Indicates the nearest neighbor of the j-th customer node, as the (j+1)-th customer node, B j+1 represents the nearest battery swap station to the (j+1)th customer node, s (j+1)r Indicates the distance from the (j+1)th customer node to the nearest battery swap station B j+1 Distance, s jD Indicates the jth customer node to warehouse D n Distance, s j(j+1) represents the distance between nodes j and j+1, q j represents the electric vehicle power at node j, p j+1 represents the demand of the (j+1)th customer, w j represents the electric vehicle load at node j, represents a binary variable, and e represents the energy consumption rate.

[0105] This application uses the ENN method to select The nearest neighbor of the j-th customer node As the (j+1)th customer node. s (j+1)r Indicates the distance from the (j+1)th customer node to the nearest battery swap station B j+1 Distance, s jD Indicates the jth customer node to warehouse D n When the battery capacity and load capacity constraints are met and the vehicle does not take a detour, Will be added to S n In addition, if vehicle k takes a detour, that is, The vehicle will first visit B j+1 , and then serve customers If the electric vehicle is about to run out of power, that is, the power consumption from the jth customer node to the (j+1)th customer node and then to the nearest battery swap station is greater than the remaining power of the vehicle, that is, e(s j(j+1) +s (j+1)r )>q j , then electric vehicles are serving customers Before, I will visit B first j+1 ; If the load capacity constraint is not met, that is, p j+1 >w j , then the electric car will return to warehouse D n Reload and then serve the customer; if both the battery capacity and load capacity constraints are not met, that is, es jD >q j and p j+1 >w j , electric vehicles must be charged at a battery swap station before visiting warehouses and customers.

[0106] It should be noted that when this step generates the final initial electric vehicle path sequence, the battery swap station node is selected from the battery swap station site selection sequence in the initial particles for insertion.

[0107] from Figure 4 (c) It can be seen that the ENN method generates an initial feasible solution by adjusting the order of serving customers and inserting battery swap station nodes. The initial feasible solution is the electric vehicle path sequence at the beginning of the iteration. The initial feasible solution needs to be generated first, and then the sequence is updated using the destruction / insertion operator based on the initial feasible solution. If the new path sequence is better than the initial feasible solution, the path sequence is updated (that is, the original sequence is replaced by the new path sequence), otherwise it remains unchanged.

[0108] In one embodiment of the present application, according to a preset rule, nodes are selected from each subsequence and / or the battery swap station location sequence and added to the feasible solution of the corresponding warehouse to obtain a feasible solution for each warehouse, including:

[0109] Step F1.1: Input the battery swap station location sequence Y B and the sequence of electric vehicle routes serving the nth warehouse

[0110] Step F1.2: According to the preset rules, or / and battery swap station site selection sequence Y B Select the nearest neighbor customer of the jth node or / and the nearest battery swap station B j+1 Add the initial feasible solution S to the nth warehouse n and change the node from or / and battery swap station site selection sequence Y B Delete in;

[0111] Step F1.3, judgment Is it an empty set? If it is an empty set, output the initial feasible solution S for the nth warehouse. n , otherwise return to step F1.2.

[0112] Thus, the initial electric vehicle path sequence is obtained through the method of this application, and the initial feasible solution S corresponding to all warehouses is obtained. n The final initial electric vehicle path sequence is formed, such as Figure 4 (c) shown.

[0113] Step F2: Determine whether the stagnation criteria are met. If so, apply a local search strategy to the currently recorded optimal battery swap station location sequence. Otherwise, use a location update mechanism to update the battery swap station location sequence to obtain a new battery swap station location sequence.

[0114] The stagnation criterion of this embodiment is: if the performance of the global optimal battery swap station location sequence does not improve after a preset number of iterations iter, the stagnation criterion is reached, where iter is a preset number of iterations, for example iter = 0.1 * maximum number of iterations.

[0115] The basic BPSO algorithm has strong global search capabilities, but its local search capabilities are poor. Therefore, a local search strategy is introduced in the EBPSO algorithm of this application to improve the local search capability. This application adopts three operators: an exchange operator, an insertion operator, and a reversal operator. The exchange operator is used to select and exchange two elements in the location sequence of the battery swap station. The insertion operator is used to randomly select several elements from the location sequence of the battery swap station and insert them to the beginning of the sequence. The reversal operator is used to select a segment from the location sequence of the battery swap station and reverse it. Which operator to select when applying the local search strategy is a relatively mature technology in this field and will not be repeated here. This application adds a local search strategy to the search process to improve the local search capability of the basic BPSO algorithm.

[0116] When the basic BPSO algorithm is iteratively updated, the i-th cell exchange particle at the t-th iteration can use the position vector and velocity vector During the iteration process, each particle adjusts its own speed and position, where the speed is updated according to formula (16) and the position is updated according to formula (17).

[0117]

[0118]

[0119] Among them, ω is the inertia parameter, c1 and c2 are learning factors, r1 and r2 are random numbers between 0 and 1, rand() is a random number between 0 and 1 that obeys a uniform distribution, and Sig(v) represents the sigmoid transformation.

[0120] In the basic BPSO algorithm, the sigmoid transformation used to update the particle position is shown in formula (18):

[0121]

[0122] When the particle approaches the optimal position, the particle's velocity will approach 0 to stabilize the particle's position. However, according to formula (18), if Close to 0, The value of will be close to 0.5, which will cause the search to always have strong randomness, even if an excellent solution has been found. Here, the battery swap station particle is the battery swap station location sequence.

[0123] To this end, in this application, a new position update mechanism is proposed to improve the convergence of the EBPSO algorithm, as shown in formulas (19) and (20):

[0124]

[0125]

[0126] From formula (19), we can see that when the particle velocity approaches 0, The value of also tends to 0. The particle updates its position according to formula (20). Near 0 o'clock, This mechanism makes particles with large velocities more inclined to change their positions, while particles with small velocities are more inclined to remain unchanged, thereby enhancing the global search capability and convergence of the algorithm.

[0127] in, represents the speed of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t-th iteration.

[0128] This application uses an improved BPSO algorithm to update the battery swap station location sequence. It should be noted that in the BPSO algorithm of this embodiment, the battery swap station location sequence is used as a particle of the BPSO algorithm. This step is mainly used to update the battery swap station location sequence, and then the electric vehicle path sequence is updated through the following steps.

[0129] Step F3: Use the destruction operator and the insertion operator to destroy and repair the electric vehicle path sequence respectively, and generate a new electric vehicle path sequence.

[0130] The ALNS algorithm is an extension of the Large Neighborhood Search algorithm, which uses different destruction and insertion operators to destroy and reconstruct the solution, respectively. To obtain a better search neighborhood and solution, the ALNS algorithm assigns an initial weight to each destruction and insertion operator and dynamically adjusts the weight based on the operator's performance.

[0131] This application uses two destruction operators to process the optimization model, including a random destruction operator and a worst-case destruction operator.

[0132] The random destruction operator randomly removes the random min to r max The proportion of customer nodes, where r min and r max are fixed values between 0 and 1. The worst destruction operator first calculates the change Δf in the objective function when the i-th customer is removed from the current solution. i , and then remove from η min to η max The proportion of customer nodes that leads to a significant increase in the objective function, where η min and η max They are fixed values between 0 and 1 respectively.

[0133] This application uses three insertion operators, namely random operator, greedy operator and regret-3 insertion operator, through which the removed customers are reinserted to repair the solution.

[0134] The random insertion operator randomly generates insertion indices for the removed customers and reinserts them to repair the solution. The greedy insertion operator calculates the increment Δf of the objective function when the i-th removed customer is inserted into the j-th position of the current solution. ij , thus inserting the customer to the position where the increment of the objective function is the smallest. tIt is defined as the difference in the objective function when inserting customer p at the tth suboptimal position and the optimal position, that is, c t =f(s t (p))-f(s opt (p)). In the regret-3 insertion operator, three suboptimal positions are considered and c is selected t The insertion index I that maximizes the sum is used as the insertion position of customer p, as shown in formula (21):

[0135]

[0136] How to select the destruction operator and the insertion operator is a relatively mature technology in this field and will not be described here in detail.

[0137] Step F4: determine whether the new electric vehicle path sequence is better than the currently recorded optimal electric vehicle path sequence. If so, proceed to step F5; otherwise, return to step F3.

[0138] The new EV path sequence objective function value is compared with the currently recorded optimal EV path sequence objective function value. If the new EV path sequence objective function value is less than the currently recorded optimal EV path sequence objective function value, the new EV path sequence is considered superior to the currently recorded optimal EV path sequence. During the iteration process, a currently recorded optimal EV path sequence and a currently recorded optimal battery swap station location sequence are always recorded and updated in step F6.

[0139] It should be noted that if Figure 3 As shown, in step F3, the initial electric vehicle path sequence is operated at first, and the solution S is the initial electric vehicle path sequence. If the solution S is updated in step F6, the solution S is operated when step F3 is executed again.

[0140] Step F5: Update the weights and scores of the destroy operator and the insert operator.

[0141] Step F6: Use the new electric vehicle path sequence as the currently recorded optimal electric vehicle path sequence, and use the new battery swap station location sequence as the currently recorded optimal battery swap station location sequence.

[0142] In this step, since the new electric vehicle path sequence is better than the original electric vehicle path sequence, the new electric vehicle path sequence is used to update the currently recorded optimal electric vehicle path sequence.

[0143] Step F7: Determine whether the maximum number of iterations has been reached. If so, stop the iteration and output the optimal battery swap station site selection sequence and electric vehicle path sequence. Otherwise, return to step F2 and continue the iteration.

[0144] It should be noted that steps F3 to F6 are basic steps of the existing ALNS algorithm, which are relatively mature technologies in this technical field and will not be described in detail here.

[0145] This application also analyzes the optimization model and solution method (HALNS algorithm) proposed in this application through a series of experimental data, and compares the HALNS algorithm with four other heuristic baseline algorithms to verify the effectiveness and performance of the HALNS algorithm. First, the HALNS algorithm is compared with a hybrid algorithm that integrates the basic BPSO algorithm and the ALNS algorithm, as well as a hybrid algorithm that integrates the genetic algorithm (GA) and the ALNS algorithm, to verify the effectiveness of the EBPSO algorithm in solving the battery swap station site selection problem. Then, the HALNS algorithm is compared with a hybrid algorithm that integrates the EBPSO algorithm and the iterated local search algorithm (ILS) and the HVNS algorithm to verify the effectiveness of the ALNS algorithm in solving the electric vehicle path planning problem. For ease of representation, the above algorithms are respectively expressed as BPSO-ALNS, GA-ALNS, EBPSO-ILS and HVNS.

[0146] Previous battery swap station site selection-electric vehicle path planning problems did not consider range anxiety and electric vehicle cooperative behavior at the same time. Therefore, the benchmark problem instances used in previous studies cannot be directly applied to the model proposed in this application. Therefore, this application first modified 11 classic instances of vehicle path planning problems with load restrictions according to the characteristics of the optimization model of this application, generated data sets A, B and P, and then randomly generated 14 instances as data set C through computer simulation. The number of warehouses in each instance is 2 to 5. The number of battery swap stations available is 17 to 55. Customer demand has three values, namely 1, 3, and 5. The maximum load capacity of electric vehicles ranges from 16 to 75. The maximum battery capacity of electric vehicles ranges from 42 to 96. The range anxiety threshold is set to 30% of the maximum battery capacity. Each instance is named according to the dataset to which it belongs, the number of nodes, and the number of warehouses. For example, instance A-n34-d3 indicates that the instance belongs to dataset A, which contains 34 nodes and 3 warehouses.

[0147] In order to verify the effectiveness of the HALNS algorithm, this application conducted multiple experiments. First, in order to verify the robustness of the HALNS algorithm in solving the CELRP-BSS-RA model, five algorithms were compared on the "C-n19-d2" instance with a population size of 5 to 15. Figure 5As shown in Figure 3, the HALNS algorithm can obtain better solutions as the population size changes and is more stable than the other four baseline algorithms. This experiment shows that the HALNS algorithm has strong robustness in solving the model proposed in this paper.

[0148] Subsequently, on the “C-n19-d2” instance, the HALNS algorithm was compared with the other four baseline algorithms at different iteration times. The iteration trajectories of the five algorithms are shown in the figure below. Figure 6 As shown. Figure 6 It can be seen that compared with the other four baseline algorithms, the HALNS algorithm converges faster and can obtain the optimal solution in fewer iterations. This experiment shows that compared with other baseline algorithms, the HALNS algorithm has stronger convergence and local search capabilities.

[0149] Finally, the HALNS algorithm was compared with four baseline algorithms on benchmark problem instances. Each of the five algorithms was run 10 times for each instance. Tables 1, 2, and 3 present the experimental results for the five algorithms. Tables 1 and 2 report the optimal value, average, and standard deviation of the objective function. Table 3 reports the average runtime of the five algorithms in seconds. As shown in Tables 1 and 2, the HALNS algorithm outperformed the other four baseline algorithms in 23 of the 25 instances, and the standard deviation was no greater than that of the other baseline algorithms in 22 instances, demonstrating that the HALNS algorithm is more efficient and stable in solving the CELRP-BSS-RA model. Table 3 shows that due to its local search strategy, the HALNS algorithm took longer to solve 17 of the 25 instances than the other baseline algorithms, but the time taken was still within an acceptable range. Furthermore, with the continuous development of computing technologies such as parallel computing and cloud computing, the runtime of the HALNS algorithm can be significantly shortened.

[0150] Table 1 The best value, average value and standard deviation of the five algorithms (Part I)

[0151]

[0152]

[0153] Table 2 The best value, average value and standard deviation of the five algorithms (Part II)

[0154]

[0155] Table 3 Average running time of five algorithms

[0156]

[0157] This application proposes a new CELRP-BSS-RA model for the battery swap station site selection and electric vehicle path planning problem that takes into account both collaborative behavior and range anxiety. This model improves supply chain efficiency through collaborative behavior between supply and demand parties, and proposes a new range anxiety function to solve the range anxiety problem during driving. In order to effectively solve this model, this application proposes a HALNS hybrid algorithm that combines the EBPSO algorithm and the ALNS algorithm. Experimental results show that when solving the CELRP-BSS-RA model, the HALNS algorithm outperforms the other four baseline algorithms.

[0158] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A collaborative method for battery swap station site selection and electric vehicle route planning, characterized in that: The method for coordinating battery swap station site selection and electric vehicle route planning includes: Based on the relationship between the detour probability and the battery status of the electric vehicle, the relationship between the detour probability and the accumulated mileage anxiety is established; Establish an optimization model that minimizes the construction costs of battery swap stations and the driving costs of electric vehicles, as well as the additional costs caused by the driver's accumulated mileage anxiety throughout the entire journey; The solution of the optimization model is represented by a sequence of battery swap station locations and a sequence of electric vehicle routes. The optimization model is iteratively solved to obtain the optimal battery swap station location and electric vehicle route planning scheme. The iterative solution of the optimization model to obtain the optimal battery swap station site selection and electric vehicle path planning solution includes: Step F1, initializing the battery swap station site selection sequence and electric vehicle path sequence; Step F2: Determine whether the stagnation criterion is met. If so, apply a local search strategy to the currently recorded optimal battery swap station location sequence. Otherwise, use a location update mechanism to update the battery swap station location sequence to obtain a new battery swap station location sequence. Step F3: Use the destruction operator and the insertion operator to destroy and repair the electric vehicle path sequence respectively, and generate a new electric vehicle path sequence; Step F4: determine whether the new electric vehicle path sequence is better than the currently recorded optimal electric vehicle path sequence. If so, proceed to step F5; otherwise, return to step F3. Step F5: Update the weights and scores of the destroy operator and the insert operator; Step F6: Using the new electric vehicle route sequence as the currently recorded optimal electric vehicle route sequence, and using the new battery swap station location sequence as the currently recorded optimal battery swap station location sequence; Step F7: Determine whether the maximum number of iterations has been reached. If so, stop the iteration and output the optimal battery swap station site selection sequence and electric vehicle path sequence. Otherwise, return to step F2 and continue the iteration. The position update mechanism uses the following formula to update the position of the battery swap station site selection sequence: in, represents the speed of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t+1th iteration, represents the position of the dth element in the i-th battery swap station location sequence at the t-th iteration, and rand() is a random number between 0 and 1 that obeys a uniform distribution.

2. The method for collaborative site selection of battery swap stations and electric vehicle route planning according to claim 1, characterized in that: The expression of establishing the relationship between the detour probability and the accumulated mileage anxiety includes: A rectangular coordinate system is constructed with the detour probability as the Y-axis and the driving distance as the X-axis, and the area enclosed by the detour probability curve and the X-axis is used as the value of accumulated mileage anxiety.

3. The method for collaborative site selection of battery swap stations and electric vehicle route planning according to claim 1, characterized in that: The optimization model is expressed by the following formula: Among them, c B represents the construction cost of the battery swap station, V B represents the set of available battery swap stations; y i represents a binary variable. If the i-th battery swap station is built, it is equal to 1, otherwise it is 0; ε represents the driving cost per unit distance of the electric vehicle, V represents the set of all nodes, V E represents the set of electric vehicles; x ijk represents a binary variable, which is equal to 1 if arc (i, j) is traveled by vehicle k, and 0 otherwise; s ij represents the distance between nodes i and j; A ijk (ζ) represents the cumulative range anxiety of the driver of vehicle k when traveling in arc (i, j); σ represents the additional cost per unit of cumulative range anxiety.

4. The method for collaborative site selection of battery swap stations and electric vehicle route planning according to claim 1, characterized in that: The initialization of the electric vehicle path sequence further includes: Based on the distance from the customer to each warehouse, the initial EV route sequence is divided into subsequences corresponding to each warehouse; According to the preset rules, nodes are selected from each subsequence and / or battery swap station location sequence and added to the feasible solution of the corresponding warehouse to obtain the feasible solution of each warehouse and form the final initial electric vehicle path sequence.

5. The method for collaborative site selection of battery swap stations and electric vehicle route planning according to claim 4, characterized in that: According to the preset rules, the following formula is satisfied: Among them, S n Represents the nth warehouse D n The feasible solution of Indicates the nearest neighbor of the j-th customer node, as the (j+1)-th customer node, B j+1 represents the nearest battery swap station to the (j+1)th customer node, s (j+1)r Indicates the distance from the (j+1)th customer node to the nearest battery swap station B j+1 Distance, s jD Indicates the jth customer node to warehouse D n Distance, s j(j+1) represents the distance between nodes j and j+1, q j represents the electric vehicle power at node j, p j+1 represents the demand of the (j+1)th customer, w j represents the electric vehicle load at node j, represents a binary variable, and e represents the energy consumption rate.

Images