Electric vehicle travel scene ordered charging method under dynamic electricity price guidance

Abstract

The invention discloses an electric vehicle travel scene ordered charging method under the guidance of dynamic electricity price, and the method comprises the steps: carrying out the charging load space-time distribution prediction through an EV charging load space-time distribution prediction model, and obtaining the load space-time distribution characteristics of each charging station; generating an optimized electricity price strategy through a master-slave game model, and making a dynamic electricity price scheme by an electric vehicle aggregator in combination with a station charging load condition and a charging responsivity function; and simulating user travel behaviors according to the road network traffic flow data, and making a user charging plan based on the corrected electricity price information. According to the method, an ordered charging regulation and control strategy based on dynamic electricity price response is provided, a dynamic electricity price model based on charging load real-time change is constructed, meanwhile, a double-layer game model of a load aggregator and a user is established in combination with user travel space-time distribution characteristics, and an ordered charging optimization strategy giving consideration to multi-party benefits is formed; and finally, dual goals of power grid operation optimization and user economic benefits are realized.

Description

Technical Field

[0001] The present invention belongs to the field of electric power, relates to an electric vehicle charging strategy, and particularly relates to an orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance. Background Art

[0002] With the sharp increase in the penetration rate of electric vehicles, the irregular charging patterns of a large number of electric vehicles may cause significant load fluctuations, weakening the safety margin of power grid operation. The charging behaviors of most electric vehicle users are concentrated during lunch breaks and when they return home from work in the evening. Such large-scale charging operations will lead to the superposition effect of charging demand and peak power demand, resulting in the phenomenon of "peak on peak", which will further increase the peak-valley difference and network loss of the urban distribution network load. In the spatial dimension, it will also cause problems such as user charging congestion and difficult grid dispatching, which is not conducive to the balanced coordination of power supply and demand and hinders the efficient and stable operation of the power system. Despite many challenges, the large-scale popularization of electric vehicles still brings new opportunities for the development of the power grid. The fast response ability of its energy storage battery endows the device with the attribute of flexible load, making the electric vehicle have the technical characteristics of a mobile energy storage unit. In actual operation, these scattered energy storage devices can not only charge and store energy during low electricity consumption periods as conventional loads, but also supply power in reverse as distributed power sources during peak load periods of the power grid. This two-way interaction characteristic provides flexible regulation resources for the power system, but the development of its regulation potential depends on scientific management strategies. Therefore, it is necessary to build an orderly charge and discharge management system suitable for multiple scenarios, optimize the charge and discharge behaviors of electric vehicles through dynamic regulation strategies, and achieve two-way coordination between the power grid and users. This system not only needs to meet the travel charging needs of users, but also fully explore the regulation potential of electric vehicles as mobile energy storage, and improve energy utilization efficiency while ensuring the safe and stable operation of the power grid.

[0003] Current research mainly focuses on orderly charge and discharge strategies under the framework of fixed time-of-use electricity prices. Existing electricity price mechanisms mostly adopt a pre-determined time period division mode, which is difficult to adapt to the dynamic changes of power grid loads. Existing management schemes focus more on load period transfer and lack consideration of the interest coordination mechanism among power grid operators, load aggregators, and end users. Summary of the Invention

[0004] Object of the Invention: In order to overcome the deficiencies in the prior art, an orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance is provided. An orderly charging regulation strategy based on dynamic electricity price response is proposed, and a dynamic electricity price model based on the real-time change of charging load is constructed. At the same time, combined with the spatio-temporal distribution characteristics of user travel, a two-layer game model between load aggregators and users is established to form an orderly charging optimization strategy that takes into account the interests of multiple parties, and finally achieve the dual goals of optimizing power grid operation and user economic benefits.

[0005] Technical solution: To achieve the above object, the present invention provides an orderly charging method for electric vehicle travel scenarios under dynamic electricity prices, including the following steps:

[0006] S1: Predict the spatio-temporal distribution of charging load through the EV charging load spatio-temporal distribution prediction model to obtain the spatio-temporal distribution characteristics of the load at each charging station;

[0007] S2: Generate an optimized electricity price strategy through the master-slave game model, and the electric vehicle aggregator formulates a dynamic electricity price plan in combination with the charging load situation of the station and the charging response function;

[0008] S3: Simulate the travel behavior of users based on the road network traffic flow data, and formulate a user charging plan based on the corrected electricity price information.

[0009] Further, the EV charging load spatio-temporal distribution prediction model in step S1 includes a "vehicle-road-network" interaction model and an electric vehicle charging load model, and the "vehicle-road-network" interaction model includes a road topology structure model, a road resistance model, a distribution network model, and a electricity price response model.

[0010] Further, the establishment of the "vehicle-road-network" interaction model in step S1 specifically includes:

[0011] A1: Establish a road topology structure model:

[0012] The mathematical model of the traffic road network is represented by equation (1):

[0013]

[0014] In the formula, G is the sum of the road network system; V is the summary of the intersection points in G; E is the road cluster in G; K is the time period; W represents the cost of passing, and indicators such as node waiting time and passing time can be used to quantitatively model and analyze it;

[0015] The incidence matrix Z presents the traffic road resistance value, as shown in equation (2):

[0016]

[0017] In the formula, inf means that node n i and n j are not passable between them;

[0018] The specific expression form of the incidence matrix Z is in the form of equation (3):

[0019]

[0020] To build an electric vehicle spatio-temporal trajectory model, first, the travel time parameters (initial time, return time) and spatial parameters (origin and destination) of the vehicle are determined through OD matrix analysis. Secondly, the Dijkstra algorithm based on real-time traffic information is used to carry out the driving trajectory planning, and its objective functions are shown in Equations (4) and (5):

[0021]

[0022] In the formula, if n ij belongs to the passing trajectory d i (i), x ij is assigned 1, otherwise 0;

[0023] A2: Establish a road resistance model:

[0024] Since the charging load of the electric vehicles studied in this invention is mainly distributed in urban areas, the vehicle passing efficiency is not only restricted by the road congestion condition, but also affected by the regulation of road nodes such as traffic lights. When the vehicle passes through these control nodes, in addition to coping with the passing resistance generated by natural road conditions such as curves and bends, there will also be additional waiting time due to traffic lights. Therefore, in the process of calculating the total travel time of the whole journey, it is necessary to comprehensively consider the influence of road congestion (section impedance) and intersection waiting time (node impedance). Therefore, the mathematical model of road resistance is expressed as:

[0025]

[0026] In the formula, Cv i (t) is the road node impedance model; Rv ij (t) is the section impedance model;

[0027] A3: Establish a distribution network model

[0028] The EV charging load of the distribution network node n at time t is shown in the following formula

[0029]

[0030] In the formula, N represents the number of electric vehicles charging through node n during time period t; P i,t represents the charging power of the i-th electric vehicle during time period t;

[0031] A4: Establish a price responsiveness model

[0032] Electric vehicle users usually tend to choose the nearest charging facility for energy replenishment. This behavior pattern may cause overloading of the charging load in local areas, threatening the operational stability of the distribution network. In addition, during peak charging hours, the roads around charging stations in traffic-dense areas may experience additional traffic pressure due to vehicle congestion, exacerbating local congestion. Therefore, it is necessary to design a regulation strategy that includes spatial electricity price differences, guiding users' charging choices through price signals to achieve the spatio-temporal balanced distribution of the charging load. Based on this, this paper constructs an electricity price responsiveness evaluation model and establishes a dynamic regulation mechanism for electric vehicle charging behavior and regional electricity prices.

[0033] For electric vehicle users to participate in the electricity price regulation mechanism, specific conditions need to be met, that is, a response behavior will occur only when the electricity price breaks through a specific critical value, and there is a saturation critical value. When the electricity price difference exceeds this critical value, even if the electricity price difference continues to increase, the number of electric vehicle users participating in the regulation will tend to be stable and will not show continuous growth. Then the mathematical model of the EV user responsiveness index is:

[0034]

[0035] In the formula, Δc m,n (t): The charging electricity price difference between the m-th charging station and the n-th charging station at time t; ω c (t): The EV responsiveness at time t; ω c,max : The upper limit value of responsiveness; J m,n and h m,n correspond to the response trigger critical value and saturation threshold of electric vehicle users for the m-th charging station and the n-th charging station respectively.

[0036] Furthermore, in step A2, according to the saturation degree S of the vehicle flow density on urban roads, the traffic smoothness is classified into four levels: unobstructed (0 < S ≤ 0.6), slow driving (0.6 < S ≤ 0.8), dense traffic flow (0.8 < S ≤ 1.0), and traffic stagnation (1.0 < S ≤ 2.0);

[0037] Link impedance model: The link impedance models under different saturation conditions are shown in Equation (9):

[0038]

[0039] In the formula, the saturation degree is defined as Q represents the road vehicle flow condition, C represents the road carrying capacity; t0 represents the travel time of vehicles when there is no vehicle flow; α, β represent the influence factors related to impedance;

[0040] Node impedance model:

[0041]

[0042] Where c represents the signal cycle time of the urban traffic light; λ represents the ratio of the green traffic light to the red traffic light; q represents the vehicle arrival rate of the road section;

[0043] By combining equations (9) and (10) with saturation as an intermediate variable, we can obtain the road resistance calculation model:

[0044]

[0045] Furthermore, the establishment of the electric vehicle charging load model in step S1 specifically includes:

[0046] B1: EV Charging and Driving Characteristics Modeling

[0047] The energy storage capacity of electric vehicles shows a significant Weibull distribution, as shown in formula (12):

[0048]

[0049] Where, α t and β t represents the eigenvalue of the Weibull distribution, represents the energy storage capacity of vehicle i;

[0050] B2: EV power consumption model considering ambient temperature and vehicle speed

[0051] Energy consumption parameters exhibit complex fluctuations, with their dynamic changes influenced by factors such as driving behavior and climatic conditions. Most current prediction models assume a fixed energy consumption per mile, a simplification that can lead to biased predictions. However, in real-world scenarios, energy losses from frequent vehicle starts and stops, as well as increased energy consumption from air conditioning systems, can cause actual energy consumption data to deviate from the prediction model.

[0052] In order to obtain the power consumption relationship between ambient temperature and vehicle speed changes, a speed and flow correlation model based on real-time road network traffic flow was established, as shown in formula (13):

[0053]

[0054] Where, v0: vehicle speed under zero traffic density; C ij : upper limit of traffic capacity; χ: adjustment factor; q ij (t): dynamic traffic flow data; a, b, γ are road condition parameters;

[0055] By analyzing and deriving equations (13) and (15), two important correlation relationships can be further clarified. On the one hand, the specific relationship between the vehicle's driving speed and driving energy consumption can be clearly obtained, which reflects the influence law of speed change on driving energy consumption; on the other hand, the mathematical expression between the temperature and temperature control energy consumption can also be determined, which reflects the influence result of temperature change on vehicle energy consumption.

[0056]

[0057] In the formula, respectively represent the start values of heating and cooling of the vehicle temperature control system; W L 、W R : respectively represent the heating and cooling power consumption of the temperature control system; E: energy consumption under normal driving conditions; E Tp : the power consumption per kilometer at the temperature T p ;

[0058] B3: M / M / k queuing theory of charging stations

[0059] For the optimization requirement of charging station resource allocation, the present invention adopts a multi-channel service model (M / M / k) as a research tool. Through the assumptions of random arrival flow and exponential service duration, the key indicator of average delay duration is focused on, and the dynamic service capabilities of each charging station are statistically analyzed based on this model. The objective functions are shown in equations (17)-(19):

[0060]

[0061] In the formula, ε T : the number of charging loads in a certain period; g: the deployment scale of charging terminals; ρ T : the utilization efficiency of charging terminals; μ T : the scale of customers served by charging services per unit time; χ T : the limit carrying capacity of the charging station per unit time.

[0062] Furthermore, different types of electric vehicles have different charging requirements in step B1. Therefore, the present invention sets three different charging departure requirements:

[0063] Private cars: When the vehicle journey ends and the remaining available power is lower than the endurance threshold, the charging requirement will be activated, which is represented by equation (20):

[0064] C t (i) ≤ L td (i)E Tp (20)

[0065] In the formula, C t (i) is the remaining power of electric vehicle i, Ltd (i) represents the driving path length between the location of vehicle i and the upcoming destination, E Tp : power consumption per kilometer;

[0066] Taxi: Since electric taxis have a relatively high daily charging frequency and long operating mileage, when the remaining battery power is lower than the driver's psychologically preset power threshold, a charging demand will be triggered, which is represented by Equation (21):

[0067] C t (i) ≤ τC l (i) (21)

[0068] In the formula, C l (i) represents the battery capacity of the electric taxi, and the value range of τ is 0.15 - 0.35;

[0069] Pool car: According to its characteristics of multiple trips, if the condition is met, slow charging at the destination is adopted; if the charging trigger condition is not met, that is, the existing battery power cannot reach the destination during driving, a charging station needs to be found nearby for charging;

[0070] The charging duration and the end time of charging are represented by Equations (22) and (23):

[0071]

[0072] T f (i) = T a (i) + T w (i) + T c (i) (23)

[0073] In the formula, η c and P c respectively represent the EV charging efficiency (taking 0.8 - 0.9) and the charging power; T a (i): the arrival time of the electric vehicle; T c (i): the charging duration; T f (i) represents the end time of charging.

[0074] Furthermore, the specific process of predicting the spatio - temporal distribution of the charging load in step S1 includes:

[0075] C1: OD (origin - destination) analysis

[0076] To study the travel distribution law of electric vehicles in the traffic road network, for different types of electric vehicles, the OD matrix analysis method is used to construct a travel characteristic model. As the core tool for describing the traffic flow distribution or travel demand characteristics, the OD matrix can intuitively reflect the total travel demand between each origin and destination in a specific area. The travel volume of electric vehicles on different road sections at each time period can be obtained based on the traffic flow data measured by the traffic management department, and then the OD start and end matrix can be deduced by the dynamic change of the traffic volume at each time period, and the Monte Carlo sampling method is used to simulate the key parameters such as the departure and return times of electric vehicles and the location of the origin and destination of the trip. The mathematical model of the OD travel probability matrix is shown in Equation (24):

[0077]

[0078] In the formula, is the travel probability from road network node i to node j; is the number of EVs departing from node i to node j during the time period from T to T+1; is the number of all EVs departing from traffic network node i;

[0079] C2: Optimal path planning for electric vehicles based on real-time Dijkstra algorithm

[0080] Relying on the previous OD probability matrix, the origin and destination of each vehicle can be accurately located, and the full travel process simulation can be completed by combining the time parameters in the vehicle model. Since there are often multiple feasible path choices between the origin and destination, there is a direct connection between the choice of the vehicle driving path and the power consumption of the drive system and the travel efficiency. Therefore, it is necessary to establish a dynamic path planning system that integrates real-time traffic data, and online generate the optimal driving route considering the battery state and road conditions through the rolling horizon optimization algorithm, so as to ensure the accuracy of energy consumption assessment and spatio-temporal prediction.

[0081] In view of the characteristics of real-time traffic data, the present invention provides a route arrangement for the travel of electric vehicles based on the Dijkstra algorithm. The core logic of this algorithm is to iteratively expand the currently known set of shortest paths and gradually approach the desired node. The specific process is as follows:

[0082] a) Initialization: First, initialize the distance from the starting point to zero, because the distance from a node to itself is always equal to 0; for all other nodes, set their corresponding initial shortest distances to infinite values, which means that at the start of the algorithm, no effective path from the source node to these nodes has been detected;

[0083] b) Construct a priority queue: Add the starting point node to the priority queue. The priority queue is sorted according to the distance from the node to the starting point, and the node with a smaller distance has a higher priority;

[0084] c) Update distance: Traverse all adjacent nodes of this node, and calculate the distance from the starting point through this node to the adjacent node; if this distance is smaller than the previously recorded distance to the adjacent node, and the remaining battery power of the electric vehicle can support driving to the adjacent node, then update the distance array and add this adjacent node to the priority queue;

[0085] d) Path backtracking: Select the node currently closest to the starting point from the group of nodes to be processed, and move it to the queue of nodes with the confirmed optimal path; subsequently, update the shortest paths of all adjacent nodes of this node, and immediately update if a better path is found. When the group of nodes to be processed is empty, the algorithm terminates, and at this time, the shortest paths of all nodes are determined;

[0086] e) Termination: Stop the search when the priority queue is empty or the end node is reached;

[0087] By dynamically adjusting the weight parameters, this algorithm can not only give priority to ensuring the shortest driving time, but also take into account the shortest distance. At the same time, combined with the remaining battery power of the vehicle, it ensures that users can reach their destinations in the shortest time on the premise of the allowable endurance. This real-time path optimization ability is of great significance for the efficient operation of electric vehicles and the improvement of user experience.

[0088] C3: Spatiotemporal Prediction of EV Charging Load

[0089] Carrying out the spatiotemporal distribution prediction of electric vehicle charging load is the key support for electric vehicle aggregators to implement load regulation strategies. By analyzing the spatiotemporal coupling relationship between the travel patterns of electric vehicles and charging behaviors, the load distribution characteristics in different time periods and geographical regions can be accurately depicted, providing dynamic data support for aggregators to formulate orderly charging strategies. To achieve the accurate prediction of electric vehicle travel characteristics, it is necessary to carry out according to a three-stage process: the first step is to construct a travel demand model, the second step is to plan the paths of electric vehicles, and the third step is to generate charging decisions and execute charging behaviors;

[0090] Specifically include:

[0091] 1) Generate travel demand

[0092] On the basis of establishing the corresponding distribution network model and regional traffic road network model, set the departure time and return time of each vehicle according to the differences in the performance parameters of electric vehicles, and at the same time use the traffic flow matrix to locate the starting point and end point of the route, so as to realize the dynamic simulation of the travel paths of electric vehicles;

[0093] 2) Driving path planning

[0094] Based on the generated travel plan, combined with the geometric characteristics of the transportation network, a route optimization strategy is used for route design, and the travel time is calculated through a real-time road condition assessment model; according to the parameters of the vehicle-mounted power battery, after completing the current travel task, the remaining battery capacity is updated in combination with the power consumption model, and the vehicle position is used as the starting point of the subsequent travel task; the Monte Carlo simulation method is adopted for the random extraction of the multi-period traffic demand matrix, which can completely reproduce the continuous operation route of electric vehicles within the day-night cycle.

[0095] 3) Prediction load reduction

[0096] During the vehicle operation, when it is detected that the remaining power is insufficient to complete the subsequent journey, the system will automatically activate a charging request; the optimal charging station is selected through an intelligent algorithm to complete the energy replenishment operation, and the power consumption data of this period is recorded in real time; the electric vehicle charging loads of each area's charging facilities are superimposed and statistically analyzed to obtain the total load situation of the power network nodes, thereby realizing the spatio-temporal distribution estimation of the electric vehicle charging load.

[0097] Furthermore, the specific steps of step S2 include:

[0098] D1: The objective function of the upper-layer leader electric vehicle aggregator

[0099] The optimization problem of the upper-layer leader aims to maximize the aggregator's revenue, and its optimization goal is: the sum of the revenue from selling electricity to EV users and the revenue from selling electricity to the real-time power market, minus the cost of purchasing electricity from the day-ahead market and the cost of purchasing electricity in the real-time market; the objective function formula is:

[0100]

[0101] In the formula, T represents the total number of periods for EVs to participate in the electricity price response; N represents the total number of EVs; c t (i) represents the charging price of electric vehicle i at time t; p t (i) represents the charging power of electric vehicle i at time t; represents the selling electricity price and purchasing electricity price of the electric vehicle aggregator from the real-time market at time t; respectively represent the electricity purchase quantity and selling electricity quantity of the electric vehicle aggregator from the real-time market at time t; represents the day-ahead contract electricity price at time t; E t represents the day-ahead contract electricity quantity;

[0102] Constraints of the electric vehicle aggregator:

[0103] Charging price constraint:

[0104]

[0105] In the formula, Denote the upper and lower limits of the electricity price in period t; c av Denote the average electricity price during the response period;

[0106] Power balance constraint:

[0107]

[0108] In the formula, M is a sufficiently large positive number, which can be set as the maximum tradable electricity volume in the real-time market; z t Is a Boolean variable, representing the electricity energy trading situation of the electric vehicle aggregator in period t;

[0109] Energy storage device constraint:

[0110]

[0111] In the formula, u t Is a Boolean variable, representing the charge and discharge state of the energy storage device; Represent the upper limits of the charging power and discharging power of the energy storage device respectively; S t Represents the electricity level of the energy storage device in period t; S max Represents the upper limit of the capacity of the energy storage device of the electric vehicle aggregator; η + 、η - Represent the charging efficiency and discharging efficiency of the energy storage device respectively; S0 represents the initial electricity of the energy storage device;

[0112] D2: Lower-level follower EV user objective function

[0113] The optimization problem of the lower-level EV users aims to maximize the expected utility, and at the same time take minimizing the charging cost as the decision-making goal, and its objective function is:

[0114]

[0115] In the formula, T a Is the EV charging period;

[0116] EV constraint conditions:

[0117] EV battery state of charge constraint:

[0118]

[0119] In the formula, δ represents the EV charging efficiency; ρ represents the desired state of charge of the EV user; E max (i) represents the battery capacity of electric vehicle i; E 0 (i) represents the initial electricity of electric vehicle i;

[0120] EV charging power constraint:

[0121] 0 ≤ p t(i) ≤ p max (i) t ∈ T a (31)

[0122]

[0123] Wherein, p max (i) represents the upper limit of the charging power of electric vehicle i;

[0124] D3: Solution method

[0125] The electric vehicle aggregator model in formulas (25)-(28) and the electric vehicle model (29)-(32) constitute a typical master-slave game framework, which essentially belongs to a two-layer optimization system. During the decision-making process on the user side, the charging electricity price c t (i) exists as a known parameter, and only the charging power p t (i) is used as a decision variable. Therefore, its decision model presents a linear characteristic. On the upper layer, the aggregator needs to dynamically adjust the electricity price strategy according to the charging load. At this time, both the electricity price parameter and the power variable are decision variables, resulting in the upper layer model having non-linear and non-convex characteristics. To solve such complex optimization problems, the present invention adopts a mathematical transformation method: First, by means of the duality principle and the KKT (Karush-Kuhn-Tucker) optimality conditions, the user-side model is transformed into a set of constraint conditions of the aggregator model; then the original two-layer structure is transformed into a single-layer mixed-integer linear programming (MILP) model; through this form transformation, the global optimal solution is calculated by the solver CPLEX.

[0126] Furthermore, the solution process of step D3 specifically includes:

[0127] D3-1: Equivalent transformation of the master-slave game model

[0128] In the game process between the electric vehicle aggregator and the electric vehicle, when the EV user makes a charging decision, the electricity price is used as a known parameter. At this time, by introducing the KKT conditions to replace the original linear programming formulas (29)-(32), the lower-layer optimization variables are eliminated, and finally the relationship between the charging power p t (i) and the electricity price c t (i) is obtained; Denote the dual variables of formulas (30)-(31) as μ i 、 θ it , then the KKT conditions of the corresponding linear programming formulas (29)-(32) are:

[0129]

[0130] Wherein, μ iis the dual variable of the equality constraint (30); are the lower-bound constraint and upper-bound constraint dual variables of the inequality (31), respectively; θ it is the dual variable of the equality constraint (32); the lower-level electric vehicle optimization problem is equivalently transformed into a constraint through equations (33)-(37). In the constraint equations (35) and (36), a⊥b means that at most one of the scalars a and b is strictly greater than 0, and the two equations constitute the complementary slackness condition;

[0131] D3-2: Linearization of the KKT conditions

[0132] Since the complementary slackness conditions (35) and (36) are non-linear, they need to be linearized. The present invention uses the big-M theory to reformulate (35) and (36) into the following form:

[0133]

[0134] where M is a sufficiently large positive number;

[0135] D3-3: Linearization of the objective function

[0136] The objective function contains a non-linear term c t (i)p t (i) formed by multiplying the charging price by the EV charging power. The present invention uses the strong duality theory to linearize the non-linear term c t (i)p t (i) in the upper-level optimization model; based on equations (29)-(32), we get:

[0137]

[0138] From equation (42) and the KKT condition constraints, the objective function equation (25) is transformed into:

[0139]

[0140] D3-4: Equivalent transformation of the master-slave game model between the electric vehicle aggregator and electric vehicle owners

[0141] Based on the linear programming duality theorem, KKT conditions, and big-M theory, the two-layer optimization model in the master-slave game framework is equivalently transformed into a mixed-integer linear programming:

[0142]

[0143] The constraints include: equations (26)-(28), equations (33), (34), and equations (37)-(41).

[0144] Further, in step S3, based on the spatio-temporal distribution prediction model of EV charging load, the EV electricity price responsiveness function, and the multi-objective optimization model of electric vehicle aggregators, an orderly charging strategy considering the spatio-temporal flexibility of electric vehicle charging load is proposed. The specific process includes:

[0145] E1: Obtain the spatio-temporal distribution characteristics of the load at each charging station based on the spatio-temporal prediction results of the charging load; generate an optimized electricity price strategy through the master-slave game model;

[0146] E2: The electric vehicle aggregator formulates a dynamic electricity price plan by combining the charging load situation at the station with the charging responsiveness function;

[0147] E3: Simulate the user travel behavior based on the road network traffic flow data, and formulate a user charging plan based on the corrected electricity price information.

[0148] In view of the characteristics of the unbalanced spatio-temporal distribution of charging demand in the travel scenario, the present invention proposes a method for predicting the spatio-temporal distribution of charging load by integrating road network information and distribution network information. Based on the traditional mileage power consumption model, the influence of environmental temperature on vehicle power consumption is comprehensively considered to establish a more practical electric vehicle power consumption model. At the same time, the real-time queuing information of the charging station is combined to improve the prediction accuracy of the spatio-temporal distribution of electric vehicle charging demand. Based on the charging demand prediction results, a dynamic electricity price is formulated by combining the electricity price responsiveness function and the charging demand, and the charging and driving behaviors of users are managed and guided through signals such as charging electricity prices. The invention can effectively improve the charging efficiency of regional charging stations during peak hours and achieve an even distribution of charging demand in time and space. At the same time, it provides a reasonable price guidance signal for the charging decision of vehicle owners to improve the benefits of aggregators and users. It can provide a good reference for constructing an orderly charging strategy for electric vehicles in the travel scenario.

[0149] Based on the "vehicle-road-network" interaction model and the refined modeling of the charging loads of different types of electric vehicles, the present invention accurately predicts the spatio-temporal distribution characteristics of electric vehicle charging loads. At the same time, aiming at the disorderly charging behavior of electric vehicles, based on the spatio-temporal distribution characteristics of electric vehicle charging loads, an orderly charging strategy for the travel scenario of electric vehicles under the guidance of dynamic electricity prices is constructed by combining the electricity price responsiveness function, ultimately achieving the goals of more flexible electricity prices, balanced load distribution, and optimal interests of all parties, providing a reference for electric vehicles to participate in demand-side management and achieve orderly charging.

[0150] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0151] 1. The present invention proposes an orderly charging management architecture for electric vehicles under the load aggregator model. Compared with the traditional centralized management model, the architecture proposed by the present invention can effectively coordinate the operating characteristics of a large number of decentralized charging loads with user needs, and achieve refined regulation of user charging behaviors.

[0152] 2. In view of the obvious spatio-temporal distribution imbalance characteristics of the charging demand of electric vehicles, the present invention constructs a spatio-temporal distribution prediction model for charging loads that integrates traffic road network information and distribution network information. Based on the traditional mileage power consumption model, this model also comprehensively considers the influence of environmental temperature on vehicle power consumption, establishes a more practical electric vehicle power consumption model, and combines real-time queuing information of charging stations to achieve more accurate spatio-temporal distribution prediction of charging loads. Based on the spatio-temporal distribution prediction results of charging loads, a dynamic electricity price is formulated in combination with the electricity price response function and charging demand, effectively guiding the orderly charging of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0153] Figure 1 is the architecture diagram of the orderly charging management of electric vehicles under the load aggregator model;

[0154] Figure 2 is the road network topology structure diagram;

[0155] Figure 3 is the IEEE 33-node system diagram;

[0156] Figure 4 is the flow chart of the spatio-temporal distribution prediction of EV charging loads;

[0157] Figure 5 is the master-slave game architecture diagram between electric vehicle aggregators and electric vehicle users;

[0158] Figure 6 is the flow chart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0159] The following further clarifies the present invention in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, various equivalent forms of modification of the present invention by those skilled in the art fall within the scope defined by the appended claims of this application.

[0160] Referring to Figure 1 and Figure 6 , the present invention provides an orderly charging method for electric vehicle travel scenarios under the guidance of dynamic electricity prices, which includes the following steps:

[0161] S1: Perform spatio-temporal distribution prediction of charging loads through the spatio-temporal distribution prediction model of EV charging loads, and obtain the spatio-temporal distribution characteristics of the loads at each charging station;

[0162] The spatio-temporal distribution prediction model of EV charging load includes a "vehicle-road-network" interaction model and an EV charging load model. The "vehicle-road-network" interaction model includes a road topology structure model, a road resistance model, a distribution network model, and a price responsiveness model.

[0163] The establishment of the "vehicle-road-network" interaction model specifically includes:

[0164] A1: Establish a road topology structure model:

[0165] Considering the characteristics of urban spatial expansion, a circular layout is selected in the modeling process. The present invention uses graph theory tools to study the structural characteristic parameters of the road network and their interaction relationships. The topological structure of the traffic road network is as Figure 2 shown.

[0166] Figure 2 In which, V1 to V9 represent road intersection points, and one-way line segments and two-way line segments represent one-way driving sections and two-way driving sections respectively. Therefore, the mathematical model of the traffic road network is represented by Equation (1):

[0167]

[0168] In the formula, G is the total of the road network system; V is the summary of the intersection points in G; E is the road cluster in G; K is the time period cycle; W represents the passing cost, and node waiting time, passing time and other indicators can be used to quantify and model analyze it;

[0169] The incidence matrix Z presents the Figure 2 traffic road resistance values in, as shown in Equation (2):

[0170]

[0171] In the formula, inf means that it is impassable between node n i and n j ;

[0172] The specific expression form of the incidence matrix Z is in the form of Equation (3):

[0173]

[0174] To construct the spatio-temporal trajectory model of electric vehicles, first determine the travel time parameters (initial time, return time) and spatial parameters (starting and ending points) of the vehicle through OD matrix analysis, and secondly use the Dijkstra algorithm based on real-time traffic information to carry out driving trajectory planning. Its objective functions are as shown in Equations (4) and (5):

[0175]

[0176] In the formula, if nij Belongs to the passage trajectory d i (i), x ij Assign value to 1, otherwise 0;

[0177] A2: Establish a road resistance model:

[0178] Since the electric vehicle charging load studied in this invention is mainly distributed in urban areas, the vehicle traffic efficiency is not only restricted by the road congestion, but also affected by the regulation of road nodes such as traffic lights. When vehicles pass through these control nodes, in addition to dealing with the traffic resistance caused by natural road conditions such as bends and curves, there will also be additional waiting time due to traffic lights. Therefore, in the process of calculating the total time consumption of the whole journey, it is necessary to comprehensively consider the impact of road congestion (road section impedance) and the waiting time at the intersection (node impedance). Therefore, road resistance The mathematical model is expressed as:

[0179]

[0180] Where, Cv i (t) is the road node impedance model; Rv ij (t) is the road section impedance model;

[0181] Based on the saturation level S of traffic density on urban roads, traffic flow is classified into four levels: unobstructed (0 < S ≤ 0.6), slow traffic (0.6 < S ≤ 0.8), dense traffic (0.8 < S ≤ 1.0), and stagnant traffic (1.0 < S ≤ 2.0).

[0182] Road section impedance model: The road section impedance model under different saturation conditions is shown in formula (9):

[0183]

[0184] In the formula, saturation is defined as Q represents the traffic condition of the road, C represents the carrying capacity of the road; t0 represents the travel time of a vehicle in the absence of traffic; α and β represent the influencing factors related to impedance;

[0185] Nodal impedance model:

[0186]

[0187] Where c represents the signal cycle time of the urban traffic light; λ represents the ratio of the green traffic light to the red traffic light; q represents the vehicle arrival rate of the road section;

[0188] By combining equations (9) and (10) with saturation as an intermediate variable, we can obtain the road resistance calculation model:

[0189]

[0190] A3: Establish a distribution network model

[0191] Electric vehicles are replenished with electrical energy through charging stations in the road network, and their charging behavior characteristics will simultaneously affect the spatio-temporal distribution of the grid load, the traffic flow of electric vehicles on the road, and the power consumption load of the charging stations. Therefore, it is necessary to construct a distribution network model adapted to the road network under the framework of the dynamic traffic network according to the energy interaction characteristics between electric vehicles and the grid. The present invention uses the distribution network model of the IEEE 33-node as shown in Figure 3 as the electrical energy supply end to conduct electrical energy transactions with the charging infrastructure.

[0192] The EV charging load of the distribution network node n at time t is shown by the following formula:

[0193]

[0194] In the formula, N represents the number of electric vehicles charged through node n during time period t; P i,t represents the charging power of the i-th electric vehicle during time period t;

[0195] A4: Establish a price responsiveness model

[0196] Electric vehicle users usually tend to choose the nearest charging facility for energy replenishment. This behavior pattern may cause overload of the charging load in local areas, threatening the operation stability of the distribution network. In addition, during peak charging periods, the roads around the charging stations in traffic-intensive areas may generate additional traffic pressure due to vehicle aggregation, exacerbating local congestion. Therefore, it is necessary to design a regulation strategy including spatial electricity price differences, and guide users' charging choices through price signals to achieve the spatio-temporal balanced distribution of the charging load. Based on this, this paper constructs a price responsiveness evaluation model and establishes a dynamic regulation mechanism for electric vehicle charging behavior and regional electricity prices.

[0197] For electric vehicle users to participate in the electricity price regulation mechanism, specific conditions need to be met, that is, a response behavior will occur when the electricity price breaks through a specific critical value, and there is a saturation critical value. When the electricity price difference exceeds this critical value, even if the electricity price difference continues to increase, the number of electric vehicle users participating in the regulation will tend to be stable and will not show continuous growth. Then the mathematical model of the EV user responsiveness index is:

[0198]

[0199] In the formula, Δc m,n (t): The charging electricity price difference between the m-th charging station and the n-th charging station during time period t; ω c (t): The EV responsiveness during time period t; ω c,max : The upper limit value of responsiveness; J m,n and h m,nThey respectively correspond to the response trigger threshold and saturation threshold of electric vehicle users for the m-th charging station and the n-th charging station.

[0200] The establishment of the electric vehicle charging load model specifically includes:

[0201] B1: Modeling of EV charging and driving characteristics

[0202] The battery capacity of an electric vehicle is an important factor affecting its charging behavior. Based on the statistical analysis of the NHTS2017 dataset released by the US Federal Highway Administration, the energy storage capacity of electric vehicles shows a significant Weibull distribution, as shown in Equation (12):

[0203]

[0204] In the formula, α t and β t represent the characteristic values of the Weibull distribution, represents the energy storage capacity of vehicle i;

[0205] The parameter characteristic values for multi-category vehicles are given in Table 1:

[0206] Table 1 Gamma distribution parameters and battery capacity of different types of electric vehicles

[0207]

[0208] Different types of electric vehicles have different charging requirements. Therefore, the present invention sets three different charging departure requirements:

[0209] Private cars: When the vehicle journey ends and the remaining available power is lower than the endurance threshold, the charging requirement will be activated, as represented by Equation (20):

[0210] C t (i) ≤ L td (i)E Tp (13)

[0211] In the formula, C t (i) is the remaining power of electric vehicle i, L td (i) represents the driving path length between the location of vehicle i and the upcoming destination, E Tp : power consumption per kilometer;

[0212] Taxis: Since electric taxis have a relatively high daily charging frequency and long operating mileage, when the remaining power is lower than the power threshold preset by the driver, the charging requirement will be triggered, as represented by Equation (21):

[0213] C t (i) ≤ τC l (i) (14)

[0214] In the formula, C l (i) represents the battery capacity of the electric taxi, and the value range of τ is 0.15 - 0.35;

[0215] For public vehicles: According to their characteristics of multiple trips, if the formula is satisfied, the slow charging form at the destination is adopted; if the charging trigger condition is not met, that is, the existing power cannot reach the destination during driving, it is necessary to find a charging station nearby for charging;

[0216] The charging duration and the charging end time are expressed by formulas (22) and (23):

[0217]

[0218] T f (i) = T a (i) + T w (i) + T c (i) (16)

[0219] In the formula, η c and P c respectively represent the EV charging efficiency (taking 0.8 - 0.9) and the charging power; T a (i): The arrival time of the electric vehicle; T c (i): The charging duration; T f (i) represents the charging end time.

[0220] B2: EV power consumption model considering ambient temperature and vehicle speed

[0221] The energy consumption parameters show complex fluctuation characteristics, and their dynamic changes are restricted by factors such as driving behavior and climate conditions. Most current prediction models adopt the assumption of fixed energy consumption per unit mileage, and this simplified treatment may lead to deviations in the prediction results. However, in actual application scenarios, the energy loss during frequent start-stop of the vehicle and the increased power consumption caused by the use of the air conditioning system will cause the actual energy consumption data to deviate from the prediction model.

[0222] To obtain the relationship between the power consumption and the changes in ambient temperature and vehicle speed, a speed-flow correlation model relying on the real-time road network traffic flow is established, as shown in formula (13):

[0223]

[0224] In the formula, v0: The vehicle driving speed under zero traffic density; C ij : The upper limit of the traffic capacity; χ: The adjustment factor; q ij (t): Dynamic traffic flow data; a, b, γ are road condition parameters;

[0225] By analyzing and deriving equations (13) and (15), two important correlation relationships can be further clarified. On the one hand, the specific relationship between the vehicle's driving speed and driving energy consumption can be clearly obtained, which reflects the influence law of speed change on driving energy consumption; on the other hand, the mathematical expression between air temperature and temperature control energy consumption can also be determined, which reflects the influence result of air temperature change on vehicle energy consumption.

[0226]

[0227] In the formula, respectively represent the start values of heating and cooling of the vehicle's temperature control system; W L 、W R : respectively represent the heating and cooling power consumption of the temperature control system; E: the energy consumption under normal driving conditions; E Tp : the power consumption per kilometer at the air temperature T p .

[0228] B3: M / M / k queuing theory of charging stations

[0229] For the optimization requirements of charging station resource allocation, the present invention uses a multi-channel service model (M / M / k) as a research tool. Through the assumptions of random arrival flow and exponential service duration, the key indicator of average delay duration is focused on, and the dynamic service capabilities of each charging station are statistically analyzed based on this model. The objective functions are shown in equations (17)-(19):

[0230]

[0231]

[0232] In the formula, ε T : the number of charging loads in a certain period; g: the deployment scale of charging terminals; ρ T : the utilization efficiency of charging terminals; μ T : the scale of customers served by charging services per unit time; ψ T : the limit carrying capacity of the charging station per unit time.

[0233] The specific process of predicting the spatio-temporal distribution of charging loads includes:

[0234] C1: OD (origin-destination) analysis

[0235] To study the travel distribution law of electric vehicles in the traffic road network, an OD matrix analysis method is adopted to construct a travel characteristic model for different types of electric vehicles. As the core tool for describing the traffic flow distribution or travel demand characteristics, the OD matrix can intuitively reflect the total travel demand between each origin and destination in a specific area. The travel volume of electric vehicles on different road sections at different time periods can be obtained based on the traffic flow data measured by the traffic management department. Then, the OD origin-destination matrix can be deduced by dynamically changing the traffic volume during the time period, and the Monte Carlo sampling method is used to simulate the departure and return times of electric vehicles and the key parameters such as the travel origin and destination locations. The mathematical model of the OD travel probability matrix is shown in Equation (24):

[0236]

[0237] In the formula, is the travel probability from road network node i to node j; is the number of EVs departing from node i to node j during the time period from T to T + 1; is the number of all EVs departing from traffic network node i;

[0238] C2: Optimal path planning for electric vehicles based on the real-time Dijkstra algorithm

[0239] Relying on the previous OD probability matrix, the travel origin and destination of each vehicle can be accurately located, and the full travel process simulation can be completed by combining the time parameters in the vehicle model. Since there are often multiple feasible path options between the origin and destination, there is a direct connection between the choice of the vehicle's driving path and the power consumption of the drive system and the travel efficiency. Therefore, it is necessary to establish a dynamic path planning system that integrates real-time traffic data, and online generate the optimal driving route considering the battery state and road conditions through the rolling horizon optimization algorithm, so as to ensure the accuracy of energy consumption assessment and spatio-temporal prediction.

[0240] In view of the characteristics of real-time traffic data, the present invention provides a route arrangement for the travel of electric vehicles based on the Dijkstra algorithm. The core logic of this algorithm is to iteratively expand the currently known set of shortest paths and gradually approach the desired node in steps. The specific process is as follows:

[0241] a) Initialization: First, initialize the distance from the starting point to zero, because the distance from a node to itself is always equal to 0; for all other nodes, set their corresponding initial shortest distances to an infinite value, which means that at the start of the algorithm, no effective path from the source node to these nodes has been detected;

[0242] b) Construct a priority queue: Add the starting point node to the priority queue. The priority queue is sorted according to the distance from the node to the starting point, and the node with a smaller distance has a higher priority;

[0243] c) Update distance: Traverse all adjacent nodes of this node, and calculate the distance from the starting point through this node to the adjacent node; if this distance is smaller than the previously recorded distance to the adjacent node, and the remaining power of the electric vehicle can support driving to this adjacent node, then update the distance array and add this adjacent node to the priority queue;

[0244] d) Path backtracking: Select the node currently closest to the starting point from the group of nodes to be processed, and move it to the queue of nodes with the confirmed optimal path; then update the shortest paths of all adjacent nodes of this node, and immediately update if a better path is found. When the group of nodes to be processed is empty, the algorithm terminates, and at this time, the shortest paths of all nodes are determined;

[0245] e) Termination: Stop the search when the priority queue is empty or the end node is reached;

[0246] By dynamically adjusting the weight parameters, this algorithm can not only give priority to ensuring the shortest driving time, but also take into account the shortest distance. At the same time, combined with the remaining power of the vehicle, it ensures that users can reach their destinations in the shortest time on the premise of the allowable endurance. This real-time path optimization ability is of great significance for the efficient operation of electric vehicles and the improvement of user experience.

[0247] C3: Spatiotemporal prediction of EV charging load

[0248] Carrying out the spatiotemporal distribution prediction of electric vehicle charging load is the key support for electric vehicle aggregators to implement load regulation strategies. By analyzing the spatiotemporal coupling relationship between the travel patterns of electric vehicles and charging behaviors, the load distribution characteristics in different time periods and geographical regions can be accurately depicted, providing dynamic data support for aggregators to formulate orderly charging strategies. To achieve the accurate prediction of electric vehicle travel characteristics, it is necessary to carry out according to a three-stage process: the first step is to construct a travel demand model, the second step is to plan the paths of electric vehicles, and the third step is to generate charging decisions and execute charging behaviors;

[0249] As Figure 4 shown, the specific steps of EV charging load prediction are as follows:

[0250] 1) Generate travel demand

[0251] On the basis of establishing the corresponding distribution network model and regional traffic road network model, set the departure time and return time of each vehicle according to the differences in the performance parameters of electric vehicles, and at the same time use the traffic flow matrix to locate the starting point and end point of the route, so as to realize the dynamic simulation of the travel paths of electric vehicles;

[0252] 2) Driving path planning

[0253] Based on the generated travel plan, combined with the geometric characteristics of the transportation network, a route optimization strategy is used for route design, and the travel time is calculated through a real-time traffic condition assessment model; according to the parameters of the vehicle-mounted power battery, the remaining battery capacity is updated in combination with the power consumption model after completing the current travel task, and the vehicle position is used as the starting point for subsequent travel tasks; the Monte Carlo simulation method is used to randomly extract the multi-period traffic demand matrix, which can completely reproduce the continuous operation route of electric vehicles within the day-night cycle.

[0254] 3) Prediction load reduction

[0255] During the vehicle operation, when it is detected that the remaining power is not enough to complete the subsequent journey, the system will automatically activate a charging request; an optimal charging station is selected through an intelligent algorithm to complete the energy replenishment operation, and the power consumption data for this period is recorded in real time; the electric vehicle charging loads of each region's charging facilities are superimposed and statistically analyzed to obtain the total load situation of the power network nodes, thereby realizing the estimation of the spatio-temporal distribution of the electric vehicle charging load.

[0256] S2: Generate an optimized electricity price strategy through a master-slave game model. The electric vehicle aggregator formulates a dynamic electricity price plan in combination with the charging load situation of the station and the charging response function.

[0257] Refer to Figure 5 , step S2 specifically includes:

[0258] D1: The objective function of the upper-layer leader, the electric vehicle aggregator

[0259] The optimization problem of the upper-layer leader aims to maximize the aggregator's revenue. Its optimization goal is: the sum of the revenue from selling electricity to EV users and the revenue from selling electricity to the real-time power market, minus the cost of purchasing electricity from the day-ahead market and the cost of purchasing electricity in the real-time market; the objective function formula is:

[0260]

[0261] In the formula, T represents the total number of periods for EVs to participate in electricity price response; N represents the total number of EVs; c t (i) represents the charging price of electric vehicle i at time t; p t (i) represents the charging power of electric vehicle i at time t; represents the selling electricity price and purchasing electricity price of the electric vehicle aggregator from the real-time market at time t; respectively represent the electricity purchase volume and selling electricity volume of the electric vehicle aggregator from the real-time market at time t; represents the day-ahead contract electricity price at time t; E t represents the day-ahead contract electricity volume;

[0262] Constraints of the electric vehicle aggregator:

[0263] Charging price constraint:

[0264]

[0265] In the formula, represents the upper and lower limits of the electricity price in period t; c av represents the average electricity price during the response period;

[0266] Power balance constraint:

[0267]

[0268] In the formula, M is a sufficiently large positive number, which can be set as the maximum tradable electricity volume in the real-time market; z t is a Boolean variable, representing the electricity energy trading situation of the electric vehicle aggregator in period t;

[0269] Energy storage device constraint:

[0270]

[0271] In the formula, u t is a Boolean variable, representing the charging and discharging state of the energy storage device; represent the upper limits of the charging power and discharging power of the energy storage device respectively; S t represents the electricity level of the energy storage device in period t; S max represents the upper limit of the capacity of the energy storage device of the electric vehicle aggregator; η + 、η - represent the charging efficiency and discharging efficiency of the energy storage device respectively; S0 represents the initial electricity of the energy storage device;

[0272] D2: Objective function of the lower-level follower EV users

[0273] The optimization problem of the lower-level EV users aims to maximize the expected utility, and at the same time takes minimizing the charging cost as the decision-making goal. Its objective function is:

[0274]

[0275] In the formula, T a is the EV charging period;

[0276] EV constraint conditions:

[0277] EV battery state of charge constraint:

[0278]

[0279] In the formula, δ represents the EV charging efficiency; ρ represents the desired state of charge of the EV user; E max (i) represents the battery capacity of electric vehicle i; E 0(i) represents the initial power of electric vehicle i;

[0280] EV charging power constraint:

[0281] 0 ≤ p t (i) ≤ p max (i) t ∈ T a (31)

[0282]

[0283] In the formula, p max (i) represents the upper limit of the charging power of electric vehicle i;

[0284] D3: Solving method

[0285] The electric vehicle aggregator model in equations (25)-(28) and the electric vehicle model (29)-(32) form a typical master-slave game framework, which essentially belongs to a two-layer optimization system. In the user-side decision-making process, the charging electricity price c t (i) exists as a known parameter, and only the charging power p t (i) is used as a decision variable. Therefore, its decision model shows a linear characteristic. On the upper layer, the aggregator needs to dynamically adjust the electricity price strategy according to the charging load. At this time, both the electricity price parameter and the power variable are decision variables, resulting in the upper layer model having non-linear and non-convex characteristics. To solve such complex optimization problems, the present invention adopts a mathematical transformation method: First, with the help of the duality principle and the KKT (Karush-Kuhn-Tucker) optimality conditions, the user-side model is transformed into a set of constraint conditions of the aggregator model; then the original two-layer structure is transformed into a single-layer mixed-integer linear programming (MILP) model; through this form transformation, the global optimal solution is calculated by the solver CPLEX.

[0286] The solving process of step D3 specifically includes:

[0287] D3-1: Equivalent transformation of the master-slave game model

[0288] In the game process between the electric vehicle aggregator and the electric vehicle, when the EV user makes a charging decision, the electricity price is used as a known parameter. At this time, by introducing the KKT conditions to replace the original linear programming in equations (29)-(32), the lower-layer optimization variables are eliminated, and finally the relationship between the charging power p t (i) and the electricity price c t (i) is obtained; Denote the dual variables of equations (30)-(31) as μ i 、 θ it , then the KKT conditions of the corresponding linear programming in equations (29)-(32) are:

[0289]

[0290] where μ i is the dual variable of the equality constraint (30); are the lower bound constraint and the upper bound constraint dual variables of the inequality (31), respectively; θ it is the dual variable of the equality constraint (32); the lower-level electric vehicle optimization problem is equivalently transformed into a constraint through equations (33)-(37). In the constraint equations (35) and (36), a⊥b means that at most one of the scalars a and b is strictly greater than 0, and the two equations form a complementary slackness condition;

[0291] D3-2: Linearization of KKT conditions

[0292] Since the complementary slackness conditions (35) and (36) are non-linear, they need to be linearized. In the present invention, the big-M theory is used to re-express equations (35) and (36) in the following form:

[0293]

[0294]

[0295] where M is a sufficiently large positive number;

[0296] D3-3: Linearization of the objective function

[0297] The objective function contains a non-linear term c t (i)p t (i) formed by multiplying the charging price by the EV charging power. In the present invention, the strong duality theory is used to linearize the non-linear term c t (i)p t (i) in the upper-level optimization model; based on equations (29)-(32), the following is obtained:

[0298]

[0299] From equation (42) and the KKT condition constraint equations, the objective function equation (25) is transformed into:

[0300]

[0301] D3-4: Equivalent transformation of the master-slave game model between the electric vehicle aggregator and the electric vehicle owners

[0302] Based on the linear programming duality theorem, the KKT conditions, and the big-M theory, the two-layer optimization model in the master-slave game framework is equivalently transformed into a mixed-integer linear programming:

[0303]

[0304] The constraints include: Equations (26)-(28), Equations (33), (34), and Equations (37)-(41).

[0305] S3: Simulate the user travel behavior based on the road network traffic flow data, and formulate the user charging plan based on the corrected electricity price information.

[0306] Refer to Figure 6 , based on the EV charging load spatio-temporal distribution prediction model, the EV electricity price responsiveness function, and the multi-objective optimization model of the electric vehicle aggregator, an orderly charging strategy considering the spatio-temporal flexibility of the electric vehicle charging load is proposed. The specific process includes:

[0307] E1: Obtain the spatio-temporal distribution characteristics of the load at each charging station based on the spatio-temporal prediction results of the charging load; generate the optimized electricity price strategy through the master-slave game model;

[0308] E2: The electric vehicle aggregator formulates a dynamic electricity price plan by combining the charging load situation at the station with the charging responsiveness function;

[0309] E3: Simulate the user travel behavior based on the road network traffic flow data, and formulate the user charging plan based on the corrected electricity price information.

Claims

1. An orderly charging method for the travel scenarios of electric vehicles under the guidance of dynamic electricity prices, characterized in that The steps include: S1: Use the EV charging load spatiotemporal distribution prediction model to predict the spatiotemporal distribution of charging load and obtain the spatiotemporal distribution characteristics of the load at each charging station; S2: The optimized electricity price strategy is generated through the master-slave game model. The electric vehicle aggregator formulates a dynamic electricity price plan based on the charging load of the site and the charging responsiveness function. S3: Simulate user travel behavior based on road network traffic flow data and formulate user charging plans based on the revised electricity price information.

2. The orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance according to claim 1, characterized in that The EV charging load spatiotemporal distribution prediction model in step S1 includes a "vehicle-road-network" interaction model and an electric vehicle charging load model. The "vehicle-road-network" interaction model includes a road topology model, a road resistance model, a distribution network model, and an electricity price responsiveness model.

3. The orderly charging method for the electric vehicle travel scenario under dynamic electricity price guidance according to claim 2, wherein The establishment of the "vehicle-road-network" interaction model in step S1 specifically includes: A1: Establish road topology model: The mathematical model of the traffic network is expressed by formula (1): Where G is the total road network system; V is the intersection summary in G; E is the road cluster in G; K is the time period; W represents the cost of travel; The correlation matrix Z presents the traffic road resistance value, as shown in formula (2): where inf represents node n i and n j are impassable; The specific expression of the correlation matrix Z is in the form of formula (3): To construct the spatiotemporal trajectory model of electric vehicles, the travel time parameters and spatial parameters of the vehicles are first determined through OD matrix analysis. Then, the Dijkstra algorithm based on real-time traffic information is used to carry out driving trajectory planning. Its objective function is shown in Equations (4) and (5): Wherein, if n ij belongs to the passing trajectory d i (i), x ij is assigned 1, otherwise 0; A2: Establish a road resistance model: Road resistance The mathematical model is expressed as: where, Cv i (t) is the road node impedance model; Rv ij (t) is the link impedance model; A3: Build a distribution network model The EV charging load at distribution network node n at time t is as follows: where N represents the number of electric vehicles charged through node n during period t; P i,t represents the charging power of the i-th electric vehicle during period t; A4: Establishing an electricity price responsiveness model The mathematical model of EV user responsiveness index is: Where, Δc m,n (t): the charging price difference between the m-th charging station and the n-th charging station during the t period; ω c (t): the EV responsiveness during the t period; ω c,max : the upper limit value of responsiveness; J m,n and h m,n correspond to the response trigger critical value and the saturation threshold of the electric vehicle users for the m-th charging station and the n-th charging station respectively.

4. The orderly charging method for the electric vehicle travel scenario under dynamic electricity price guidance according to claim 3, characterized in that, In step A2, traffic flow is classified into four levels based on the saturation level S of traffic density on urban roads: unobstructed (0 < S ≤ 0.6), slow traffic (0.6 < S ≤ 0.8), dense traffic (0.8 < S ≤ 1.0), and stagnant traffic (1.0 < S ≤ 2.0); Road section impedance model: The road section impedance model under different saturation conditions is shown in formula (9): In the formula, the saturation is defined as Q represents the road traffic flow condition, C represents the carrying capacity of the road; t0 represents the passing time of the vehicle when there is no traffic flow; α and β represent the impedance-related influence factors; Nodal impedance model: Where c represents the signal cycle time of the urban traffic light; λ represents the ratio of green traffic lights to red traffic lights; q represents the vehicle arrival rate on the road section; By combining equations (9) and (10) with saturation as an intermediate variable, we can obtain the road resistance calculation model: 。 5. The orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance according to claim 4, characterized in that, The establishment of the electric vehicle charging load model in step S1 specifically includes: B1: EV Charging and Driving Characteristics Modeling The energy storage capacity of electric vehicles shows a significant Weibull distribution, as shown in formula (12): where α t and β t represent the characteristic values of the Weibull distribution, represents the energy storage capacity of vehicle i; B2: EV power consumption model considering ambient temperature and vehicle speed In order to obtain the power consumption relationship between ambient temperature and vehicle speed changes, a speed and flow correlation model based on real-time road network traffic flow was established, as shown in formula (13): where \(v_0\) is the vehicle driving speed under zero traffic density; \(C\) ij is the upper limit of traffic capacity; \(\chi\) is the adjustment factor; \(q\) ij (t) is Dynamic traffic flow data; a, b, γ are road condition parameters; By analyzing and deducing Equations (13) and (15), we can further clarify two important correlations: In the formula, respectively represent the starting values for heating and cooling of the vehicle temperature control system; W L , W R : respectively represent the heating and cooling power consumptions of the temperature control system; E: the energy consumption under normal driving conditions; E Tp : the power consumption per kilometer at the air temperature T p ; B3: Charging Station M / M / k Queuing Theory The objective function is shown in equations (17)-(19): where ε T : the number of charging loads in a certain period; g: the deployment scale of charging terminals; ρ T : the utilization efficiency of charging terminals; μ T : the customer scale of charging services per unit time; ψ T : the limit carrying capacity of the charging station per unit time.

6. The orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance according to claim 5, characterized in that In step B1, three different charging departure demands are set: Private cars: When the vehicle's journey ends and the remaining available power is lower than the endurance threshold, the charging demand will be activated, as represented by equation (20): C t (i) ≤ L td (i) E Tp (20) Where C t (i) is the remaining power of electric vehicle i, L td (i) represents the driving path length between the location of vehicle i and the upcoming destination, E Tp : power consumption per kilometer; Taxis: When the remaining power is lower than the power threshold preset by the driver, the charging demand will be triggered, as represented by equation (21): C t (i) ≤ τC l (i) In Equation (21), C l (i) represents the battery capacity of the electric taxi; Public vehicles: According to their characteristics of multiple trips, if the condition is met, slow charging at the destination is adopted; if the charging trigger condition is not met, that is, the existing power cannot reach the destination during driving, a charging station needs to be found nearby for charging; The charging duration and the charging end time are represented by equations (22) and (23): T f (i) = T a (i) + T w (i) + T c (i)(23) where η c and P c represent the EV charging efficiency and charging power respectively; T a (i): the arrival time of the electric vehicle; T c (i): Charging duration; T f (i) represents the end moment of charging.

7. A method for orderly charging of an electric vehicle travel scenario under dynamic electricity price guidance according to claim 6, characterized in that, The specific process of predicting the spatio-temporal distribution of the charging load in step S1 includes: C1: OD analysis The mathematical model of the OD travel probability matrix is shown in equation (24): In the formula, is the travel probability from road network node i to node j; is the number of EVs departing from node i to node j during the period from T to T + 1; is the number of all EVs departing from traffic network node i; C2: Optimal path planning for electric vehicles based on the real-time Dijkstra algorithm The specific process is as follows: a) Initialization: First, initialize the distance from the starting point to zero; for all other nodes, set their corresponding initial shortest distances to infinity, which means that at the start of the algorithm, no effective path from the source node to these nodes has been detected; b) Construct a priority queue: Add the starting node to the priority queue, and the priority queue is sorted according to the distance from the node to the starting point; c) Update the distance: Traverse all adjacent nodes of this node, and calculate the distance from the starting point through this node to the adjacent node; if this distance is smaller than the previously recorded distance to the adjacent node, and the remaining power of the electric vehicle can support driving to this adjacent node, then update the distance array and add this adjacent node to the priority queue; d) Path backtracking: Select the node currently closest to the starting point from the group of nodes to be processed, and move it to the queue of nodes with the confirmed optimal path; then update the shortest paths of all adjacent nodes of this node, and immediately update if a better path is found. When the group of nodes to be processed is empty, the algorithm terminates, and at this time, the shortest paths of all nodes have been determined; e) Termination: When the priority queue is empty or the end node is reached, stop the search; C3: Spatio-temporal prediction of EV charging load Specifically, it includes: 1) Generate travel demands Based on the establishment of the corresponding distribution network model and regional traffic road network model, set the departure time and return time of each vehicle according to the differences in the performance parameters of electric vehicles. At the same time, use the traffic flow matrix to locate the starting point and end point of the route, so as to realize the dynamic simulation of the travel path of electric vehicles; 2) Driving path planning Based on the generated travel plan, combined with the geometric characteristics of the traffic network, use the path optimization strategy to design the route, and calculate the travel time through the real-time road condition evaluation model; according to the parameters of the in-vehicle power battery, update the remaining battery capacity after completing the current travel task, and use the vehicle position as the starting point for subsequent travel tasks; adopt the Monte Carlo simulation method to randomly extract the multi-period traffic demand matrix, which can completely reproduce the continuous operation route of electric vehicles within the day-night cycle; 3) Predictive load reduction During the operation of the vehicle, when it is detected that the remaining power is insufficient to complete the subsequent journey, the system will automatically activate a charging request; select the optimal charging station through an intelligent algorithm to complete the energy replenishment operation, and record the power consumption data during this period in real time; superimpose and statistically analyze the electric vehicle charging loads of charging facilities in each region to obtain the total load situation of the power network nodes, and then realize the estimation of the spatio-temporal distribution of the electric vehicle charging load.

8. A method for orderly charging of an electric vehicle travel scenario under dynamic electricity price guidance according to claim 7, characterized in that, The specific steps of step S2 include: D1: The objective function of the upper-layer leader electric vehicle aggregator The optimization problem of the upper-layer leader aims to maximize the aggregator's revenue. Its optimization goal is: the sum of the revenue from selling electricity to EV users and the revenue from selling electricity to the real-time power market, minus the cost of purchasing electricity from the day-ahead market and the cost of purchasing electricity from the real-time market; the objective function formula is: Where T represents the total number of time periods for EVs to participate in the electricity price response; N represents the total number of EVs; c t (i) represents the charging price of electric vehicle i at time t; p t (i) represents the charging power of electric vehicle i at time t; represents the selling and purchasing electricity prices of the EV aggregator from the real-time market at time t; respectively represent the electricity purchasing and selling quantities of the EV aggregator from the real-time market at time t; represents the day-ahead contract electricity price at time t; E t represents the day-ahead contract electricity quantity; Constraints of the electric vehicle aggregator: Charging price constraint: In the formula, represents the upper and lower limits of the electricity price in period t; c av represents the average electricity price during the response period; Power balance constraint: where M is a positive number; z t is a Boolean variable representing the electricity energy trading situation of the electric vehicle aggregator during period t; Energy storage device constraint: where, u t is a Boolean variable representing the charge and discharge state of the energy storage device; respectively represent the upper limits of the charging power and discharging power of the energy storage device; S t represents the power level of the energy storage device at time t; S max represents the upper limit of the capacity of the energy storage device of the electric vehicle aggregator; η + , η - respectively represent the charging efficiency and discharging efficiency of the energy storage device; S0 represents the initial power of the energy storage device; D2: The objective function of the lower-layer follower EV users The optimization problem of the lower-layer EV users aims to maximize the expected utility, and at the same time takes minimizing the charging cost as the decision-making goal. Its objective function is: where T a is the EV charging period; EV constraints: EV battery state of charge constraint: where δ represents the EV charging efficiency; ρ represents the state of charge expected by EV users; E max (i) represents the battery capacity of electric vehicle i; E 0 (i) represents the initial power of electric vehicle i; EV charging power constraint: 0 ≤ p t (i) ≤ p max (i) t ∈ T a (31) where p max (i) represents the upper limit of the charging power of electric vehicle i; D3: Solving method With the help of the duality principle and the KKT optimality conditions, convert the user-side model into a set of constraint conditions of the aggregator model; Furthermore, convert the original two-layer structure into a single-layer mixed-integer linear programming model; through this form conversion, calculate the global optimal solution through the solver CPLEX.

9. The orderly charging method for the electric vehicle travel scenario under dynamic electricity price guidance according to claim 8, characterized in that The specific solving process of step D3 includes: D3-1: Equivalent transformation of the master-slave game model In the game between the electric vehicle aggregator and electric vehicles, when EV users make charging decisions, the electricity price is a known parameter. At this time, by introducing the KKT conditions to replace the original linear programming equations (29)-(32) and eliminating the lower-layer optimization variables, the charging power p is finally obtained. t (i) The relationship with the electricity price c t (i); Denote the dual variables of equations (30)-(31) as μ i 、 θ it , then the KKT conditions for the corresponding linear programming equations (29)-(32) are: where, μ i is the dual variable of the equality constraint (30); are the lower bound constraint and the upper bound constraint dual variables of the inequality (31) respectively; θ it is the dual variable of the equality constraint (32); the lower-level electric vehicle optimization problem is equivalent to the constraint through equations (33)-(37). In the constraint equations (35) and (36), a⊥b means that at most one of the scalars a and b is strictly greater than 0, and the two equations constitute the complementary slackness condition; D3-2: Linearization of the KKT conditions Use the big-M theory to reformulate equations (35) and (36) into the following form: Where M is a positive number; D3-3: Linearization of the objective function The strong duality theory is used to linearize the non - linear term c t (i)p t (i) in the upper - layer optimization model; based on equations (29)-(32), we get: From equation (42) and the KKT condition constraint equations, the objective function equation (25) is transformed into: D3-4: Equivalent transformation of the electric vehicle aggregator and the master-slave game model of electric vehicles Based on the linear programming duality theorem, the KKT conditions and the big-M theory, equivalently transform the two-layer optimization model under the master-slave game framework into a mixed-integer linear programming: The constraints include: equations (26)-(28), equations (33), (34) and equations (37)-(41).

10. The orderly charging method for an electric vehicle travel scenario under dynamic electricity price guidance according to claim 9, characterized in that, In step S3, based on the EV charging load spatio-temporal distribution prediction model, the EV electricity price responsiveness function, and the multi-objective optimization model of the electric vehicle aggregator, an orderly charging strategy considering the spatio-temporal flexibility of the electric vehicle charging load is proposed. The specific process includes: E1: Obtain the spatio-temporal distribution characteristics of the load of each charging station based on the spatio-temporal prediction results of the charging load; generate an optimized electricity price strategy through the master-slave game model; E2: The electric vehicle aggregator formulates a dynamic electricity price plan by combining the charging load situation of the station and the charging responsiveness function; E3: Simulate the user travel behavior based on the road network traffic flow data, and formulate the user charging plan based on the corrected electricity price information.

Images