A method and system for optimizing the dispatch of a vehicle for relocation and charging in a multi-level parking garage

Abstract

The present application relates to static traffic optimization technical field, for a kind of vehicle in three-dimensional parking building shift and charging scheduling optimization method and system.The method includes: establishing multiple data sets;Introduce obstruction matrix and distance matrix;Settable scheduling indicator, decision variable and constraint condition, to maximize continuous parking duration and minimize charging cost as joint target, vehicle scheduling charging problem is modeled as mixed integer linear programming model;The initial scheduling solution of model is solved by branch and bound method;Design large neighborhood search algorithm, identify high interference vehicle by structural disturbance influence function and generate disturbance subset, generate multiple candidate scheduling schemes, comprehensively measure transposition distance, path obstacle cost and charging cost, update solution using greedy repair strategy;Through multiple rounds of disturbance and update, output optimal parking scheduling and charging scheme.The present application can effectively improve parking space utilization, scheduling efficiency and system operation robustness.

Description

Technical Field

[0001] This invention belongs to the field of static traffic management and optimization technology, specifically relating to an optimization method and system for vehicle relocation and charging scheduling within a multi-level parking garage. Background Technology

[0002] With the continuous improvement of the electrification level of urban public transportation, electric buses are gradually becoming the main source of transportation capacity for urban trunk lines. In order to meet the needs of centralized parking and recharging of large numbers of electric buses, multi-story bus parking garages with multi-level structures have been widely built in urban transportation hub areas to improve land utilization and achieve intensive resource management.

[0003] The scheduling optimization problem of multi-level bus parking garages can be summarized as follows: under multi-dimensional constraints (such as vehicle station time, remaining battery power, available parking spaces, charging pile distribution, and physical path obstacles), rationally allocate parking spaces and charging resources for hundreds of buses, organize their orderly entry and exit, and facilitate relocation and recharging, thereby achieving certain optimization objectives (such as minimizing vehicle path conflicts, maximizing parking space utilization, minimizing charging waiting time, and balancing charging resource allocation). This optimization problem directly relates to the garage's throughput efficiency, vehicle operation safety, and energy utilization level, and has significant practical implications for ensuring the efficient operation of the urban transportation system.

[0004] The scheduling and charging optimization problem of multi-level parking garages is essentially an NP-hard combinatorial optimization problem. Given a large number of vehicles, dense time periods, and diverse constraints, finding an accurate solution is extremely difficult. Existing research mostly considers charging scheduling and parking space allocation in planar parking lots or vertical lift parking garages based on static models, failing to effectively account for the dynamic arrival / departure characteristics of vehicles and the limitations on vertical access and occlusion constraints between parent and child parking spaces within the multi-level parking garage. Therefore, it is difficult to adapt to the actual needs of multi-vehicle scheduling in large multi-level parking garages.

[0005] Algorithms for this type of combinatorial optimization problem can be mainly categorized into two types: exact algorithms and heuristic algorithms. Exact algorithms mainly include branch and bound, cutting plane method, and decomposition coordination algorithm. These methods, through rigorous mathematical modeling and branch search mechanisms, can theoretically obtain the optimal solution. However, when the number of vehicles is large and the time intervals are dense, their computational complexity increases dramatically, often making it difficult to complete in real-world time. Heuristic algorithms mainly include particle swarm optimization, tabu search, and simulated annealing. These algorithms typically explore the solution space quickly through randomized perturbations and heuristic repair operations, and can obtain high-quality feasible solutions in a relatively short time, making them suitable for large-scale complex instances. However, their disadvantages include a tendency to get trapped in local optima, and the algorithm's performance depends on the rationality of neighborhood design and parameter tuning.

[0006] Therefore, how to balance vehicle dynamics, charging resource constraints, path conflicts and occlusion relationships in the context of multi-level parking garages, and combine appropriate optimization modeling and solution mechanisms to form a systematic scheduling and charging optimization method, has become a key problem that urgently needs to be solved. Summary of the Invention

[0007] This invention proposes an optimization method and system for vehicle relocation and charging scheduling within a multi-level parking garage. The method models vehicle and energy scheduling within the garage as a mixed-integer linear programming (MILP) model. A branch and bound (B&B) method is used to numerically calculate the MILP model, generating an initial scheduling solution that satisfies all feasibility constraints. Then, a large neighborhood search (LNS) algorithm is introduced, and a perturbation-repair framework is used to dynamically optimize the vehicle scheduling order and parking path. This approach achieves efficient scheduling optimization considering multiple vehicles, multiple time periods, and multiple resource constraints, solving problems such as parking space resource conflicts, path obstruction restrictions, uneven charging resource allocation, and low operating efficiency in existing multi-level parking garage scheduling and charging management systems.

[0008] In an embodiment of the present invention, an optimized method for vehicle relocation and charging scheduling within a multi-level parking garage includes the following steps:

[0009] S1. Establish a basic data structure based on the parking lot situation and divide it into multiple data sets; introduce an obstacle matrix and a distance matrix; set schedulable indicators and decision variables related to parking space allocation and charging, and bind the stable parking judgment, energy replenishment and entry of vehicle scheduling as constraints. With the goal of maximizing continuous parking time and minimizing charging cost, the scheduling and charging planning problem of vehicles in the multi-level parking garage is modeled as a mixed integer linear programming (MILP) model.

[0010] S2. Based on the current parking and charging status, and according to the constraints set in step S1, perform preliminary numerical solutions on the MILP model using the branch and bound method to obtain an initial scheduling solution that satisfies all constraints.

[0011] S3. Design a large neighborhood search algorithm to identify highly disruptive vehicles and generate a disturbance subset through a structural disturbance influence function. Replace and generate multiple candidate scheduling schemes. Use a local optimization objective function to comprehensively measure the swap distance, path obstacle cost, and charging cost. Under parallel feasibility assessment, a greedy repair strategy is adopted to update the solution. Finally, through multiple rounds of disturbance and update, the optimal parking scheduling and charging scheme is output when the termination condition is met.

[0012] In an embodiment of the present invention, an optimized system for vehicle relocation and charging scheduling within a multi-level parking garage, based on the aforementioned optimization method, specifically includes the following modules:

[0013] The model building module constructs a mixed integer linear programming (MILP) model to establish a feasible matching relationship between vehicles and parking spaces; and sets an initial objective function for scheduling optimization, maximizing continuous parking time and minimizing charging costs; and sets constraints including stable parking constraints, energy constraints, and entry determination constraints.

[0014] The initial solution solving module is used to generate a set of initial scheduling solutions that satisfy all constraints using the branch and bound method.

[0015] The scheduling solution optimization module is used to design a large neighborhood search algorithm. It identifies highly disruptive vehicles and generates a disturbance subset through a structural disturbance influence function, and replaces it to generate multiple candidate scheduling schemes. It uses a local optimization objective function to comprehensively measure the swap distance, path obstacle cost, and charging cost. Under parallel feasibility assessment, a greedy repair strategy is adopted to update the solution. Finally, through multiple rounds of disturbance and update, the optimal parking scheduling and charging scheme is output when the termination condition is met.

[0016] As can be seen from the above technical solutions, compared with the prior art, the present invention has the following advantages:

[0017] 1. Based on multi-objective mixed integer linear programming technology, this invention constructs an initial scheduling and charging model for electric vehicles in a multi-level parking garage. Taking into account practical constraints such as vehicle loading and unloading time windows, battery capacity constraints, charging power constraints, parking space usage feasibility and charging priority, it can effectively generate an initial solution scheme with high feasibility and excellent resource utilization.

[0018] 2. This invention introduces a Large Neighborhood Search (LNS) algorithm based on the initial scheduling solution, dynamically adjusting the vehicle scheduling order and parking space allocation through a perturbation-repair framework. Specifically, in each iteration, the system prioritizes identifying and handling highly disruptive vehicles, reducing potential conflicts and resource competition through relocation repair operations. This overcomes the limitations of the initial model's solution accuracy, uncovers better neighborhood solutions, and improves overall scheduling performance and robustness.

[0019] 3. This invention constructs a disturbance subset selection mechanism based on disturbance influence, which prioritizes vehicles with high transposition frequency and dense path conflicts for optimization, thereby focusing disturbance resources and improving local optimization efficiency.

[0020] 4. This invention designs a local optimization objective function to perform a weighted fusion evaluation of factors such as vehicle swapping distance, path conflict degree, and charging cost, so that the disturbance replacement strategy can achieve a balance between path feasibility, system scheduling stability, and energy consumption economy, thereby reducing the impact of scheduling changes on the overall system operation.

[0021] 5. This invention employs a parallel perturbation and greedy insertion repair mechanism. Multiple replacement candidate paths are generated for each vehicle in the perturbation subset, and feasibility is evaluated using a local optimization objective function. Under the premise of satisfying scheduling constraints, the optimal insertion path is selected from the candidate paths based on a greedy criterion, achieving rapid repair after perturbation and updating of the current solution. This mechanism significantly improves the parallel processing efficiency and continuous feasibility of the solution during scheduling optimization, ensuring the stability and convergence speed of the overall scheme.

[0022] 6. This invention sets up a multi-round perturbation and convergence judgment mechanism to terminate the optimization process in a timely manner when the perturbation is invalid, the objective function converges, or the upper limit of the number of rounds is reached, so as to ensure that the algorithm has controllable computational overhead and stable convergence performance under complex constraints. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an optimized method for vehicle relocation and charging scheduling within a multi-level parking garage, as provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of an update scheduling scheme provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram illustrating an obstacle that obstructs the view between parent and child parking spaces, provided in an embodiment of the present invention.

[0026] Figure 4 This is a grid diagram representation of the final complete parking schedule and route arrangement scheme provided in the embodiments of the present invention. The horizontal axis represents the parking space number (column I), and the vertical axis represents the time period within the scheduling cycle (row T). The value in each grid cell represents the vehicle number parked in the corresponding time period and the corresponding parking space. If it is blank, it means that the parking space is vacant during that time period. Detailed Implementation

[0027] The technical solution of the present invention will be further described clearly and in detail below with reference to the embodiments and accompanying drawings. Obviously, the embodiments described below are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example

[0029] Please see Figure 1 , Figure 2 This embodiment provides an optimized method for vehicle relocation and charging scheduling within a multi-level parking garage, comprising the following steps:

[0030] S1. Establish a basic data structure based on the parking lot situation and divide it into multiple data sets; introduce an obstacle matrix and a distance matrix; set schedulable indicators and decision variables related to parking space allocation and charging, and bind the stable parking judgment, energy replenishment and entry of vehicle scheduling as constraints. With the goal of maximizing continuous parking time and minimizing charging cost, the scheduling and charging planning problem of vehicles in the multi-level parking garage is modeled as a mixed integer linear programming (MILP) model.

[0031] The basic data structure established in this step includes parameters such as vehicle set, parking space set, time period set, charging pile set, and maximum charging power. Key decision variables include parking space allocation, charging indication, and battery level, which serve as constraints to ensure stable parking determination, energy replenishment, and entry binding for vehicle scheduling. Furthermore, with the joint objective of maximizing continuous parking time and minimizing charging costs, the scheduling and charging planning problem of vehicles within the multi-level parking garage is modeled as a MILP model.

[0032] In this embodiment, step S1 specifically includes the following steps:

[0033] S11. Establish a basic data structure based on the parking lot situation, and divide the basic data structure into multiple data sets according to the modeling requirements; the multiple data sets include vehicle set, parking space set, time period set, charging pile set, maximum charging power, and maximum battery capacity.

[0034] The basic data structure is based on the actual situation of a comprehensive parking lot in Yantian District, Shenzhen. Taking the actual operation data of this comprehensive parking lot in Yantian District, Shenzhen as an example, the parking lot has 7 floors and is equipped with 140 charging piles, which can simultaneously meet the parking and charging needs of more than 300 electric buses. Considering fire safety and other restrictions, all AC and DC charging piles are concentrated on floors 1 to 3. In this embodiment, given the high consistency of spatial structure and charging pile configuration across different floors, to improve modeling and solution efficiency, the overall problem is divided into representative sub-problems with 6 parking spaces per floor, totaling 42 parking spaces across 7 floors, for modeling and optimized solution.

[0035] In multiple datasets, the vehicle set is denoted as V = {1, 2, ..., N}, where each element represents a vehicle to be scheduled; the parking set is denoted as P = {1, 2, ..., M}, covering all available parking spaces on each floor of the multi-level parking garage; the entire scheduling cycle is divided into multiple discrete time periods, forming a time period set T = {1, 2, ..., H}, to model the parking and movement states of vehicles in different time periods; for each electric vehicle i ∈ V, This indicates the maximum capacity of the vehicle's battery. This indicates the initial charge level of the vehicle's battery; parking spaces with charging capabilities constitute a collection of charging stations. Parking spaces with charging capabilities in the system are divided into two categories: fast charging pile set A and slow charging pile set B, both of which are subsets of charging pile set C. Each parking space p∈A or p∈B corresponds to a maximum charging power parameter per unit time. or Used to describe its charging capability; N A N B These represent the maximum concurrent capacity of Class A charging piles (fast charging piles) and Class B charging piles (slow charging piles), respectively; N all This indicates the total number of charging stations.

[0036] S12. Introduce the obstacle matrix and distance matrix as constraints for MILP. The obstacle matrix is ​​used to indicate whether there is a physical obstacle between any two parking spaces, and the distance matrix is ​​used to quantify the path cost or time cost required for a vehicle to move from one parking space to another.

[0037] To address the challenges of judging driving conflicts and quantifying the cost of switching paths in self-propelled multi-level parking garages, this embodiment introduces two auxiliary matrix structures—an obstacle matrix and a distance matrix—to enhance the model's expressive power in spatial feasibility constraints and path cost modeling.

[0038] Where the obstruction matrix B∈{0,1} M×M This matrix is ​​used to indicate whether there are equipment obstructions or structural obstacles between any two parking spaces. It consists of 0-1 variables. Figure 3 This illustrates a situation where adjacent parking spaces in a parent-child parking system obstruct each other. In this case, the obstruction matrix element B[p,q] = 1, indicating that a vehicle cannot move directly from one parking space to another, reflecting a passage obstacle between the two spaces, such as the inability to move directly from parking space p to parking space q. If the obstruction matrix element is 0, it means there is no obstruction between the two parking spaces, and vehicles can pass directly.

[0039] Distance matrix D∈R M×MThe distance matrix, denoted as D[p,q], is used to quantify the cost of vehicle relocation, specifically the path cost or relocation time cost required for a vehicle to move from one parking space to another. In this embodiment, the Manhattan distance is used as the metric to represent the path cost or relocation cost required for a vehicle to move between different parking spaces, thereby characterizing the spatial movement cost of vehicles during path scheduling and parking space allocation.

[0040] In this embodiment, the obstacle matrix, as one of the auxiliary matrices, is used to quantify the total number of conflicts generated by vehicles on the path in the perturbation influence function, and serves as an important component of the local optimization objective function to measure the obstacle cost index in the scheduling scheme. The obstacle matrix is ​​embedded into the MILP model in the form of the perturbation influence function and the local optimization objective function.

[0041] The distance matrix, as another auxiliary matrix, will serve as an important component of the local optimization objective function, used to measure the path distance cost in the scheduling scheme. The distance matrix is ​​embedded into the MILP model in the form of a local optimization objective function.

[0042] S13. Set a dispatchable indicator, whose value represents whether each vehicle is actually in the dispatch area at each time of the dispatch cycle, and embed it into the MILP model in the form of entry binding constraints and berth capacity feasibility constraints to ensure that each vehicle is in an effective dispatch state during the actual working period.

[0043] To accurately depict the availability status of each vehicle within the scheduling cycle and avoid resource conflicts or scheduling deviations caused by missing status information, this embodiment pre-defines a set of schedulable indicators, denoted as W. i,t ∈{0,1}, used to describe whether each vehicle is actually within the scheduling area at each time point in the scheduling cycle. Specifically, the schedulable indicator W i,t Composed of 0-1 variables, it represents whether vehicle i actually enters the control range of the dispatching system at time t. i,t A value of 1 indicates that vehicle i is in a schedulable state at time t, while a value of 0 indicates that vehicle i is not within the scheduling range at time t and does not participate in scheduling. This avoids vehicles outside the scheduling period being incorrectly included in the scheduling process, reducing resource conflicts and scheduling deviations.

[0044] Furthermore, to simplify boundary scheduling modeling and improve computational efficiency, this embodiment implements preset control over vehicle online status during the boundary periods of the scheduling cycle. Specifically, vehicles numbered in the first 1 / 4 are considered out of the scheduling area at the first and last moments of the entire scheduling cycle; vehicles numbered in the second 2 / 4 are considered offline at the first two moments and the last two moments; and vehicles numbered in the last 1 / 2 are considered online throughout the entire scheduling cycle. This setting is achieved by defining a predefined schedulable indicator W.i,t The value of is selected and embedded into the MILP model in the form of entry binding constraints. This preset rule can also effectively ensure that each vehicle is in an effective scheduling state during its actual working period, preventing resource conflicts and scheduling failures caused by incorrect online status calibration; and can more realistically simulate the operating rules of vehicles with some differences in online and offline times in actual operation, improving the engineering adaptability and simulation rationality of the model.

[0045] In summary, this embodiment introduces three auxiliary modeling structures—obstacle matrix, distance matrix, and schedulable indicator—in the modeling process, which enhances the MILP model's ability to express spatial travel constraints, path costs, and scheduling feasibility states.

[0046] S14. Construct a mixed-integer linear programming (MILP) model for integrated parking and charging scheduling.

[0047] In the constructed MILP model, key decision variables such as parking space allocation, charging indication, power level, time-of-use pricing, and swapping decisions are set to characterize the spatial location, charging behavior, and swapping operations of vehicles within the scheduling cycle. Constraints such as stable parking determination, energy replenishment, and entry binding are introduced to ensure the physical feasibility and logical consistency of the solution. The model aims to maximize continuous parking time and minimize charging costs as a joint objective, thereby improving scheduling efficiency and parking space utilization stability. The establishment of the MILP model specifically includes the following steps:

[0048] S141. Define key decision variables to characterize the vehicle's spatial location, charging behavior, and relocation actions within the parking garage's scheduling cycle. The defined key decision variables specifically include:

[0049] Parking space allocation variable x v,p,t ∈{0,1} indicates whether vehicle v occupies parking space p during time period t. If the value is 1, it means that parking space p is occupied by vehicle v.

[0050] Charging indicator variable y v,t ∈{0,1} indicates whether vehicle v is charging during time period t. If the value is 1, it means that it is charging.

[0051] Electricity level variable s v,t ∈R + , represents the remaining battery level of vehicle v in time period t, used to track its energy changes;

[0052] Static indicator variable s v,p,t ∈{0,1} is used to indicate whether vehicle v remains stationary in parking space p during consecutive time periods t and t+1, which is convenient for statistical analysis of continuous parking behavior;

[0053] Fast charging usage indicator variable Ai,t ∈{0,1}, used to indicate whether vehicle i connects to a Class A charging pile at time t;

[0054] Slow charging occupancy indicator variable B i,t ∈{0,1}, used to indicate whether vehicle i connects to a Class B charging pile at time t;

[0055] Charging difference variable δ i ∈R + , This represents the difference in battery charge that vehicle i failed to make up at the end of the scheduling cycle. It can be introduced into the model as a penalty slack variable to measure the difference when the charging plan fails to meet the full charge requirement.

[0056] Real-time charging power variable r A,t r B,t ∈R + , These represent the actual charging power that Class A and Class B charging piles can provide within time period t, respectively. This variable is dynamically adjusted according to changes in grid load, time-of-use pricing, and power supply strategies to characterize the differences in the power supply capacity of charging piles at different time periods, thereby more accurately reflecting the time-varying characteristics of vehicle charging rates.

[0057] The above variables are the main decision variables in the MILP model constructed in this embodiment, and are used as a whole to support the establishment and solution of the subsequent objective function and scheduling constraints.

[0058] S142. Set multiple objective functions, including maximizing the continuous parking time of a vehicle in the same parking space and minimizing the charging cost under time-of-use pricing conditions, and imposing penalties on vehicles that fail to complete charging before the operating period; combine the multiple objective functions to construct a MILP model.

[0059] In this embodiment, in order to improve the overall operating efficiency of the scheduling system and the stability of parking space usage, and further consider the economy and availability of vehicle charging, the objective function is set as a joint optimization form. This includes maximizing the continuous dwell time of vehicles in the same parking space to reduce the frequency of parking space changes and improve scheduling continuity, as well as minimizing the charging cost under time-of-use pricing conditions, and imposing penalties on vehicles that fail to complete charging before working hours. This achieves a balance between system operating efficiency, economy and vehicle availability, guiding the system to prioritize emergency vehicle needs.

[0060] By jointly modeling the above multiple objective functions, the obtained MILP model can achieve a balance between stability, economy, and usability. The specific expression of the constructed MILP model is as follows:

[0061]

[0062] In the formula, V represents the vehicle set, P represents the parking set, T represents the set of discrete time segments into which the scheduling period is divided, and the set of operating periods for vehicle i. s v,p,t This indicates whether vehicle v remains in the same parking space within two adjacent time periods. A value of 1 indicates that the vehicle remains stationary, while a value of 0 indicates that the vehicle has changed position. v,p,t x v,p,t+1 These indicate whether vehicle v occupies parking space p during time period t and time period t+1, respectively. If the value is 1, it means that parking space p is occupied by vehicle v. This indicates the maximum capacity of the vehicle's battery. Indicates the initial charge of vehicle i's battery; δ i This represents the uncharged battery level of vehicle i, serving as a penalty slack. Parking spaces with charging capabilities are divided into two categories: fast-charging station set A and slow-charging station set B, both subsets of charging station set C. Parking spaces with charging capabilities constitute the charging station set. Each parking space p∈A or p∈B corresponds to a maximum charging power parameter per unit time. or r A,t r B,t N represents the actual charging power that Class A and Class B charging piles can provide within time period t; A N B These represent the maximum concurrent capacity of Class A charging piles and Class B charging piles, respectively; N all The total number of charging stations is represented by w1, w2, and w3, which are weights representing the degree of emphasis on maximizing continuous parking time, minimizing charging costs under time-of-use pricing, and minimizing penalties for not being fully charged, respectively.

[0063] S143. Set multiple constraints for the MILP model.

[0064] In a preferred embodiment of the present invention, the multiple constraints set in step S143 specifically include:

[0065] (1) Set up joint constraints for stable parking determination to achieve accurate modeling of the continuous stable parking state of vehicles and avoid misjudging non-continuous occupation or short-term parking at a certain moment as "stable parking".

[0066] The joint constraints set are as follows:

[0067]

[0068] Among them, s v,p,tLet x be a stability indicator variable for vehicle v remaining in parking space p for two consecutive time periods t and t+1. A value of 1 indicates that the vehicle remains stationary, and a value of 0 indicates that the vehicle has changed position. v,p,t x v,p,t+1 ∈{0,1} indicates whether vehicle v occupies parking space p in time periods t and t+1, respectively. If the value is 1, it means that parking space p is occupied by vehicle v.

[0069] This set of joint constraints ensures that the stability indicator variable s remains constant when the vehicle is not in the designated parking space or has subsequently left the parking space. v,p,t The value can only be 0, thus achieving the exclusion control of non-continuous parking behavior; at the same time, it is guaranteed that the stability indicator variable can be 1 if and only if the vehicle stays in the same parking space p for two consecutive time periods t and t+1. In particular, when the vehicle does not occupy parking space p at any time, joint constraint ① and joint constraint ② will cause the right-hand side to be less than 1, which in turn makes joint constraint ③ require the left-hand side to be less than or equal to 0, eliminating misjudgments caused by intermittent parking and ensuring the accuracy of statistical continuous parking behavior.

[0070] (2) Set soft constraints for energy replenishment to adapt to the charging heterogeneity caused by the existence of various types of charging piles (such as fast charging piles and slow charging piles) and / or different states of charge (SoC) of vehicles before they arrive at the charging station, and guide as many vehicles as possible to complete the required charging tasks within the scheduling cycle.

[0071] The specific details of the energy replenishment soft constraint are as follows:

[0072]

[0073] in, This indicates the initial charge level of the vehicle's battery. Indicates the maximum capacity of the vehicle's battery; r A r B These represent the charging rates per unit time for Class A and Class B charging piles, respectively; A i,t B i,t ∈{0,1}, representing whether vehicle i connects to a Class A charging station and a Class B charging station at time t, respectively; δ i V represents the battery gap of vehicle i at the end of the scheduling period. If the vehicle fails to fully charge, the gap will be recorded and introduced into the objective function as a penalty term, thereby avoiding the model from generating infeasible solutions due to forced full charging; V represents the set of vehicles; T represents the set of time periods.

[0074] The aforementioned soft constraint on energy replenishment can guide as many vehicles as possible to complete their required charging tasks within the scheduling cycle, while allowing a small number of vehicles to fail to fully charge under constrained conditions. The corresponding energy gap is determined by δ. iThe objective function penalty is represented and incorporated to achieve a flexible balance between the feasible solution space and charging demand; at the same time, it ensures that the scheduling system can make full use of the resources of different types of charging piles to realize the dynamic strategy of "hybrid charging path" and allow vehicles to flexibly switch between different types of charging piles according to the availability of parking spaces.

[0075] (3) Set entry binding constraints to ensure that the charging scheduling logic is consistent with the actual arrival status of the vehicle, and prevent the system from arranging for the vehicle to use any type of charging pile before the vehicle has arrived at the parking lot.

[0076] The specific entry binding constraints are as follows:

[0077]

[0078] Among them, A i,t B i,t ∈{0,1}, representing whether vehicle i connects to a Class A charging station and a Class B charging station at time t, respectively; W i,t ∈{0,1} indicates whether vehicle i actually enters the scheduling area at time t; V represents the vehicle set; T represents the time period set.

[0079] This constraint is used to ensure that vehicles are not assigned any type of charging station resource before they actually enter the parking lot, thereby maintaining the physical feasibility and logical consistency of the scheduling scheme.

[0080] (4) Set charging capacity constraints to ensure that the number of vehicles connected to different types of charging piles in any given time period does not exceed the number of their physical available charging piles.

[0081] The specific charging capacity constraints are as follows:

[0082]

[0083] Where, N A N B V represents the maximum concurrent capacity of Class A charging piles and Class B charging piles, respectively; V represents the set of vehicles; and T represents the set of time periods.

[0084] This constraint, by setting capacity limits for Class A and Class B charging piles respectively, reflects the physical limitations of the heterogeneous configuration of charging piles in the parking garage, and avoids resource overflow or vehicle waiting caused by improper allocation of charging pile types during the scheduling process.

[0085] (5) Set real-time charging power constraints to ensure that the output power of various charging piles does not exceed their rated upper limit at any time, thereby maintaining the physical rationality and power security of the charging scheduling scheme while taking into account time-of-use electricity prices and grid load fluctuations.

[0086] To characterize the impact of time-of-use electricity pricing and grid load fluctuations on the output power of charging piles, and to ensure that the model can reflect the dynamic characteristics of charging power at different time periods, the following real-time charging power constraints are set:

[0087]

[0088] Where, r A,t r B,t These represent the actual charging power that Class A and Class B charging piles can provide during time period t, respectively. These represent the maximum output power of Class A and Class B charging piles under design conditions, respectively; T represents the time period set.

[0089] (6) Set parking space capacity feasibility constraints to ensure that all vehicles in the system that are not in operation and are not charging have enough parking spaces without charging function available at any given time period.

[0090] The specific constraints on berth capacity feasibility are as follows:

[0091]

[0092] Among them, A i,t B i,t ∈{0,1}, representing whether vehicle i connects to a Class A charging station and a Class B charging station at time t, respectively; W i,t ∈{0,1}, indicating whether vehicle i actually enters the scheduling area at time t; N all N represents the total number of charging stations; A N B V represents the maximum concurrent capacity of Class A charging piles and Class B charging piles, respectively; V represents the set of vehicles; and T represents the set of time periods.

[0093] This constraint ensures that, at each scheduling time, these vehicles can be legally placed in the limited number of non-charging parking spaces by limiting the number of vehicles that are not working and not participating in charging, thereby preventing the over-occupancy of parking space resources.

[0094] (7) Set operating time constraints to ensure that vehicles must leave the parking garage to perform transportation tasks during their operating time, avoid continuing to occupy parking spaces or connect to charging piles during operating time, and prevent conflicts with parking space allocation and charging resources.

[0095] To accurately depict the vehicle's operational status during transportation tasks and prevent it from continuing to occupy parking spaces or charge during operating hours, the following operating period constraints are set:

[0096]

[0097] Among them, Hi This represents the set of operating periods for vehicle i, meaning that the vehicle needs to be out performing tasks during this period and cannot remain in the parking garage; ∑ p∈P x v,p,t =0 indicates that a vehicle cannot occupy any parking space during operating hours; A i,t =0,B i,t =0 indicates that the vehicle cannot connect to a Class A or Class B charging station during this period.

[0098] This constraint ensures that vehicles must leave the depot to perform tasks during operating hours, avoiding conflicts with parking space allocation and charging resources in the scheduling system, thereby improving the model's fit to actual operating scenarios.

[0099] S2. Based on the current parking and charging status, and according to the constraints set in step S1, perform preliminary numerical solutions on the MILP model using the B&B branch and bound method to obtain an initial feasible solution that satisfies all constraints, i.e., the initial scheduling solution.

[0100] In this embodiment, the initial feasible solution serves as the input basis for subsequent optimization steps, and its generation process specifically includes the following steps:

[0101] S21. First, the MILP model is subjected to linear relaxation (LP Relaxation), that is, the integer constraint of integer variables is temporarily ignored and they are treated as continuous variables, thereby quickly obtaining a relaxed solution for the linear programming problem. This relaxed solution provides a lower bound estimate of the objective function for the B&B branch and bound framework and serves as a reference for branching and pruning operations.

[0102] S22. Within the branch and bound framework, construct a search tree where each node represents a subproblem under a set of variable value constraints. For each node, first solve the model (i.e., the relaxation model) of the subproblem: if all integer variables in the solution satisfy integer property, the corresponding solution is considered feasible; if there are non-integer variables, select one of these variables to branch, generating a new subproblem. Simultaneously, maintain the currently known upper bound (UB) and the explored lower bound (LB) of the optimal feasible solution during the search process. When the lower bound of a node can no longer be better than the current optimal solution, the corresponding branch is pruned to reduce the search scale.

[0103] S23. To control computational overhead and avoid lengthy searches, this embodiment introduces a relative gap (MIP Gap) termination mechanism as a condition for early convergence of the subproblem model solution. Specifically, the relative gap is defined as:

[0104] Gap = (UB - LB) ÷ |UB|

[0105] When the gap value is less than the user-preset threshold ε, the current solution can be considered to have sufficient superiority. Under the premise of ensuring that all constraints are met, the system can terminate the search in advance and output the current optimal feasible solution as the initial feasible solution that satisfies all constraints.

[0106] In this embodiment, considering the large model size and numerous variables in the multi-level parking garage scheduling scenario, the gap threshold ε is set to 0.5 to ensure the rationality of the solution while controlling the computation time. This setting of the threshold ε indicates that if the relative error between the current solution and the explored optimal lower bound LB does not exceed 50% (i.e., the preset threshold ε for the relative gap), the system considers the current solution to satisfy the ε-optimality condition, thereby prematurely terminating the subproblem model solving process.

[0107] S3. Design a Large Neighborhood Search (LNS) algorithm to identify highly disruptive vehicles and generate a disturbance subset through a structural disturbance influence function. Replace the subset to generate multiple candidate scheduling schemes. Use a local optimization objective function to comprehensively measure the swap distance, path obstacle cost, and charging cost. Under parallel feasibility assessment, a greedy repair strategy is adopted to update the solution. Finally, through multiple rounds of disturbance and update, the optimal parking scheduling and charging scheme is output when the convergence threshold, maximum number of iterations, or invalid disturbances are met.

[0108] In this embodiment, the large neighborhood search algorithm is a hybrid heuristic optimization framework, as shown in Table 1. Its core idea is to periodically break the structure of some variables in the current solution (called the "perturbation" stage) to release part of the solution space, thereby avoiding getting trapped in local optima; and to reconstruct high-quality feasible solutions in new neighborhoods with the help of constraint reconstruction and objective function guidance (called the "repair" stage), thereby improving the globality and robustness of the overall solution.

[0109] Table 1. Iterative solution framework for mixed-integer optimization problems based on large neighborhood search.

[0110]

[0111]

[0112] In this embodiment, "perturbation" refers to an optimization strategy that involves making structural modifications to the original scheduling solution to explore a better solution space. Perturbation operations typically generate new candidate solutions by shuffling the parking time, location, or charging decision variables of some vehicles, allowing for subsequent repair and local search to reconstruct and optimize the solution.

[0113] In this embodiment, "high-interference vehicles" refer to individual vehicles that have a significant impact on the overall resource allocation, feasibility, stability, or sensitivity of the objective function in the system scheduling solution. These vehicles typically have characteristics such as "high overlap of parking time with other vehicles, easily causing conflicts" and "long occupation time span, frequent charging".

[0114] In this embodiment, to address the lack of feasibility in the scheduling solution caused by perturbation operations, a "greedy repair" strategy is set up to quickly recover feasible paths in the solution space after perturbation. The core idea of ​​this strategy is to select the path with the smallest local optimization objective function value among all candidate paths that meet the current insertion conditions as the optimal insertion path for the current round.

[0115] In this embodiment, for the initial scheduling solution, a disturbance influence function is first introduced to evaluate all vehicles, identifying highly disturbed vehicles and forming a disturbance subset. Then, while keeping the solution for non-disturbed vehicles fixed, a large neighborhood search (LNS) mechanism is used to perform structural replacement on the disturbed vehicles, perturbing them and generating multiple candidate scheduling schemes. Scheduling reconstruction is then performed based on these candidate schemes. Finally, the candidate schemes are evaluated in parallel, and the candidate solutions are greedily repaired and updated using a local optimization objective function to improve the quality of the scheduling solution within the local disturbance range, prioritizing the scheme with the smallest objective value to update the current solution. This process specifically includes the following steps:

[0116] S31. Design a perturbation subset selection mechanism and introduce a perturbation influence function based on the obstacle matrix to measure the importance of all vehicles to be scheduled in the current initial scheduling solution. The perturbation influence function comprehensively considers the cumulative number of times a vehicle has been swapped in the historical scheduling and the total number of times a conflict has occurred on the path, and sets a balance between the two through a weight coefficient. In this way, key vehicles that have a greater impact on the objective function value in this round of optimization are identified, and a perturbation subset is constructed accordingly.

[0117] The perturbation influence function is constructed as follows:

[0118] ψ(i)=α·freq(i)+(1-α)·conf(i)

[0119] In the formula, freq(i) represents the cumulative number of times vehicle i has been swapped in the historical scheduling; conf(i) is quantified by the obstacle matrix, representing the total number of times vehicle i has conflicted with other vehicles on the path; α∈[0,1] is an adjustable weight coefficient used to set a balance between the frequency of disturbances and the degree of conflict.

[0120] By using this disturbance influence function, the system can comprehensively characterize the "disruption" of vehicles during the scheduling process and prioritize the inclusion of highly disruptive vehicles into the disturbance subset, thereby expanding the spatial search range of feasible solutions, avoiding local solutions from getting stuck in convergence, improving the diversity and exploration capabilities of local re-optimization, and thus improving the pertinence of the disturbance and the overall search efficiency.

[0121] After calculating the disturbance influence function, the system selects a set of high-interference vehicles from the entire set of vehicles to be scheduled based on the calculation results. This set constitutes the disturbance subset required for this round of optimization and is denoted as V. perturb ∈V, where V represents the entire set of objects to be scheduled, V perturb This represents the subset of vehicles whose behavior needs to be rearranged during this round of disturbance.

[0122] S32. In each round of local re-optimization, the large neighborhood search algorithm performs a structural replacement operation on the original scheduling scheme of the vehicles in the determined disturbance subset. Under the premise of keeping the scheduling solution of the non-disturbed vehicles unchanged, the large neighborhood search algorithm attempts to reallocate parking positions for the disturbed vehicles, thereby generating a set of candidate replacement schemes. Each candidate replacement scheme represents a new scheduling solution obtained after local adjustment based on the original solution.

[0123] In its implementation, the large neighborhood search algorithm will target each vehicle i∈V within the perturbation subset. perturb The current solution's parking location information x i The original parking position variable x of the disturbed vehicle is replaced and adjusted by ∈P, thereby generating one or more new alternative solutions. This operation essentially involves changing the original parking position variable x of the disturbed vehicle while keeping the solution for the undisturbed vehicle unchanged. i Replace with a new set of variables x' in the candidate solution space i ∈P, forming a candidate replacement scheme X′={x' i |i∈V perturb}∪{x j |j∈V\V perturb Multiple candidate replacement schemes are generated by setting different perturbation magnitudes, path search rules, or heuristic construction strategies, denoted as C = {X}. (1) ,X (2) ,...,X (k)}, where each X (k) This represents a candidate scheduling solution where only the positions of some vehicles are adjusted based on the original scheduling solution.

[0124] S33. In order to accurately evaluate the merits of scheduling schemes during the disturbance process, a local optimization objective function based on the obstacle matrix and distance matrix is ​​introduced in the scheduling scheme evaluation stage. Multiple optimization objectives, including path distance cost and obstacle cost, are comprehensively calculated to obtain the objective function value corresponding to the candidate scheduling scheme, and this value is used as the metric for evaluating the merits of the scheduling scheme.

[0125] The local optimization objective function is defined as follows:

[0126]

[0127]

[0128] In the formula, d k (i,j) represents the path distance cost for vehicle i to move from its current position to candidate parking space j in the k-th round of perturbation, d k (i,j) are the elements that make up the distance matrix; b k (i,j) represents the obstacle cost on the path, used to measure the congestion or restriction of the path, b k (i,j) constitute the elements of the hindering matrix; Δs v,p,t ΔA i,t ΔB i,t , Δδ i It represents the change of the candidate solution relative to the current solution; w4 and w5 are regularization weights used to balance "global goal improvement" and "local executability"; x v,p,t This indicates whether vehicle v is parked in parking space p at time t.

[0129] In this embodiment, step S33 uses the initial scheduling solution obtained from the MILP model in step S2 through the branch and bound method. Based on this, an updated candidate solution is constructed by rearranging some vehicles after perturbation. During the LNS optimization process, the x-axis of the unperturbed vehicles... v,p,t Keep it unchanged, only regenerate x′ for disturbed vehicles. v,p,t The solution is then substituted into the local optimization objective function for evaluation, thereby comparing the merits of different candidate solutions while ensuring overall feasibility, and gradually improving the quality of the scheduling solution.

[0130] S34. Based on the evaluation of candidate solutions, in order to improve search efficiency and avoid getting trapped in local optima, the perturbation replacement process adopts a parallel mechanism to simultaneously verify the constraints and calculate the objective function value of multiple schemes in the candidate solution set. Under the premise of satisfying all scheduling constraints, the system further adopts a greedy repair strategy to select the candidate solution with the optimal objective function value from the candidate solution set as the update solution for the current round, and completes one LNS local optimization iteration.

[0131] The specific steps are as follows:

[0132] For each candidate solution x v,p,t ∈C, call the local optimization objective function in step S33;

[0133] After obtaining the objective function values ​​corresponding to all candidate solutions, let:

[0134]

[0135] This represents a candidate solution that minimizes the objective function;

[0136] The acceptance mechanism adopts the following criteria:

[0137] like Then accept the solution and update the current solution:

[0138] If the solution is not better than the current solution, then the value of the current solution remains unchanged.

[0139] S35. After completing multiple rounds of disturbance and update operations, when the preset termination conditions are met, output the final optimized vehicle relocation and charging scheduling scheme.

[0140] In this embodiment, based on completing multiple rounds of perturbation and scheduling updates, three types of termination conditions are set to determine whether the optimization iteration should continue: First, based on the number of perturbation rounds, when η consecutive perturbation operations fail to significantly improve the objective function value, it indicates that the current solution has become stable and can be terminated early; second, based on the degree of convergence, when the decrease in the objective function value between the current round and the previous round is less than the convergence threshold ε, it indicates that the optimization has entered the convergence region, and further perturbation has limited benefits; third, based on computational resources, when the number of iterations reaches the preset maximum number of perturbation rounds k... max When this occurs, it is determined that the computing resource constraint has been triggered.

[0141] By triggering any of the aforementioned termination conditions, the system can effectively control computational overhead while maintaining global search capabilities. When a termination condition is met, the system immediately stops the LNS optimization iteration process and outputs the final vehicle relocation and charging scheduling scheme. The final result includes a complete vehicle parking route arrangement and parking schedules for each time period, such as... Figure 4 As shown.

[0142] Based on the same inventive concept, this embodiment also provides an optimized system for vehicle relocation and charging scheduling within a multi-level parking garage. This optimized system is implemented based on the above-mentioned optimization method and specifically includes the following modules:

[0143] The model building module constructs a mixed-integer linear programming (MILP) model to establish a feasible matching relationship between vehicles and parking spaces; and sets an initial objective function for scheduling optimization, maximizing continuous parking time and minimizing charging costs; and sets linear constraints including stable parking constraints, energy constraints, and entry determination constraints.

[0144] The initial solution solving module is used to generate a set of initial scheduling solutions that satisfy all constraints using the branch and bound method.

[0145] The scheduling solution optimization module is used to design a large neighborhood search algorithm. It identifies highly disruptive vehicles and generates a disturbance subset through a structural disturbance influence function, and replaces it to generate multiple candidate scheduling schemes. It uses a local optimization objective function to comprehensively measure the swap distance, path obstacle cost, and charging cost. Under parallel feasibility assessment, a greedy repair strategy is adopted to update the solution. Finally, through multiple rounds of disturbance and update, the optimal parking scheduling and charging scheme is output when the termination condition is met.

[0146] The specific implementation process of each of the above modules is detailed in steps S1-S3, and will not be repeated here.

[0147] In summary, this embodiment first introduces obstacle and distance matrices related to driving within the building, based on parameters such as vehicle set, parking space set, time period set, charging pile set, and maximum charging power. It sets decision variables such as parking space allocation, charging indication, and battery level, and combines constraints such as stable parking, energy replenishment, and entry binding. A mixed integer linear programming model for displacement and charging is established with the joint objective function of maximizing continuous parking time and minimizing charging cost. Secondly, a branch-and-bound method is used to generate initial solutions that satisfy the constraints. Based on this, a large neighborhood search algorithm is introduced to filter highly disruptive vehicles to form a disturbance subset through a disturbance influence function, generating multiple candidate solutions and using a local optimization objective function for parallel evaluation and greedy repair. Finally, through multiple rounds of perturbation and updates, a globally near-optimal scheduling and charging scheme is output. The optimization method and system of this embodiment take into account multiple vehicles, multiple time periods, and multiple resource constraints, effectively improving parking space utilization, scheduling efficiency, and system robustness.

[0148] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An optimized method for vehicle relocation and charging scheduling within a multi-level parking garage, characterized in that, Includes the following steps: S1. Establish a basic data structure based on the parking lot situation and divide it into multiple data sets; introduce an obstacle matrix and a distance matrix; set schedulable indicators and decision variables related to parking space allocation and charging, and bind the stable parking judgment, energy replenishment and entry of vehicle scheduling as constraints. With the goal of maximizing continuous parking time and minimizing charging cost, the scheduling and charging planning problem of vehicles in the multi-level parking garage is modeled as a mixed integer linear programming (MILP) model. S2. Based on the current parking and charging status, and according to the constraints set in step S1, perform preliminary numerical solutions on the MILP model using the branch and bound method to obtain an initial scheduling solution that satisfies all constraints. S3. Design a large neighborhood search algorithm to identify highly disruptive vehicles and generate a disturbance subset through the structural disturbance influence function. Replace and generate multiple candidate scheduling schemes. Use the local optimization objective function to comprehensively measure the swap distance, path obstacle cost and charging cost. Under parallel feasibility assessment, adopt a greedy repair strategy to update the solution. Finally, through multiple rounds of disturbance and update, when the termination condition is met, output the optimal parking scheduling and charging scheme. The obstacle matrix introduced in step S1 indicates whether there is a physical obstacle between any two parking spaces. It is used to quantify the total number of conflicts generated by vehicles on the path in the disturbance influence function in step S3, and to reflect the continuous parking time target in the local optimization objective function, so as to measure the obstacle cost index in the scheduling scheme. The distance matrix introduced in step S1 quantifies the path cost or transfer time cost required for a vehicle to move from one parking space to another. It is used to reflect the continuous parking duration target in the local optimization objective function in step S3, so as to measure the path distance cost index in the scheduling scheme. The steps in step S1 for establishing the MILP model include: Decision variables are set to characterize the spatial location, charging behavior, and relocation actions of vehicles within the parking garage scheduling cycle; Multiple objective functions are defined, including maximizing the continuous parking time of a vehicle in the same parking space and minimizing the charging cost under time-of-use pricing conditions, and penalizing vehicles that fail to complete charging before the operating period; the multiple objective functions are combined to construct a MILP model. Multiple constraints are set for the MILP model, including joint constraints for stable parking determination, soft constraints for energy replenishment, entry binding constraints, charging capacity constraints, real-time charging power constraints, berth capacity feasibility constraints, and operating period constraints. Among them, joint constraints are used to model the continuous and stable parking state of vehicles; Energy replenishment soft constraints are used to adapt to the charging heterogeneity caused by the presence of various types of charging piles and / or different charge states of vehicles before they arrive at the charging station in actual multi-level parking garages, while guiding vehicles to complete the required charging tasks within the scheduling cycle; entry binding constraints are used to ensure that the charging scheduling logic is consistent with the actual arrival state of the vehicle. Charging capacity constraints are used to ensure that the number of vehicles connected to different types of charging piles in any given time period does not exceed the number of their physically available charging piles. Real-time charging power constraints are used to ensure that the output power of various charging piles does not exceed their rated upper limit at any time. Parking capacity feasibility constraints are used to ensure that all vehicles in the system that are not in operation and are not charging have parking spaces without charging capabilities available at any given time. Operating hours constraints are used to ensure that vehicles must leave the parking garage to perform transportation tasks during their operating hours.

2. The optimization method according to claim 1, characterized in that, Step S1 divides the basic data structure into multiple data sets according to modeling requirements; the multiple data sets include vehicle set, parking space set, time period set, charging pile set, maximum charging power, and maximum battery capacity.

3. The optimization method according to claim 1, characterized in that, Step S1 also sets a schedulable indicator, whose value represents whether each vehicle is actually in the scheduling area at each time of the scheduling cycle, and is embedded in the MILP model in the form of entry binding constraints and berth capacity feasibility constraints.

4. The optimization method according to claim 1, characterized in that, The decision variables included are: Parking space allocation variables ∈ {0,1} indicates whether vehicle v occupies parking space p during time period t. If the value is 1, it means that parking space p is occupied by vehicle v. Charging indicator variable ∈ {0,1} indicates whether vehicle v is charging during time period t. If the value is 1, it means that it is charging. Electricity level variables , represents the remaining battery level of vehicle v in time period t, used to track its energy changes; Static indicator variable This is used to indicate whether vehicle v remains stationary in parking space p during consecutive time periods t and t+1, which facilitates the statistical analysis of continuous parking behavior. Fast charging usage indicator variable , used to indicate whether vehicle i is connected to a Class A charging pile at time t; Slow charging usage indicator variable , used to indicate whether vehicle i is connected to a Class B charging pile at time t; Charging difference variable , representing the power difference that vehicle i failed to make up at the end of the scheduling cycle, is introduced into the MILP model as a penalty slack variable to measure the difference when the charging plan fails to meet the full charge requirement; Real-time charging power variable , , representing the actual charging power that Class A charging piles and Class B charging piles can provide within the time period t, respectively.

5. The optimization method according to claim 1, characterized in that, Step S2 includes: S21. Perform linear relaxation on the MILP model to obtain a relaxed solution for a linear programming problem; S22. Under the branch and bound framework, construct a search tree. Each node of the search tree represents a subproblem under a set of variable value constraints. For each node, first solve the model of the subproblem: if all integer variables in the solution satisfy integer property, the corresponding solution is considered a feasible solution; if there are non-integer variables, select one of the variables to branch and generate a new subproblem; at the same time, maintain the upper bound UB of the currently known optimal feasible solution and the lower bound LB of the explored solution during the search process. When the lower bound of a certain node can no longer be better than the current optimal solution, the corresponding branch is pruned. S23. A relative gap termination mechanism is introduced as a condition for early convergence of the subproblem model solution. If the relative error between the current solution and the explored optimal lower bound LB does not exceed the preset threshold of the relative gap, the subproblem model solution process is terminated early.

6. The optimization method according to claim 1, characterized in that, Step S3 includes: S31. Design a disturbance subset selection mechanism, introduce a disturbance influence function based on the obstacle matrix, and measure the importance of all vehicles to be scheduled in the current initial scheduling solution; The disturbance influence function comprehensively considers the cumulative number of times a vehicle has swapped positions in historical scheduling and the total number of times a conflict has occurred on the path, and sets a balance between the two through a weighting coefficient, thereby identifying key vehicles that have a significant impact on the objective function value in this round of optimization, and constructing a disturbance subset accordingly. S32. In each round of local re-optimization, the large neighborhood search algorithm performs a structural replacement operation on the original scheduling scheme of the vehicles in the determined disturbance subset. Under the premise of keeping the scheduling solution of the non-disturbed vehicles unchanged, the large neighborhood search algorithm attempts to reallocate parking positions for the disturbed vehicles, thereby generating a set of candidate replacement schemes. Each candidate replacement scheme represents a new scheduling solution obtained after local adjustment based on the original solution. S33. In the scheduling scheme evaluation stage, a local optimization objective function based on the obstacle matrix and distance matrix is ​​introduced to comprehensively calculate multiple optimization objectives, including path distance cost and obstacle cost, so as to obtain the objective function value corresponding to the candidate scheduling scheme, and use it as a metric for the quality of the scheduling scheme. S34. Based on the evaluation of candidate solutions, the perturbation replacement process adopts a parallel mechanism to simultaneously verify the constraints and calculate the objective function value of multiple schemes in the candidate solution set. Under the premise of satisfying all scheduling constraints, a greedy repair strategy is adopted to select the candidate solution with the optimal objective function value from the candidate solution set as the update solution for the current round, thus completing a local optimization iteration of a large neighborhood search. S35. After completing multiple rounds of disturbance and update operations, when the preset termination conditions are met, output the final optimized vehicle relocation and charging scheduling scheme.

7. The optimization method according to claim 6, characterized in that, The greedy repair strategy is used to quickly recover feasible paths in the solution space after perturbation. The core idea of ​​the greedy repair strategy is to select the path with the smallest joint objective function value among all candidate paths that meet the current insertion conditions, as the optimal insertion path for the current round.

8. An optimized system for vehicle relocation and charging scheduling within a multi-level parking garage, characterized in that, The optimization system is implemented based on the optimization method according to any one of claims 1-7, and includes the following modules: The model building module constructs a mixed integer linear programming (MILP) model and establishes a feasible matching relationship between vehicles and parking spaces. The initial objective function for scheduling optimization is set to maximize continuous parking time and minimize charging cost; linear constraints are set, including stable parking constraints, energy constraints, and entry determination constraints. The initial solution solving module is used to generate a set of initial scheduling solutions that satisfy all constraints using the branch and bound method. The scheduling solution optimization module is used to design a large neighborhood search algorithm. It identifies highly disruptive vehicles and generates a disturbance subset through a structural disturbance influence function, and replaces it to generate multiple candidate scheduling schemes. It uses a local optimization objective function to comprehensively measure the swap distance, path obstacle cost, and charging cost. Under parallel feasibility assessment, a greedy repair strategy is adopted to update the solution. Finally, through multiple rounds of disturbance and update, the optimal parking scheduling and charging scheme is output when the termination condition is met.

Images