Modeling and optimization method for electric motorization crowdshipping system based on multi-agent model

Abstract

The application discloses a kind of electric motorization crowd-sourcing freight system modeling and optimization method based on multi-agent model, the method includes constructing electric motorization crowd-sourcing freight system based on multi-agent model;Electric motorization crowd-sourcing freight system is configured system operation optimization model, it includes for allowing secondary distribution dynamic vehicle load matching optimization model;For realizing the energy management of vehicle fleet, the mixed charging optimization model of fusion emergency charging and preventive charging;And for quantitatively evaluating net income to optimize vehicle capacity allocation index, empty car scheduling optimization model based on real-time expectation calculation;According to system operation optimization model, the operation optimization of electric motorization crowd-sourcing freight system based on multi-agent model is realized.The application solves the technical problems that the existing method has the operation problem caused by the coupling of electric truck endurance limit, charging facility capacity constraint and demand space-time inequality in the highly complex system operation environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban logistics, and in particular to a modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model. BACKGROUND

[0002] In recent years, the rapid growth of urban freight demand and the increasingly stringent requirements for environmental protection constitute the core contradiction of modern urban logistics development. The limitations of traditional freight modes in terms of transport capacity allocation, resource utilization, and environmental impact are increasingly apparent, and there is an urgent need to explore new transportation organization modes. The electric crowd-sourced freight system, as a product of the deep integration of shared economy and green logistics, integrates social idle transport capacity resources and clean energy vehicles, providing a new approach to solving this contradiction. This system uses electric vehicles as transport tools and uses an Internet platform to dynamically match freight demand and transport capacity resources, reducing logistics costs and improving delivery efficiency while effectively reducing carbon emissions and noise pollution, meeting the strategic needs of urban sustainable development.

[0003] However, compared with traditional freight systems, the electric crowd-sourced freight system needs to coordinate multiple factors such as the spatio-temporal distribution of freight demand, the endurance constraints of electric vehicles, the capacity limitations of charging facilities, and the dynamic allocation of transport capacity resources. These factors are coupled and dynamically changing, forming a highly complex system operating environment. There are operational difficulties caused by the coupling of electric truck endurance limitations, charging facility capacity constraints, and uneven demand spatio-temporal distribution. Therefore, it is necessary to develop a modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model in the complex environment of an electric crowd-sourced freight network, which has important practical significance and application value. SUMMARY

[0004] The present application provides a modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model to overcome the above technical problems.

[0005] To achieve the above purpose, the technical solution of the present application is:

[0006] A modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model, specifically comprising the following steps:

[0007] S1: Constructing an electric crowd-sourced freight system based on a multi-agent model;

[0008] And the construction method of the electric crowd-sourced freight system specifically comprises the following steps:

[0009] S11: Constructing a simulation spatio-temporal network of the electric crowd-sourced freight system based on a discretization method;

[0010] S12: Defining an agent model according to the simulation spatio-temporal network;

[0011] Furthermore, the intelligent agent model includes at least a cargo agent, an electric vehicle agent, and a charging station agent;

[0012] S13: Constructing interaction rules for an electric crowdsourced freight system based on an agent model;

[0013] Furthermore, the interaction rules should include at least vehicle-cargo matching rules, charging scheduling rules, and empty vehicle scheduling rules;

[0014] S2: Configure the corresponding system operation optimization model for the electric crowdsourced freight system;

[0015] The system operation optimization model includes:

[0016] A dynamic vehicle-cargo matching optimization model that allows for secondary allocation; a hybrid charging optimization model that integrates emergency charging and preventative charging for fleet energy management; and an empty vehicle scheduling optimization model based on real-time expectation calculation for quantitatively evaluating net benefits to optimize vehicle capacity space configuration indicators.

[0017] S3: Based on the system operation optimization model, optimize the operation of the electric crowdsourced freight system based on the multi-agent model.

[0018] Furthermore, the method for constructing the simulated spatiotemporal network in S11 is as follows:

[0019] S110: Considering the operational time dimension, the system operation period is evenly divided into a discrete time step set T, where T = {0, 1, ..., t... |T|}; T = 0 represents the simulation initial time of system operation;

[0020] S111: Considering the operational space dimension, a hierarchical grid method is adopted to divide the system operation area into several non-overlapping functional sub-regions B={1,...i,j}, and each functional sub-region is defined as the basic spatial unit of the agent's activity area. Each functional sub-region is further gridded to obtain a spatial grid, and the center point of the spatial grid is used as the positioning reference of the agent. At the same time, Manhattan distance is used to measure the spatial distance between agents to simulate the path spatial location characteristics of the urban road network.

[0021] S112: Based on the discrete time step set and spatial distance, the simulation spatiotemporal network of the electric crowdsourced freight system is constructed.

[0022] Furthermore, the agent model defined in S12 includes:

[0023] Goods Agent: Define the set of goods agents as Q, where Q = {q1, q2, ..., q} N}, define the characterization of each cargo qi The attribute tuple of the feature is A q And A q = (S q , R q , L q , D q , T q , W q ) ;

[0024] Wherein S q ∈{0,1,2,3,4} represents the state of goods: 0 represents the generated goods; 1 represents the goods to be taken; 2 represents the goods in transit; 3 represents the goods that have arrived; 4 represents the goods that have been withdrawn due to overtime; R q represents the record demand goods generation time; L q represents the starting point of the goods position; D q represents the end point of the goods position; T q represents the transportation distance; W q represents the cumulative waiting time;

[0025] Electric vehicle Agent: define the set of electric vehicle Agent as Z and Z = {z1, z2...z N}, define the attribute tuple of each electric vehicle z i is A z And

[0026] Wherein S z ∈{0,1,2,3,4,5,6,7} represents the operation state of electric vehicle: 0 represents the idle state; 1 represents the state of going to take goods; 2 represents the state of in transit; 3 represents the state of empty car waiting for scheduling; 4 represents the state of going to charge; 5 represents the state of charging; 6 represents the state of waiting for charging; 7 represents the state of exiting the system; L z represents the current position; D z represents the destination; represents the remaining power at t time; C z represents the battery capacity at t time; T z represents the maximum range; H z represents the distance from the destination; V z represents the speed of travel;

[0027] Charging station Agent: define the set of charging station Agent as F and F = {f1, f2...f N}, define the attribute tuple owned by each charging station f i is A f And A f = (S f , L f , N f , Mf ,G f ,V f );

[0028] wherein S f ∈{0,1} represents the available state of the charging station: 0 represents available; 1 represents unavailable; L f represents the geographical location; N f represents the total number of charging piles; M f represents the available number of charging piles; G f represents the waiting charging queue; V f represents the charging power.

[0029] Further, the interaction rules about the electric crowd-sourced freight system are constructed based on the agent model in S13, including

[0030] vehicle-freight matching rules:

[0031] The demand validity verification is performed on the vehicle-freight matching, that is, confirming that the freight starting point and the ending point are located in different system operation areas, and confirming that there is an available idle electric vehicle based on the electric vehicle Agent; when the verification is passed, the matching scheme of dynamically adjusting the vehicle-freight matching based on real-time information is obtained according to the dynamic vehicle-freight matching optimization model, and the vehicle-freight state updating rule is triggered after confirming the matching scheme;

[0032] The vehicle-freight state updating rule: the electric vehicle is converted from S z =0 to S z =1, the destination is updated to the freight location D z =L q ; the electric vehicle distance to the destination is updated to H zi ; the freight state is converted from S q =0 to S q =1; when the electric vehicle arrives at the pickup point and completes the loading, the electric vehicle state is updated to S z =2; the destination is updated to the freight destination D z =D q ; the distance is updated to T q ; the freight state is updated to S q =2; when the freight arrives at the destination, the electric vehicle state is restored to S z =0; the destination is reset to empty; the location is updated to L z =D q ; the electric vehicle power is updated according to the formula ; the freight state is updated to S q =3; if the freight waiting time exceeds the preset threshold, the freight state is updated to S z =4 and exits the system;

[0033] charging scheduling rules:

[0034] Based on the hybrid charging optimization model of fusing emergency charging and preventive charging, the set of vehicles needing charging is screened at each time step, and the optimal vehicle and charging station matching pair is determined for the charging station assignment;

[0035] And after confirming the optimal vehicle and charging station matching pair, the charging dispatching update rule is triggered;

[0036] The charging dispatching update rule updates the state of the vehicle needing charging to S z =4; the destination is set to the charging station location D z =L f ; the distance is updated to H zf ; the vehicle's location after arriving at the charging station is updated to L z =L f , and the current remaining power is recorded at the same time; according to the charging station Agent, it is confirmed whether the charging station has available charging piles; if the charging station has available charging piles, i.e. M f =0, the vehicle immediately starts charging, and the state of the vehicle is updated to S z =5; otherwise, it is added to the waiting queue G f , and the state is updated to S z =6; the charging duration of the vehicle needing charging is calculated according to the formula ; when the charging is completed, the number of available charging piles M f of the charging station is increased by 1; the state of the vehicle is restored to S z =0; the power of the vehicle is updated to full power ; the location of the vehicle remains at the charging station L z =L f ;

[0037] Empty vehicle dispatching rule:

[0038] The preset monitoring module monitors the continuous idle duration of the idle vehicle in the electric vehicle Agent, i.e. S z =0, and judges whether it exceeds the preset duration threshold μ; if not, no operation is performed; if yes, the empty vehicle dispatching optimization model based on real-time expectation calculation is triggered for vehicle dispatching;

[0039] And after confirming the dispatched vehicle, the empty vehicle dispatching update rule is triggered;

[0040] The state of the dispatched vehicle is updated to S z =3; the destination is set to the target area center D z =i; after the dispatched vehicle arrives at the dispatching point, the state of the dispatched vehicle is restored to S z =0; the power of the dispatched vehicle is updated according to the formula ; the location of the dispatched vehicle is updated to L z= i; determining whether the non-occupancy time ratio of the scheduled vehicle in a unit time exceeds a preset time ratio threshold; if yes, the system determines that the scheduled vehicle leaves, and updates the state of the scheduled vehicle to S z = 7 and removes it from the preset active fleet queue; otherwise, no processing is performed.

[0041] Further, the dynamic vehicle-load matching optimization model for allowing secondary allocation in S2 includes a vehicle-load matching objective function and a first constraint condition:

[0042] The expression of the vehicle-load matching objective function is

[0043]

[0044] In the formula, R', Z' represent defined extended sets; R U represents a set of unallocated loads; R A represents a set of loads that have completed preliminary matching but for which the vehicle has not yet arrived at the pickup point; Z I represents a set of idle vehicles; Z P represents a set of in-transit vehicles that have completed preliminary matching but for which the vehicle has not yet arrived at the pickup point; U qz represents matching utility; x qz represents a decision variable of whether load q is matched with vehicle z and x qz ∈ {0, 1}; W q represents a waiting time normalization value of load q; H qz represents vehicle-load distance; λ q represents a vehicle-load distance weight coefficient; λ w represents a load waiting time weight coefficient for balancing timeliness and economy; λ p represents a penalty weight coefficient; δ represents a secondary allocation penalty coefficient; q z represents a decision variable of whether vehicle z has completed matching; y qz represents a decision variable of determining whether electric vehicle z has completed matching with load q at the previous time step;

[0045] The first constraint condition includes

[0046] An allocation uniqueness constraint that ensures that unallocated loads must be allocated:

[0047]

[0048] In the formula, a q represents an identification variable of whether a load has been allocated;

[0049] A re-allocation times constraint that limits a load to be re-allocated at most once:

[0050]

[0051] wherein: b q is a decision variable indicating whether the identified goods have been reassigned;

[0052] Energy feasibility constraint that guarantees the remaining energy of the vehicle is sufficient to reach the nearest charging station after completing the transportation task:

[0053]

[0054] wherein: is the current remaining mileage of the vehicle; H qz is the distance from the vehicle to the goods; H ij is the transportation mileage of the goods; H zf is the distance to the nearest charging station.

[0055] Further, the hybrid charging optimization model for implementing the integrated emergency charging and preventive charging of the vehicle fleet in S2 includes a hybrid charging objective function and a second constraint condition;

[0056] The expression of the hybrid charging strategy objective function is

[0057]

[0058] T zf = t zf + t zw + t zc

[0059]

[0060]

[0061] wherein: min Z represents a charging station allocation planning model with the goal of minimizing the total time of charging activities; Z c represents the final set of vehicles to be charged; represents the emergency charging queue; Z norm represents the set of idle vehicles whose energy is higher than the preset emergency threshold; represents the set of vehicles whose energy is lower than the preset emergency threshold θ and joined the preventive charging queue according to the stratified threshold function N; avg E represents the average energy of the idle vehicle fleet; n1, n2 represent the set of vehicles determined according to the stratified threshold function N and n ∈ n1, n2; x zf represents a decision variable indicating whether vehicle z is allocated to charging station f and x zf ∈ {0, 1}; T zf represents the charging activity time; t zf represents the driving time; t zwrepresents the queuing waiting time; t zc represents the actual charging time;

[0062] The second constraint is to ensure that the remaining battery of the vehicle is sufficient to reach the designated charging station, which is a unique allocation constraint and an energy accessibility constraint:

[0063]

[0064] Further, the S2-based real-time expected calculation-based empty vehicle scheduling optimization model for quantitatively evaluating the net benefit optimization of the space configuration of the vehicle is expressed as

[0065]

[0066] E za = p t · t at -p s · t as -p r · T za

[0067] In the formula, p za represents the selection probability of the operating area where the vehicle is guided to perform empty vehicle scheduling; C za represents the selection intensity after the conversion of the expected value E za ; E za represents the expected net benefit of the vehicle z scheduled to the operating area a; p t · t at represents the expected service income; p t represents the unit distance service price; t at represents the average service distance of the operating area a; p s · t as represents the waiting cost; p s represents the unit time opportunity cost; t as represents the average matching waiting time of the area a; p r · T za represents the scheduling cost; p r represents the unit time driving cost; T za represents the vehicle empty driving time.

[0068] Beneficial effects: The application provides a modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model. The method realizes information closed loop and constraint consistency of "vehicle and cargo matching, charging scheduling, and empty vehicle scheduling" in a unified discrete space-time and multi-agent framework, avoids infeasibility and suboptimality problems caused by fragmented optimization of the electric crowd-sourced freight system, reduces the risk of cargo waiting and overage, improves service coverage and completion rate through a dynamic vehicle and cargo matching optimization model for allowing secondary allocation, minimizes the total energy supplement time of "traveling + queuing + charging" through a hybrid charging optimization model for realizing integrated emergency charging and preventive charging for vehicle fleet energy management, thereby relieving peak congestion and improving site load balancing, and the empty vehicle scheduling optimization model based on real-time expectation calculation can inhibit operation area clustering, reduce secondary empty driving, and optimize the allocation of transport capacity space through empty vehicle scheduling considering expected net income and probabilistic selection. The application can endogenously evolve SOC and achieve station accessibility in the matching and scheduling links, thereby ensuring the feasibility of range and service continuity, reducing vehicle waiting and empty driving, relieving charging queuing, improving completion rate and site load balancing, helping enterprises to reduce costs and increase efficiency and green operation, and enabling large-scale deployment in multi-region scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0070] Figure 1 The flowchart of the modeling and optimization method for the electric crowd-sourced freight system based on the multi-agent model. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0072] The present embodiment provides a modeling and optimization method for an electric crowd-sourced freight system based on a multi-agent model, specifically including the following steps:

[0073] S1: Construct an electric crowd-sourced freight system based on a multi-agent model.

[0074] Specifically, the simulation framework is constructed based on multi-agent in this embodiment, which provides a basis for the construction of the electric crowd-sourced freight transportation system; the multi-agent modeling method provides a bottom-up analysis perspective for the electric crowd-sourced freight transportation system, which can reveal the macroscopic emergent characteristics of the system by simulating the autonomous behavior and interaction process of micro individuals. In this embodiment, a modular design is used to pre-construct the simulation framework, which includes four functional modules of environment configuration, simulation running, data recording and result analysis, and the system evolution is promoted through a synchronous updating mechanism.

[0075] A. Simulation environment configuration module: the basic running environment is constructed in a parameterized way, which can divide the region into multiple levels of spatial units; the transportation demand is based on historical data to construct a demand generation matrix, and the system generates freight agents according to the probability distribution of the matrix at each time step and randomly determines the start and end point positions; the vehicle fleet size is determined through pre-set simulation experiments, considering the vehicle exit rate and freight waiting time to find the optimal configuration; the spatial distribution of electric vehicles and charging stations adopts the known demand-oriented configuration method, which enables the freight demand-intensive areas to obtain higher resource allocation probability, ensuring the spatial matching of supply and demand of transport capacity; B. Simulation running module: a discrete event-driven mechanism is used to coordinate the behavior of each agent, that is, at each time step, the system triggers vehicle and freight matching, charging scheduling and empty vehicle scheduling events in parallel according to the current state information and pre-set decision rules, which updates the state attributes of related agents and promotes the evolution of the system; C. Data recording module: the state transition, event sequence and performance indicators of the agent are captured in real time, and the memory of the agent with completed life cycle is recycled; D. Result analysis module: the system performance is evaluated from the dimensions of freight transportation efficiency, vehicle utilization rate and charging resource allocation, which provides quantitative basis for strategy optimization.

[0076] The method for constructing the electric crowd-sourced freight transportation system comprises the following steps:

[0077] S11: constructing a simulation space-time network of the electric crowd-sourced freight transportation system based on a discretization method; specifically including: S110: considering the operation time dimension, evenly dividing the system operation period into a set of discrete time steps T and T = {0, 1..., t... |T|}; T = 0 represents the initial simulation time of system operation and each time step represents a fixed time interval;

[0078] S111: Considering the operational space dimension, a hierarchical grid method is adopted to divide the system operation area into several non-overlapping functional sub-regions B = {1,...i,j}, and each functional sub-region is defined as the basic spatial unit of the agent's activity area. Each functional sub-region is further gridded to obtain a spatial grid, and the center point of the spatial grid is used as the positioning reference of the agent. At the same time, Manhattan distance is used to measure the spatial distance between agents to simulate the path spatial location characteristics of the urban road network. The hierarchical grid method in this embodiment is a well-known existing technology and will not be described in detail here.

[0079] S112: Based on the set of discrete time steps and spatial distance, the simulation spatiotemporal network of the electric crowdsourced freight system is constructed. This embodiment can construct the simulation spatiotemporal network by considering the network characteristics of known urban roads and combining the set of discrete time steps and spatial distance through spatiotemporal network technology. The spatiotemporal network technology in this embodiment is a well-known existing technology and will not be described in detail here.

[0080] S12: Define an intelligent agent model based on the simulated spatiotemporal network, and the intelligent agent model includes at least a cargo agent, an electric vehicle agent, and a charging station agent;

[0081] Goods Agent: Define the set of goods agents as Q, where Q = {q1, q2, ..., q} N}, define the characterization of each cargo q i The attribute tuple of the feature is A q And A q =(S q ,R q ,L q D q ,T q W q );

[0082] Where S q ∈{0,1,2,3,4} represents the cargo status: 0 represents generated cargo; 1 represents cargo awaiting pickup; 2 represents cargo in transit; 3 represents delivered cargo; 4 represents cargo returned after timeout; R q Indicates the generation time of the required goods; L q Indicates the starting point of the goods; D q Indicates the final destination of the goods; T q Indicates the transport distance; W q Indicates the cumulative waiting time;

[0083] Electric Vehicle Agent: Define the set of electric vehicle agents as Z, where Z = {z1, z2, ..., zn}. N Define each electric vehicle z iA z and

[0084] wherein S z ∈{0,1,2,3,4,5,6,7} represents the operating state of the electric vehicle: 0 represents an idle state; 1 represents a state of going to pick up goods; 2 represents a state of transportation; 3 represents a state of empty vehicle waiting for scheduling; 4 represents a state of going to charge; 5 represents a state of charging; 6 represents a state of waiting for charging; 7 represents a state of exiting the system; L z represents the current position; D z represents the destination; represents the remaining power at time t; C z represents the battery capacity at time t; T z represents the maximum range; H z represents the distance to the destination; V z represents the driving speed;

[0085] Charging station agent: define the set of charging station agents as F and F = {f1, f2... f N}, define the attribute tuple owned by each charging station f i as A f and A f = (S f , L f , N f , M f , G f , V f );

[0086] wherein S f ∈{0,1} represents the available state of the charging station: 0 represents available; 1 represents unavailable; L f represents the geographic location; N f represents the total number of charging piles; M f represents the available number of charging piles; G f represents the waiting charging queue; V f represents the charging power;

[0087] S13: constructing interaction rules about the electric crowd-sourced freight transportation system based on the agent model, and the interaction rules at least include vehicle-goods matching rules, charging scheduling rules and empty vehicle scheduling rules;

[0088] The vehicle-goods matching rules:

[0089] The matching process first needs to verify the demand effectiveness of the vehicle-goods matching, that is, to confirm that the starting point and the end point of the goods are located in different system operation areas, and that there is an available idle electric vehicle based on the electric vehicle Agent; when the verification is passed, the matching scheme of dynamically adjusting the vehicle-goods matching based on real-time information is obtained according to the dynamic vehicle-goods matching optimization model, and the vehicle-goods state updating rule is triggered after confirming the matching scheme;

[0090] The vehicle-goods state updating rule: the electric vehicle is converted from S z =0 to S z =1, and the destination is updated to the location D z =L of the goods; q The distance of the electric vehicle to the destination is updated to H zi ; the state of the goods is converted from S q =0 to S q =1; when the electric vehicle arrives at the pickup point and completes the loading, the state of the electric vehicle is updated to S z =2; the destination is updated to the destination D z =D of the goods; q The distance is updated to T q ; the state of the goods is updated to S q =2; when the goods arrive at the destination, the state of the electric vehicle is restored to S z =0; the destination is reset to empty; the location is updated to L z =D q ; the electric quantity of the electric vehicle is updated according to the formula ; the state of the goods is updated to S q =3; if the waiting time of the goods exceeds the preset threshold, the state of the goods is updated to S z =4 and exits the system;

[0091] The charging scheduling rule:

[0092] Based on the hybrid charging optimization model of fusing emergency charging and preventive charging, a set of vehicles needing charging is screened at each time step, and the optimal matching pair of vehicle and charging station is determined for the set of vehicles;

[0093] And after confirming the optimal matching pair of vehicle and charging station, the charging scheduling updating rule is triggered;

[0094] The charging scheduling updating rule: the state of the vehicle needing charging is updated to S z =4; the destination is set to the location L z =D of the charging station; f The distance is updated to H zf ; the location of the vehicle is updated to L z =L fand record the current remaining power at the same time; determine whether there is an available charging pile in the charging station according to the charging station Agent; if the charging station has an available charging pile, i.e., M f = 0, the vehicle immediately starts charging, and the state of the vehicle is updated to S z = 5; otherwise, the vehicle joins the waiting queue G f and the state is updated to S z = 6; the charging duration of the vehicle in need of charging is calculated according to the formula ; when the charging is completed, the number of available charging piles M f of the charging station is increased by 1; the state of the vehicle is restored to S z = 0; and the power of the vehicle is updated to full power The position of the vehicle remains at the charging station L z = L f ;

[0095] The empty vehicle scheduling rule is:

[0096] The preset monitoring module monitors the continuous idle duration of the idle vehicle in the electric vehicle Agent, i.e., S z = 0, and determines whether it exceeds the preset duration threshold μ; if not, no operation is performed; if yes, the vehicle scheduling is triggered based on the real-time expectation calculation empty vehicle scheduling optimization model;

[0097] After confirming the scheduled vehicle, the empty vehicle scheduling update rule is triggered;

[0098] The state of the scheduled vehicle is updated to S z = 3; the destination is set to the target area center D z = i; after the scheduled vehicle arrives at the scheduling point, the state of the scheduled vehicle is restored to S z = 0; the power of the scheduled vehicle is updated according to the formula ; the position of the scheduled vehicle is updated to L z = i; it is determined whether the non-occupancy time ratio of the scheduled vehicle within a unit time exceeds the preset time ratio threshold; if yes, the system determines that the scheduled vehicle leaves, and the state of the scheduled vehicle is updated to S z = 7 and is removed from the preset active vehicle team queue; otherwise, no operation is performed. Specifically, in the embodiment, the non-occupancy time ratio of the vehicle within a unit time exceeds the threshold, i.e., the proportion of the time in which the vehicle is in a non-operating state such as non-cargo carrying, non-scheduling and non-charging to the total duration; the active vehicle team refers to a set of vehicles that are currently within the control range of the system platform, can participate in vehicle and cargo matching, scheduling and charging distribution.

[0099] S2: configuring a corresponding system operation optimization model for the electrified crowdsourcing freight system, and the system operation optimization model comprises: a dynamic vehicle-load matching optimization model for allowing secondary allocation; a hybrid charging optimization model for realizing fleet energy management by fusing emergency charging and preventive charging; and an empty vehicle scheduling optimization model based on real-time expectation calculation for quantitatively evaluating net benefits to optimize vehicle capacity allocation indicators;

[0100] Specifically, the embodiment proposes a dynamic vehicle-load matching optimization model allowing secondary allocation, the core of the strategy lies in expanding the range of matchable objects, that is, in addition to the traditional unallocated load set R U and the idle vehicle set Z I , the set of loads R A that have completed preliminary matching but the vehicles have not yet arrived at the pickup point and the corresponding set of vehicles on the way Z P are also included in the re-optimization range, thereby defining the extended sets R' and Z' and R' ∈ {q|R U ∪R A}, Z' ∈ {z|Z I ∪Z P}, so that the system can dynamically adjust the matching scheme based on real-time information, and the embodiment models the vehicle-load matching problem as a dynamically updated weighted bipartite graph maximum weight matching problem, and the objective function, i.e., the vehicle-load matching objective function, is:

[0101]

[0102] In the formula, R', Z' represent the defined extended sets; R U represents the unallocated load set; R A represents the set of loads that have completed preliminary matching but the vehicles have not yet arrived at the pickup point; Z I represents the idle vehicle set; Z P represents the set of vehicles on the way that have completed preliminary matching but the vehicles have not yet arrived at the pickup point; U qz represents the matching utility; x qz represents the decision variable of whether the load q is matched with the vehicle z and x qz ∈{0,1}; W q represents the waiting time normalization value of the load q; H qz represents the vehicle-load distance; λ q represents the vehicle-load distance weight coefficient; λ w represents the load waiting time weight coefficient for balancing timeliness and economy; λ p represents the penalty weight coefficient; δ represents the secondary allocation penalty coefficient; q z represents the 0-1 decision variable of whether the vehicle z has completed matching (if z ∈ Z P is going to the location of the load, then qz = 1 ; otherwise, q z = 0); y qz represents a 0-1 decision variable that determines whether the electric vehicle z has completed matching with the goods q at the last time step (e.g., y qz = 1 indicates that the electric vehicle z has been assigned to the goods q before; y qz = 0 indicates that the electric vehicle z has not served the goods q originally, i.e., for reassignment);

[0103] The first constraint condition includes

[0104] An assignment uniqueness constraint (3) is ensured that the unassigned goods must be assigned:

[0105]

[0106] In the formula: a q represents an identification variable of whether the goods have been assigned; a q = 1 indicates that the goods have completed a matching once; a q = 0 indicates that the goods have not completed a matching once;

[0107] A reassignment times constraint (4) is limited that the goods are re-assigned at most once to avoid frequent adjustment:

[0108]

[0109] In the formula: b q represents a decision variable of whether the identified goods have been re-assigned; if the electric vehicle z e Z P is going to the location of the goods q e R A , and the goods q e R A have been re-assigned once, then b q = 1; otherwise, b q = 0;

[0110] An energy feasibility constraint (5) is ensured that the remaining energy of the vehicle is sufficient to reach the nearest charging station after completing the transportation task:

[0111]

[0112] In the formula: represents the current remaining mileage of the vehicle; H qz represents the distance from the vehicle to the goods; H ij represents the goods transportation mileage; H zf represents the distance to the nearest charging station.

[0113] In this embodiment, the above optimization problem is solved by using the Kuhn-Munkres algorithm, which can find an augmented path by constructing an equivalent subgraph and iterating in O(n3 )time complexity, the Kuhn-Munkres algorithm is executed at each decision-making moment, and the specific steps include: a. initialization stage, constructing a vehicle-goods bipartite graph, obtaining the utility value of all feasible matches, and forming a weight matrix; b. top label setting, initializing the row top label as the maximum value of each row and the column top label as zero; c. equivalence subgraph construction, constructing an equivalence subgraph based on the current top label; d. augmentation path search, using the Hungarian algorithm to find an augmented path to expand the matching; e. top label adjustment, if a perfect match is not found, obtain the relaxation variable and update the top label; f. iterative optimization, repeat steps c-e until the maximum weight matching is found. Through the above dynamic matching strategy, the system can respond to state changes in real time, continuously optimize resource allocation under the premise of ensuring service continuity, and effectively improve overall operating efficiency.

[0114] In this embodiment, a hybrid charging optimization model integrating emergency charging and preventive charging is proposed, that is, the foresighted management of vehicle fleet energy can be realized through a double-layer decision mechanism: the first layer is emergency charging triggering: for vehicles with power lower than the emergency threshold θ in the idle vehicle set Z idle , immediately join the emergency charging queue to ensure timely energy replenishment of low-power vehicles; the second layer is preventive charging scheduling: foresighted charging decision is made based on the overall energy state of the vehicle fleet, and the average power of the idle vehicle fleet is calculated, expressed as:

[0115]

[0116] wherein, Z norm represents the set of idle vehicles with power higher than the emergency threshold; the preventive charging scale is determined according to the hierarchical threshold function N, and its expression is:

[0117]

[0118] By selecting the N vehicles with the lowest power in Z norm , that is, the vehicle set determined according to the hierarchical threshold function N, whose power is lower than the preset emergency threshold θ and joins the preventive charging queue

[0119]

[0120] wherein: n1, n2 represent the vehicle set determined according to the hierarchical threshold function N and n ∈ n1, n2;

[0121] Further, the final set of vehicles to be charged is:

[0122]

[0123] In this embodiment, the goal of allocating charging stations is to minimize the total charging activity time. An integer programming model, i.e., a hybrid charging objective function, is constructed, and its expression is:

[0124]

[0125] In the formula: x qz ∈{0,1} are decision variables, representing whether vehicle z is assigned to charging station f; and the charging activity time T. zf It consists of three parts:

[0126] T zf =t zf +t zw +t zc (11)

[0127] In the formula: t zf Indicates travel time; t zw Indicates queuing waiting time; t zc Indicates the actual charging time;

[0128] By establishing allocation uniqueness constraint (12) and energy accessibility constraint (13), i.e., the second constraint condition, to ensure that the vehicle has sufficient remaining battery power to reach the designated charging station, the expression is as follows:

[0129]

[0130] In this embodiment, considering the NP-hard nature of the problem, the Particle Swarm Optimization (PSO) algorithm is used to solve the objective function of the hybrid charging strategy. This algorithm simulates bird flock foraging behavior and iteratively searches for an approximate optimal solution in the solution space. The algorithm's execution steps include: a. Population initialization: randomly generating N particles, each particle's position vector representing a vehicle-charging station allocation scheme, and its velocity vector controlling the search direction; b. Fitness evaluation: calculating the total charging time for each particle's corresponding allocation scheme as its fitness value; c. Optimal solution update: recording each particle's historical optimal position and the population's global optimal position; d. Velocity and position update: updating particle velocity and position based on individual experience and collective intelligence; e. Boundary handling: correcting out-of-bounds particles to ensure the feasibility of the allocation scheme; f. Iterative optimization: repeating steps b until the maximum number of iterations or the convergence condition is reached. Through this proactive hybrid charging strategy, the system can smooth the temporal distribution of charging demand, balance the load on charging stations, and maintain high availability of the entire fleet while ensuring timely charging for vehicles with low battery levels.

[0131] This embodiment proposes an empty vehicle scheduling optimization model based on real-time expectation calculation, which can guide the allocation of transportation capacity resources by quantitatively evaluating the net benefits of empty vehicle scheduling. The core of this strategy is to construct an expectation function that comprehensively considers both benefits and costs.

[0132] E za = p t · t at -p s · t as -p r · T za (14)

[0133] wherein: E za represents the expected net benefit of dispatching vehicle z to operating area a; p t · t at represents expected service revenue; p t represents the unit distance service price; t at represents the average service distance of operating area a; p s · t as represents the waiting cost; p s represents the unit time opportunity cost; t as represents the average matching waiting time of area a; p r · T za represents the dispatch cost; p r represents the unit time driving cost; T za represents the vehicle empty driving time;

[0134] To avoid the flock effect caused by deterministic dispatching, a probability selection mechanism based on the softmax function is adopted in this embodiment: that is, the expected value E za is converted into selection intensity C za by exponential transformation according to formula (15), and then normalized by formula (16) to obtain the selection probability p za of each area, and the expression is as follows:

[0135]

[0136]

[0137] According to the dynamic dispatching strategy based on real-time expected calculation, the adaptive rule algorithm can be used to solve the empty vehicle dispatching problem in this embodiment: that is, by performing the following steps at each decision-making time: a. reachability analysis, determine the reachable area set according to the current position L z and the remaining power of the vehicle, and exclude the current area to avoid invalid dispatching; b. expected value calculation, for each reachable area a, obtain the average service distance t at , the average matching time t as , and calculate the empty driving time T za, the expected net income is calculated according to formula (14); c. Probability distribution construction, the expected value vector is converted into a probability distribution by applying a softmax transformation, and a higher selection probability is obtained in a high expected value area; d. Random decision generation, weighted random sampling is performed based on the probability distribution to determine the target scheduling area; e. Abnormal processing, if there is no reachable area (insufficient power), the vehicle state is marked as needing to be charged, triggering the charging scheduling process; f. Decision execution, the selected target area is output, guiding the vehicle to carry out empty car scheduling. Through the above-mentioned dynamic scheduling strategy based on real-time expectations, the system can intelligently configure the transport capacity resources according to the real-time supply and demand state and economic indicators, improve service coverage while controlling scheduling costs, and achieve efficient use of transport capacity resources.

[0138] S3: According to the system operation optimization mechanism, the operation optimization of the electric crowd-sourcing freight system based on the multi-agent model is realized. In this embodiment, through the multi-Agent integrated simulation framework, the problem that the existing method excessively relies on traditional mathematical programming and is difficult to depict the autonomous decision and complex interaction of subjects such as goods, electric vehicles and charging stations is solved. At the same time, based on the Agent simulation of the fuel scene, the electric special constraints (such as SOC evolution, partial charging, site capacity / queuing, station accessibility after task completion, etc.) are not considered, which leads to distorted evaluation, difficult strategy landing, and difficult comparison between different strategies under the unified standard. The method described in this embodiment forms a unified multi-Agent modeling and simulation evaluation basis under the electric background, and describes the operation mechanism of the vehicle in the same simulation space-time network.

[0139] In this embodiment, through the adjustable matching-supplementing-scheduling linkage mechanism, i.e., the system operation optimization mechanism: a) The vehicle and cargo matching lacks controlled adjustment (such as limited redistribution in the transit stage) according to system information updates, which is difficult to break through the limitation of "one-time allocation" and is easy to cause service continuity damage and empty driving rise; b) The charging scheduling is mostly passive response, which is difficult to anticipate balanced site load and improve charging resource utilization efficiency, and is easy to form queuing congestion in peak period and pile idle in trough period; c) The empty car scheduling is difficult to consider the space-time supply and demand imbalance and power accessibility at the same time, and is easy to cause regional clustering and secondary empty driving. In this embodiment, by constructing a strategy system that can be adjusted in a rolling time domain according to state updates, the matching can be adjusted, the supplementing can be planned, and the feasibility and efficiency can be maintained under the power and capacity constraints.

[0140] System layer cooperative optimization: existing methods often optimize vehicle-load matching, charging scheduling and empty vehicle scheduling separately, ignoring the internal relationship and synergistic effect among the three: that is, matching does not provide energy consumption / arrival prediction to the energy supplement layer, the energy supplement result is not timely fed back to the vehicle availability and spatial distribution of the matching layer, and scheduling does not absorb the matching history and charging plan for pre-configuration. Thus, energy infeasibility, station capacity conflict and regional supply and demand imbalance may easily occur. The method described in the embodiment establishes information closed loop and constraint consistency (energy accessibility, matching uniqueness and station capacity, etc.) among the three strategies, and can realize cooperative optimization in a unified rolling integrated framework, ensuring the linkage improvement and stable feasibility of system level (waiting, completion rate, empty driving, queuing and load balancing).

[0141] In summary, the method described in the embodiment realizes information closed loop and constraint consistency of "vehicle-load matching-charging scheduling-empty vehicle scheduling" in a unified discrete space-time and multi-agent framework, avoiding infeasibility and suboptimality caused by separate optimization of electric crowd-sourced freight transportation system; through the dynamic vehicle-load matching strategy for allowing secondary allocation, the risk of freight waiting and overage is reduced, and service coverage and completion rate are improved; through the hybrid charging strategy for realizing energy management of vehicle fleet by integrating emergency charging and preventive charging, the total energy supplement time of "driving + queuing + charging" can be minimized, peak congestion can be alleviated, and station load balancing can be improved; based on real-time expectation calculation, the empty vehicle scheduling strategy can inhibit operation area clustering, reduce secondary empty driving and optimize space allocation of transport capacity by considering expected net income and using probabilistic selection for empty vehicle scheduling; the invention can endogenously evolve SOC and achieve station accessibility in matching and scheduling links, ensuring the feasibility of driving range and service continuity, thereby reducing vehicle waiting and empty driving, alleviating charging queuing, improving completion rate and station load balancing, helping enterprises to reduce costs and increase efficiency and green operation, and having minute-level rolling solution and modular integration capability, facilitating online deployment and scale expansion in multiple regions / multiple centers; reducing comprehensive operation cost and improving vehicle turnover efficiency at the system level, and supporting sustainable development of green logistics and urban distribution with electrification, which has significant application prospects.

[0142] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for modeling and optimization of an electrified crowdshipping freight system based on multi-agent models, characterized in that, Specifically, the following steps are included: S1: Construct an electric crowdsourced freight system based on a multi-agent model; The method for constructing the electric crowdsourced freight system specifically includes the following steps: S11: Constructing a simulation spatiotemporal network for an electric crowdsourced freight system based on a discretization method; S12: Define the agent model based on the simulated spatiotemporal network; Furthermore, the intelligent agent model includes at least a cargo agent, an electric vehicle agent, and a charging station agent; S13: Constructing interaction rules for an electric crowdsourced freight system based on an agent model; Furthermore, the interaction rules should include at least vehicle-cargo matching rules, charging scheduling rules, and empty vehicle scheduling rules; S2: Configure the corresponding system operation optimization model for the electric crowdsourced freight system; The system operation optimization model includes: A dynamic vehicle-cargo matching optimization model that allows for secondary allocation; a hybrid charging optimization model that integrates emergency charging and preventative charging for fleet energy management; and an empty vehicle scheduling optimization model based on real-time expectation calculation for quantitatively evaluating net benefits to optimize vehicle capacity space configuration indicators. A dynamic vehicle-cargo matching optimization model that allows for secondary allocation, including the vehicle-cargo matching objective function and the first constraint condition: The expression for the vehicle-cargo matching objective function is as follows: wherein: denotes the defined extended set; denotes the unassigned goods set; denotes the set of goods that have completed the preliminary matching but the vehicle has not arrived at the pickup point; denotes the set of idle vehicles; denotes the set of on-the-way vehicles that have completed the preliminary matching but the vehicle has not arrived at the pickup point; denotes the matching utility; denotes the goods denotes the vehicles denotes the decision variable whether a vehicle ; denotes the waiting time normalized value of a good ; denotes the vehicle-goods distance; denotes the vehicle-goods distance weight coefficient; denotes the good waiting time weight coefficient for balancing the timeliness and economy; denotes the penalty weight coefficient; denotes the quadratic assignment penalty coefficient; denotes the decision variable whether a vehicle has completed the matching; denotes the decision variable whether an electric vehicle has completed the matching with a good at the last time step; The first constraint includes The allocation uniqueness constraint ensures that unallocated goods must be allocated: In the formulae: an identification variable indicating whether the goods have been allocated or not; Constraint limiting the number of reassignments a shipment to a maximum of one: In the formulae: denotes a decision variable indicating whether the identified goods have been redistributed or not. Energy feasibility constraint to ensure that the vehicle has sufficient remaining battery power to reach the nearest charging station after completing its transportation task: wherein: represents the current remaining range of the vehicle; represents the distance to the goods of the vehicle; represents the goods transport range; represents the distance to the nearest charging station; represents the set of charging station agents; represents a charging station individual; S3: Based on the system operation optimization model, optimize the operation of the electric crowdsourced freight system based on the multi-agent model.

2. The method of claim 1, wherein, The method for constructing the simulated spatiotemporal network in S11 is as follows: S110: considering the operation time dimension, evenly dividing the system operation period into a set of discrete time steps and ; denotes the initial moment of simulation of system operation; S111: considering the operation space dimension, the hierarchical grid method is adopted to divide the system operation area into a plurality of non-overlapping functional sub-areas and define each functional sub-area as a basic space unit of the agent activity area, and further grid division of each functional sub-area obtains a space grid, and the center point position of the space grid is taken as the positioning reference of the agent, and the Manhattan distance is used to measure the spatial distance between the agents to simulate the path space position characteristics of the urban road network; S112: Based on the discrete time step set and spatial distance, the simulation spatiotemporal network of the electric crowdsourced freight system is constructed.

3. The method of claim 2, wherein, The agent models defined in S12 include: Cargo Agent: define the set of cargo agents as and define the attribute tuple for characterizing each cargo feature as and ; wherein represents the state of the goods: 0 represents generated goods; 1 represents goods to be picked up; 2 represents goods in transit; 3 represents goods that have been delivered; 4 represents goods that have been exited due to overtime; represents the generation time of the goods to be recorded; represents the starting point position of the goods; represents the end point position of the goods; represents the distance of the transportation; represents the cumulative waiting time; Electric Vehicle Agent: define the set of electric vehicle agents as and , define the attribute tuple of each electric vehicle as and ; wherein represents the operating state of the electric vehicle: 0 represents an idle state; 1 represents a state of going to pick up goods; 2 represents a state of transportation; 3 represents a state of empty vehicle waiting for dispatch; 4 represents a state of going to charge; 5 represents a state of charging; 6 represents a state of waiting for charging; 7 represents a state of exiting the system; represents the current position; represents the destination; represents the remaining power at the moment; represents the battery capacity at the moment; represents the maximum range; represents the distance to the destination; represents the driving speed; Charging Station Agent: define the set of charging station agents as and , define the attribute tuple owned by each charging station as and ; wherein represents the available state of the charging station: 0 represents available; 1 represents unavailable; represents the geographic location; represents the total number of charging piles; represents the available number of charging piles; represents the waiting charging queue; represents the charging power.

4. The method of claim 3, wherein, S13 constructs interaction rules for the electric crowdsourced freight system based on an agent model, including... Vehicle-cargo matching rules: The vehicle-cargo matching is validated for demand validity, that is, the origin and destination of the goods are located in different system operation areas, and the availability of idle electric vehicles is confirmed based on the electric vehicle agent. Once the validation is successful, a matching scheme is obtained based on real-time information and dynamically adjusted according to the dynamic vehicle-cargo matching optimization model, and the vehicle-cargo status update rule is triggered after the matching scheme is confirmed. The vehicle cargo status updating rule: the electric vehicle is converted from to , and the destination is updated to the cargo location ; The electric vehicle distance to the destination is updated to ; the cargo status is converted from to ; when the electric vehicle reaches the pickup point and completes loading, the electric vehicle status is updated to ; the destination is updated to the cargo destination ; the distance is updated to ; and the cargo status is updated to ; When the goods arrive at the destination, the electric vehicle state is restored to ; the destination is reset to null; Location update is ; The electric power of the electric vehicle is according to the formula The update; the cargo status is updated to ; If the cargo waiting time exceeds a preset threshold, the cargo state is updated as and exits the system; Charging scheduling rules: Based on a hybrid charging optimization model that integrates emergency charging and preventive charging, the set of vehicles that need charging is selected at each time step, and charging stations are assigned to them to determine the optimal vehicle-charging station matching pair. And after confirming the optimal vehicle and charging station match, the charging scheduling update rules are triggered; The charging scheduling update rule: update the state of the screened vehicle needing charging to ; the destination is set to the charging station location ; the distance is updated to ; the position of the vehicle after arriving at the charging station is updated to , and the current remaining power is recorded at the same time; according to the charging station Agent, it is confirmed whether there is an available charging pile in the charging station; If the charging station has available charging piles, then... The vehicle immediately began charging and updated its status. Otherwise, add it to the waiting queue. And update the status to ; The charging duration of the vehicle in need of charging is calculated according to the formula The number of available charging piles of the charging station when the charging is completed is increased by 1; the vehicle state is restored to ; and the vehicle power is updated to full power ; Vehicle position maintenance at charging station ; Empty train dispatching rules: The system monitors idle vehicles in the electric vehicle agent through a pre-installed monitoring module. The continuous idle time is determined, and it is judged whether it exceeds a preset time threshold. If not, no action is taken; if so, the empty vehicle scheduling optimization model based on real-time expectation calculation is triggered to schedule vehicles. And after confirming the dispatched vehicle, the empty vehicle dispatch update rule is triggered; Update the status of the scheduled vehicle to ; destination set to the target zone center ; After the scheduled vehicle arrives at the scheduling point, the state of the scheduled vehicle is restored to ; the power of the scheduled vehicle is updated according to the formula ; the position of the scheduled vehicle is updated to ; it is judged whether the non-occupancy time ratio of the scheduled vehicle within a unit time exceeds a preset time ratio threshold; if yes, the system determines that the scheduled vehicle leaves, updates its state to and removes it from the preset active vehicle team queue; otherwise, no processing is performed.

5. The multi-agent model based electrically motorized crowd sourced freight system modeling and optimization method of claim 4, wherein, The hybrid charging optimization model in S2, which integrates emergency charging and preventative charging for fleet energy management, includes the hybrid charging objective function and the second constraint condition. The expression for the hybrid charging objective function is as follows: In the formula: represents the charging station allocation planning model aiming at minimizing the total charging activity time; represents the final set of vehicles to be charged; represents the emergency charging queue; represents the set of idle vehicles whose electric quantity is higher than the preset emergency threshold value; represents the set of vehicles determined according to the hierarchical threshold function whose electric quantity is lower than the preset emergency threshold value and joined in the prevention charging queue; represents the average electric quantity of the idle vehicle team; represents the set of vehicles determined according to the hierarchical threshold function and ; represents the decision variable of whether the vehicle is allocated to the charging station and ; represents the charging activity time; represents the driving time; represents the queuing waiting time; represents the actual charging time; The second constraint is to ensure that the remaining power of the vehicle is sufficient to reach the designated charging station, and the distribution uniqueness constraint and the energy accessibility constraint: 。 6. The multi-agent model based electrically motorized crowd-sourced freight system modeling and optimization method of claim 5, wherein, The empty car scheduling optimization model based on real-time expected calculation for quantitatively evaluating the net benefit optimization of space allocation of transport capacity in S2 is expressed as In the formula, represents the selection probability corresponding to the operation area guiding the vehicle to carry out empty dispatching; represents the expected value after conversion; represents the expected net income of the vehicle dispatching to the operation area ; represents the expected service income; represents the service price per unit distance; represents the average service distance of the operation area ; represents the waiting cost; represents the opportunity cost per unit time; represents the average matching waiting time of the area ; represents the dispatching cost; represents the driving cost per unit time; represents the vehicle empty driving time.​​​

Images