An electric vehicle charging coordination and regulation method and system

Abstract

The present application relates to electric vehicle charging management technical field, specifically relates to a kind of electric vehicle charging coordination control method and system.The method is by constructing the linearization power flow model of radial distribution network, photovoltaic output and electric vehicle charging load are included in unified power flow calculation framework, obtain node voltage and line power distribution;The decision-making process of electric vehicle charging station is modeled as Markov decision process, to each charging station as independent agent, according to the space-time load distribution of charging station, power flow constraint and demand response signal dynamically adjusts charging pricing strategy;While introducing communication network modeling and communication cost quantification mechanism, information interaction overhead is included in reward function design, so that Markov decision process realizes collaborative trade-off between profit maximization and communication resource consumption.The present application realizes the intelligentization, order and efficient charging control of electric vehicle group in vehicle network interaction environment under the condition of traffic, power and communication multi-network coordination.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle charging management technology, specifically to a method and system for coordinated regulation of electric vehicle charging based on spatiotemporal extended user balance and demand response. Background Technology

[0002] With the widespread adoption of electric vehicles, their proportion as a new type of mobile load connected to the distribution network is constantly increasing, becoming a key factor affecting the operation characteristics and safety of the distribution network. During peak hours and other specific periods, the simultaneous charging of a large number of electric vehicles can easily lead to operational risks such as node voltage exceeding limits and line overload. Furthermore, the charging demand of electric vehicles exhibits significant spatiotemporal coupling characteristics, being closely related not only to the vehicle's departure time, travel route, and initial state of charge, but also constrained by traffic congestion and the geographical distribution of charging facilities.

[0003] Existing electric vehicle charging regulation methods mainly employ centralized optimization or heuristic scheduling approaches, aiming to smooth peak and valley loads by setting charging periods, optimizing charging power, or implementing price incentives. However, these methods typically assume that electric vehicle charging load can be directly controlled, neglecting user autonomy and actual travel patterns, thus limiting the effectiveness of regulation strategies in real-world scenarios. Furthermore, traditional methods often fail to jointly model traffic flow evolution and power distribution network flow, lacking simultaneous consideration of traffic network conditions and power distribution network operating conditions. This can easily lead to inconsistencies between traffic optimization results and power grid operating conditions, making it difficult to support large-scale, multi-period vehicle-to-grid interactive control requirements.

[0004] To characterize the travel and charging behavior of electric vehicles, existing technologies have proposed a unified modeling of vehicle route selection, time selection, and charging selection through a spatiotemporally extended traffic network. In this model, traffic conditions and vehicle behavior are expressed spatiotemporally, which can be used to analyze the dynamic flow distribution of vehicles in the traffic network. For example, the Chinese invention patent CN120087727B, entitled "A Highway Network Planning Method and System Based on an Extended Traffic Network Flow Allocation Model," provides a theoretical basis for modeling at the traffic system level.

[0005] Therefore, although existing methods have considered electric vehicle behavior, traffic network characteristics, and power distribution network operational constraints to varying degrees, most solutions remain at the level of hierarchical optimization or sequential decision-making, failing to achieve joint decision-making and real-time coupling between the traffic network and the power network. They also lack an effective closed-loop feedback mechanism between charging pricing, power grid operational constraints, and user travel and charging behavior, making it difficult to form a stable dynamic collaborative control framework. Furthermore, with the expansion of charging station scale and the increase in demand response interaction frequency, information interaction between charging stations and between charging stations and the power distribution network becomes increasingly frequent, and its interactive operation relies on the support of communication networks. However, existing technologies typically assume that information can be transmitted in real time without cost or constraints, without modeling and analyzing the information transmission volume, bandwidth consumption, and communication costs on the communication links. In large-scale charging network scenarios, frequent price updates and state interactions may bring significant communication burdens, affecting the overall economic efficiency and scalability of the system. Therefore, the efficiency and stability of electric vehicle charging collaborative control still need to be improved. Summary of the Invention

[0006] Purpose of the invention: In view of the technical problems existing in the background art, the present invention provides a method for coordinated regulation of electric vehicle charging, which solves the problem that the charging regulation efficiency and stability of electric vehicles are still not high. The present invention also discloses an electric vehicle charging coordinated regulation system.

[0007] Technical solution: In a first aspect, the present invention provides a method for coordinated regulation of electric vehicle charging, the method comprising: The optimal travel and charging routes for electric vehicles are calculated based on the traffic distribution model of the extended transportation network, thereby obtaining the spatiotemporal load distribution of each charging station at different times. A communication network model between electric vehicle charging stations is constructed. Based on the communication network model, the total amount of information transmitted on each communication edge is determined, and then the communication cost of all communication links is obtained. The decision-making process of the electric vehicle charging system is modeled as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. In the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; a reward is given when the total charging demand remains within a specified range; and no compensation is given when the total charging demand falls below a set lower limit. The reward function is configured to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on a radial distribution network, thus coupling the operation of the transportation system and the power system.

[0008] Furthermore, including: The construction of the communication network model between electric vehicle charging stations includes: The information network between electric vehicle charging stations is modeled as a communication network CN. It is assumed that each controllable switch is equipped with a communication exchange CS for sending and receiving information. Each CS is considered a node in the communication network, and the set of all communication nodes is denoted as . The controllable switch is a controllable switch device corresponding to the information acquisition and control node of the communication network or the power distribution side.

[0009] Considering that each communication node needs to both broadcast local operating status information and receive control signals from neighboring nodes or the upper-level power grid, the communication link between any two adjacent nodes must support bidirectional communication. The set of bidirectional edges in the communication network is denoted as […]. any one of the edges Represents a node With nodes There is a two-way communication link between them, therefore, ; Since information exchange originates from load transmission, voltage feedback, and demand response signal transmission between charging stations and distribution network nodes, the set of commodity pairs is defined as follows: ; in, Indicates a collection of charging stations. Represents the set of nodes in the distribution network. For charging stations With distribution network nodes The mapping coefficients, when =1 indicates a charging station Access Node , Represents a pair of goods. As the source node for sending information, For the destination node of information reception, a set This represents all node pairs that actually exchange information during the coupled operation of transportation and power systems; In terms of communication topology, a depth-first search (DFS) algorithm is used for each product pair. Search for its shortest communication path, let... Indicates the product pair The set of communication edges contained in the corresponding shortest communication path, and path edge association variables defined based on path information: ; Among them, if the communication side Belongs to path If the edge carries 1 bit of information, then the edge carries no information flow for the item pair; otherwise, the edge does not carry information flow for the item pair.

[0010] Furthermore, including: The step of determining the total amount of information transmitted on each communication edge based on the communication network model, and then obtaining the communication cost of all communication links, includes: The total amount of information transmitted on each communication edge is represented as: ; in, Indicates communication edge At any moment The total number of bits transmitted cumulatively; Indicates the product pair At any moment The number of source-to-destination transmitted bits; Therefore, the overall communication cost of a communication network can be expressed as the sum of the amount of information transmitted on all communication links: ; in, Indicates time The total communication cost of a communication network, an indicator used to quantify the communication resource consumption resulting from information exchange during the coupling and control of transportation and power; commodity pairs At any moment Number of transmitted bits Based on current pricing actions Pricing action compared to the previous moment The differences, as well as the changes in distribution network node load and voltage caused by the pricing action, are jointly determined, specifically as follows: ;in, These represent the basic number of bits per instance for price broadcasting, load reporting, and voltage feedback, respectively. The preset relative price change threshold, This indicates that the relative changes of each price component for each charging station and each charging type in the price action vector are calculated item by item. This is an indicator function that triggers a price broadcast only when the relative price change reaches the threshold; otherwise, this term is zero. express The load change quantification level after rounding up; express The quantization level of voltage change after rounding up; For information quantization resolution; Indicates charging station At any moment Load changes; Indicates distribution network node At any moment Voltage changes; and Each product pair The charging stations and distribution network nodes associated with the corresponding information interaction; for non-price broadcast product pairs, the price broadcast item is 0; for price broadcast product pairs, the same price broadcast for the same recipient is only counted once in its corresponding product pair to avoid the same price broadcast being counted repeatedly in all product pairs.

[0011] Furthermore, including: The method of dynamically adjusting the pricing strategy based on the spatiotemporal load distribution of the charging station, treating each electric vehicle charging system as an intelligent agent, includes: The agent's decisions depend on the current observable state, which contains enough information to predict future transitions, thus satisfying the Markov property assumption. The Markov decision process consists of: State space: at time step Electric vehicle charging station The state vector integrates local operating variables, fast and slow charging demand from the traffic allocation model of the extended transportation network, and grid-side power status. The state vector is defined as follows: ; in, This represents the total charging load at the previous moment. and To meet the current needs of fast charging and slow charging, and The prices for fast charging and slow charging at the previous moment, The electricity purchase price at the previous moment. Electric vehicle charging stations At any moment The upper limit of the demand response load threshold is set in fast charging mode. Electric vehicle charging stations At any moment The lower limit of the demand response load threshold is set in fast charging mode. Electric vehicle charging stations At any moment The upper limit of the demand response load threshold is set in slow charging mode. Electric vehicle charging stations At any moment The lower limit of the demand response load threshold is set in slow charging mode. As a current demand response signal, express Always connected to the power distribution network nodes voltage, express Always connected to the power distribution network The lines in The corresponding active current, This refers to the total number of electric vehicle charging stations in an electric vehicle charging system. Total number of time steps; Motion space: Electric vehicle charging station The action corresponds to the time. The price pair for fast charging and slow charging is represented as follows: ; in, and Charging stations At any moment The prices for fast charging and slow charging are set; therefore, the agent's policy network outputs the prices for fast charging and slow charging to achieve an optimal balance between revenue and demand response participation.

[0012] Furthermore, including: The reward function is configured to take into account energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links, including: The single-step reward function at time 1 Defined as: in, Indicates charging station At any moment Energy trading revenue; Indicates charging station At any moment Rewards or penalties obtained based on demand response mechanisms; Indicates time The total communication cost of a communication network; The communication cost weighting coefficient is used to adjust the trade-off between revenue and communication resource consumption. Therefore, by maximizing the cumulative reward function, the agent can learn an adaptive joint pricing strategy, thereby maximizing its own profit while ensuring network stability and meeting system-level demand response objectives.

[0013] Furthermore, including: The energy trading revenue is expressed as follows: ; in, Electric vehicle charging stations At any moment The revenue from fast charging energy trading Electric vehicle charging stations At any moment The revenue from slow-charging energy trading and These represent the purchase price of electricity and the incentive coefficient for demand response, respectively. and For fast and slow charging loads from the traffic distribution model of the extended traffic network; The rewards or penalties obtained based on the demand response mechanism are represented as follows: ; Among them, binary variables Indicates electric vehicle charging station Is it connected to the distribution network node? , , The upper and lower limits of the demand response load threshold for each mode are defined. , For fast charging mode, Slow charging mode For fast or slow charging loads from traffic allocation models of extended transportation networks.

[0014] Furthermore, including: The steady-state operating parameters of the radial distribution network are achieved by constructing a linearized power flow model, including: This represents a radial distribution network, as shown in the diagram. ,in, Indicates a collection of power distribution lines. Represents the set of busbars; A linear power flow model is constructed, described as follows: ;in, Indicates power distribution network Middle node Is it a reference bus, i.e., a balancing node? If yes, the value is 1; otherwise, the value is 0. Indicates power distribution network The reactive power exchanged between the substation and the upper-level power grid; For power distribution network Central route reactive power, For power distribution network Middle node Corresponding to the reactive power output of the photovoltaic unit, For power distribution network Central route reactive power, For power distribution network Middle node Total reactive load, Indicates the terminal node; Represents a node The set of child nodes, i.e., all nodes that satisfy the path nodes A set; This represents the set of numbers for the power distribution network, i.e. Indicates the first A power distribution network; Indicates power distribution network The active power exchanged between the substation and the upper-level power grid; For power distribution network Central route active power, For power distribution network Middle node The corresponding active power output of the photovoltaic unit, For power distribution network Middle node Total active load; ; in, Indicates the first Nodes in a power distribution network voltage amplitude, Represents a node voltage amplitude, Indicates the first Lines in a power distribution network The resistance, Indicates the first Lines in a power distribution network The reactance; ; in, For power distribution network Middle node The maximum value of the active power output corresponding to the photovoltaic unit; ; in, Represents network Middle node The power factor corresponding to the photovoltaic unit; ; in, Indicates parameters Follows probability distribution ; ; in, Represents a node In time Voltage amplitude at that time Represents a node The minimum value of the voltage amplitude, Represents a node The maximum value of the voltage amplitude; in, and They represent the power distribution network. Central route At any moment The active and reactive power flow.

[0015] Furthermore, including: The demand response mechanism is obtained by integrating the flow allocation model of the extended transportation network and the linearized power flow model based on the radial distribution network, including: The total active power load of the distribution network is expressed as: ; in, For power distribution network Middle node In time Total active power load at time For power distribution network Middle node In time The base load at that time This represents the load demand of fast charging stations obtained from the traffic distribution model of the extended transportation network. This represents the load demand of slow charging stations obtained from the traffic distribution model of the extended transportation network. To determine the charging station Is it connected to the power distribution network? nodes The corresponding mapping coefficients, when Time indicates charging station Connected to the power distribution network nodes , A collection of electric vehicle charging stations. It represents a set of time segments; thus, it integrates the traffic distribution model and the linearized flow model of the extended traffic network to form a coordinated demand response framework, thereby treating each electric vehicle charging station as an intelligent agent that can dynamically adjust its pricing strategy based on demand impact signals.

[0016] A second aspect of the present invention provides an electric vehicle charging coordinated control system, the system comprising: The spatiotemporal load distribution determination module is used to calculate the optimal travel and charging routes for electric vehicles based on the traffic allocation model of the extended transportation network, thereby obtaining the spatiotemporal load distribution of each charging station in different time periods. The communication network construction module is used to construct a communication network model between electric vehicle charging stations, determine the total amount of information transmitted on each communication edge based on the communication network model, and then obtain the communication cost of all communication links. The decision-making module models the decision-making of the electric vehicle charging system as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. During the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; a reward is given when the total charging demand remains within a specified range; and no compensation is given when the total charging demand falls below a set lower limit. The reward function is configured to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on a radial distribution network, thus coupling the operation of the transportation system and the power system.

[0017] Finally, the present invention also provides a computer system including a memory, a processor, and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of the method as described above.

[0018] Beneficial effects: Compared with the prior art, the present invention has the following advantages: This application models and analyzes the information transmission volume, bandwidth usage, and communication cost on the communication link, so that in the scenario of large-scale charging network, frequent price updates and status interactions do not affect the overall economic efficiency and scalability of the system. This application constructs a linearized power flow model based on a radial distribution network to reflect power flow constraints. This model is combined with the flow allocation model of the extended transportation network, namely the ETAP-UE model, to establish a coordinated demand response (DR) framework, thereby coupling the operation of the transportation system and the power system. After integrating these two structures, each electric vehicle charging station acts as an intelligent agent, capable of dynamically adjusting its pricing strategy based on DR signals and network status. The design of the state space vector enables the agent to simultaneously perceive the traffic and power network states, thereby capturing their spatiotemporal coupling characteristics. The setting of the action space vector allows the agent's strategy network output price to achieve an optimal balance between revenue and DR participation. The reward function quantifies the instantaneous revenue of each EVCS at a single time step, comprehensively considering the results of energy trading and DR participation, thereby maximizing its own profit while ensuring network stability and meeting the system-level DR objective. This achieves joint modeling of the operation status of the transportation network and distribution network for electric vehicle charging regulation, and realizes synchronous optimization of travel behavior and charging behavior. This makes traditional optimization methods and reinforcement learning algorithms converge stably and have more sufficient exploration capabilities in high-dimensional continuous control scenarios.

[0019] Therefore, this application introduces communication network modeling and communication cost constraints into the transportation-electricity coupling optimization framework, and constructs a multi-network collaborative mechanism of transportation, electricity and communication, which becomes a key issue in improving the efficiency and stability of electric vehicle charging collaborative regulation. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a simplified flowchart of the electric vehicle charging coordinated control method according to an embodiment of the present invention; Figure 2 This is a flowchart of the electric vehicle charging coordinated regulation method based on spatiotemporal extended user balance and demand response as described in an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the convergence curves of reinforcement learning under demand response and no demand response conditions, as described in an embodiment of the present invention. Figure 4 This is a schematic diagram of the electric vehicle charging coordinated control system according to an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention integrates intelligent collaborative control methods based on spatiotemporal travel characteristics, power distribution network flow constraints, and demand response signals to guide the orderly charging behavior of electric vehicle groups and ensure the safe and stable operation of the power distribution network. Therefore, the electric vehicle charging collaborative control method based on spatiotemporally extended user balance and demand response, such as... Figure 1 As shown, the method includes: Step 1: Calculate the optimal travel and charging routes for electric vehicles based on the traffic distribution model of the extended traffic network, thereby obtaining the spatiotemporal load distribution of each charging station at each time period. Step 2: Construct a communication network model between electric vehicle charging stations, determine the total amount of information transmitted on each communication edge based on the communication network model, and then obtain the communication cost of all communication links; Step 3: Model the decision-making of the electric vehicle charging system as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. In the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; a reward is given when the total charging demand remains within a specified range; and no compensation is given when the total charging demand falls below a set lower limit. The reward function is set to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on the radial distribution network, thus achieving coupling between the transportation system and the power system.

[0024] Specifically, such as Figure 2 As shown, it includes the following steps: Step S1: Collect the traffic network topology, distribution network node parameters, electric vehicle travel demand and initial state of charge (SoC), and construct a spatiotemporal extended traffic network model that considers the traffic-electricity coupling characteristics; Step S2: Calculate the optimal travel and charging routes for electric vehicles based on the extended transportation network, and obtain the spatiotemporal load distribution of each charging station in each time period; The specific implementation of steps S1 and S2 is as described in the extended traffic network traffic allocation model in authorization announcement number CN120087727B. In this model, the main problem passes the real-time state of the traffic network to subproblems. The subproblems use a shortest path algorithm to find the optimal spatiotemporal path and return this optimal path as an alternative travel plan to the main problem. After iteration, the optimal traffic flow distribution of the ETAP-UE model is finally obtained. Specifically: The aforementioned method for coordinated regulation of electric vehicle charging based on spatiotemporal extended user balance and demand response, The main components of the proposed spatiotemporal extended transportation network model include the concept of the Expanded Transportation Network (ETN), the electric vehicle origin-destination (EOD) pair, and the structure of spatiotemporal travel paths.

[0025] The system's global optimum is achieved through an iterative framework: the main problem updates traffic distribution, while subproblems identify optimal travel options for electric vehicle users. Since the number of feasible paths is finite, the convergence of this iterative approach is guaranteed.

[0026] ETN is represented as a three-dimensional directed graph. ,in, and These represent the extended node set and the arc set, respectively. The node set includes real network nodes at different time steps and virtual charging nodes corresponding to slow charging activities. The arc set... There are two types of paths: 1) Temporal arcs: describe the transition of the same node (or virtual charging node) between different time periods; 2) Spatial arcs: represent the actual movement between different locations within a given time interval.

[0027] It is important to note that all virtual slow charging nodes are geographically identical to their corresponding physical nodes. Each expansion node is defined as ,in Represents physical or virtual nodes. Represents a discrete time interval. Arc Connecting nodes and , recorded as ,in, This indicates that the starting and ending points are at different locations. The set of road connections is denoted as... It includes node pairs connected by actual roads. .

[0028] Within the ETN framework, each electric vehicle traveler corresponds to an extended trip origin-destination pair. ,in, This represents the set of all EOD pairs. Unlike the traditional Traffic Assignment User Balance (TAP-UE) model, the EOD pairs based on ETN simultaneously include spatial, temporal, and energy dimensions. Each EOD pair can be represented as: ,in, and These represent the starting and ending points, respectively. Indicates the expected departure time. This represents the initial state of charge (SoC). Under range constraints, the set of feasible spatiotemporal arcs for an electric vehicle is denoted as... ,in This indicates a feasible expansion path. The corresponding spatial path index is... Its arc flow rate is expressed as .gather With variables Together, they defined all feasible travel paths that satisfy time and energy constraints.

[0029] The system's global optimum is achieved through an iterative framework: the master problem is responsible for updating the traffic distribution. Specifically, the master problem controls the overall distribution of electric vehicle traffic on feasible spatiotemporal paths, and its formal definition is as follows: in, For the current path, For a set of paths, EOD In the path The total travel cost is defined as: (6) The additional cost of the extended path is: (7) The user balance condition is: in, and They represent time respectively Roads during the time period

[0030] Electric vehicle traffic flow and charging stations The number of waiting vehicles; and These represent traffic congestion and service waiting time functions, respectively; parameters and These correspond to free-flow driving time and basic charging time, respectively. and These refer to the capacity of roads and charging stations, respectively. For EOD The travel demand; This is the time cost conversion factor; This represents the minimum generalized travel cost in equilibrium. , , These are binary indicators representing the path-node-time relationship.

[0031] The goal of the main problem is to achieve user balance, i.e., all chosen paths have equal and minimum generalized travel costs. The total system cost is... The subproblem is used to determine the minimum-cost spatiotemporal path for each EOD pair, and its mathematical form is: (10) The constraints are:

[0032] in, , Electricity prices for slow and fast charging stations; Represents a node ,time The electric vehicle's state of charge (SoC); These represent the charging amount; Indicates the maximum and minimum charging power ( (Indicates discharge via V2G). Choose variables for the path; It is a large constant (Big-M), used for relaxation; This is the minimum SoC threshold constraint.

[0033] The objective function minimizes the total generalized travel cost, including travel time, waiting time, and charging costs. The constraints describe the evolution and feasibility of the State of Charge (SoC) under fast and slow charging scenarios, while the time cost is defined by the last equation. To maintain computational stability, each travel period is limited to one hour. If the electric vehicle's journey exceeds this duration, the system decomposes the path into multiple EOD pairs, connected by virtual nodes to ensure time consistency.

[0034] Equation (12) assumes that departure and arrival occur within the same time interval. For longer journeys, the model decomposes the journey into consecutive segments to maintain the consistency of SoC evolution.

[0035] For details on the design and application of the spatiotemporal extended transportation network model, please refer to the relevant content of the authorization announcement number CN120087727B. This application will not elaborate further here.

[0036] Step S3: Construct a communication network model between electric vehicle charging stations, determine the total amount of information transmitted on each communication edge based on the communication network model, and then obtain the communication cost of all communication links; Set up a set of communication nodes With communication link set Based on the information interaction needs generated during the traffic-power coupling control process, the information transmission relationship between nodes is determined, the information transmission volume on each communication link is calculated, and the total communication cost of the communication network is obtained. In this embodiment, the information network is modeled as a communication network (CN). It is assumed that each controllable switch is configured with a communication switch (CS) for sending and receiving information. Each CS can be considered a node in the communication network, and the set of all communication nodes is denoted as [the set of nodes]. .

[0037] Considering that each communication node needs to both broadcast local operating status information and receive control signals from neighboring nodes or the upper-level power grid, the communication link between any two adjacent nodes must support bidirectional communication. The set of bidirectional edges in the communication network is denoted as […]. any one of the edges Represents a node With nodes There is a two-way communication link between them, and .

[0038] In the transportation-power coupled regulation process, information exchange mainly originates from load uploads, voltage feedback, and demand response signal transmission between charging stations and distribution network nodes. Therefore, the set of commodity pairs is defined as: (17) in, Indicates a collection of charging stations. Represents the set of nodes in the distribution network. For charging stations With distribution network nodes The mapping coefficients, when =1 indicates a charging station Access Node . Represents a pair of goods. As the source node for sending information, The destination node for receiving information. (Set) This represents all node pairs that actually exchange information during the transportation-electricity coupling operation.

[0039] Subsequently, in terms of communication topology, a depth-first search (DFS) algorithm is used for each product pair. Search for its shortest communication path. Let... Represents node pairs The set of communication edges contained in the corresponding shortest communication path. Based on the path information, path edge association variables can be defined: (18) in, Indicates from node Transmit to node The number of information bits, in bits. If the communication side Belongs to path If the edge carries 1 bit of information, then the edge carries no information flow for the item pair; otherwise, the edge does not carry information flow for the item pair.

[0040] Furthermore, The total amount of information transmitted on each communication edge is represented as: (19); in, Indicates communication edge At any moment The total number of bits transmitted cumulatively; Represents path-edge related variables; Indicates the product pair At any moment The number of source-destination transmitted bits.

[0041] Therefore, the overall communication cost of a communication network can be expressed as the sum of the amount of information transmitted on all communication links: (20); in, Indicates time The total communication cost of the communication network. This indicator is used to quantify the communication resource consumption caused by information interaction during the traffic-power coupled regulation process, and to provide a quantitative basis for subsequently incorporating communication costs into the reward function to achieve synergistic optimization of revenue and communication overhead.

[0042] To clarify the causal coupling mechanism between pricing actions and communication resource consumption, import pairs At any moment Number of transmitted bits Based on current pricing actions Pricing action compared to the previous moment The differences, as well as the changes in distribution network node load and voltage caused by the pricing action, are jointly determined, specifically as follows: (21); among which, These represent the basic number of bits per instance for price broadcasting, load reporting, and voltage feedback, respectively. This is a preset threshold for relative price change, for example, 5%, which is a dimensionless proportion. This indicates that the relative changes of each price component for each charging station and each charging type in the price action vector are calculated item by item. This is an indicator function that triggers a price broadcast only when the relative price change reaches the threshold; otherwise, this term is zero. express The load change quantification level after rounding up; express The quantization level of voltage change after rounding up; For information quantization resolution; Indicates charging station At any moment Load changes; Indicates distribution network node At any moment Voltage changes; and Each product pair The charging stations and distribution network nodes associated with the corresponding information interactions; for non-price broadcast product pairs, the price broadcast item is set to 0; for price broadcast product pairs, the same price broadcast to the same recipient is counted only once in its corresponding product pair, avoiding duplicate counting of the same price broadcast in all product pairs. Therefore, the pricing action... Decide The value of is then used to determine the cumulative number of bits on the communication side through formula (19). The total communication cost is determined by formula (20). In other words, to clarify the causal coupling mechanism between pricing actions and communication resource consumption, this embodiment further stipulates that: the commodity pair defined before the aforementioned formula (18) Source-destination transmission bit count In fact, it is a time variable, denoted as... At every moment The value of is dynamically determined by the pricing actions of the intelligent agent and the resulting changes in load and voltage, and consists of the following three parts. The following... and Each product pair The charging stations and power distribution network nodes associated with the corresponding information interactions.

[0043] (a) Price broadcast trigger: The agent at time Output pricing action ,in This indicates that the relative change of each price component in the price action vector for each charging station and each charging type is calculated item by item. Only if the relative change magnitude relative to the action at the previous moment satisfies... ( When a preset relative price change threshold (e.g., 5%, a dimensionless ratio) is reached, the agent broadcasts the new price to all connected users. The corresponding number of broadcast bits for a single price-broadcasting item pair is... Otherwise, the original price is maintained and no broadcast is triggered; this item is 0. This item only applies to price broadcast product pairs; for non-price broadcast product pairs, this item is 0. For price broadcast product pairs, the same price broadcast for the same recipient is only counted once in its corresponding product pair to avoid the same price broadcast being counted repeatedly in the product pair summation of formula (19). That is: (twenty two) (b) Load reporting item: The number of bits of load information reported by the charging station to the distribution network node is positively correlated with the quantification level of load change, that is: (twenty three) in, express The load change quantification level after rounding up, and .when When the load changes little, this item is 0; when the load changes little, this item is at a lower quantification level.

[0044] (c) Voltage feedback term: The number of bits in the voltage feedback is positively correlated with the quantization level of the node voltage change, i.e.: (twenty four) in, express The quantization level of the voltage change after rounding up, and ; These represent the basic number of bits per instance for price broadcasting, load reporting, and voltage feedback, respectively. For information quantification resolution.

[0045] In conclusion, At any moment The specific values ​​are: (25) According to the revised formula (19), the communication edge At any moment cumulative number of bits By each commodity pair Path-edge related variables Therefore, With pricing action And the change; and then through formula (20), the total communication cost Similarly, this is a pricing action. The function. The complete causal chain is expressed as: The arrows above correspond to the ETAP-UE model solution, the mapping of the charging station load to the distribution network node by formula (30) and the PDN power flow calculation, the generation of the number of transmission bits for commodity pairs, the calculation of the cumulative number of communication edges by formula (19), and the calculation of the total communication cost by formula (20).

[0046] That is, every pricing decision made by the intelligent agent Ultimately, by redistributing user behavior and power flow, the situation will change. This changes the actual bit capacity of each communication link. Total communication cost ,make Become about The real variables rather than background constants.

[0047] Therefore Weighted coefficient Add single-step rewards Then, as shown in the subsequent formula (38), the agent will face a real cost trade-off in policy learning: drastic or frequent price adjustments may increase energy trading revenue in the short term. However, this will simultaneously trigger price broadcasting, cause user path reallocation, and thus amplify load fluctuations and voltage changes, making... A significant increase; conversely, a steady price adjustment or a price adjustment at the threshold. Maintaining the original price may forgo some short-term profit opportunities, but it can significantly reduce communication costs. The maximum entropy reinforcement learning process based on the SoftActor-Critic (SAC) algorithm can guide the policy to tend to a coordinated control mode of "only triggering price changes when necessary and reporting state information on demand" by maximizing the cumulative reward, as shown in formula (41), thereby achieving a coordinated trade-off between maximizing revenue and consuming communication resources.

[0048] In this embodiment, based on the specific mechanism described above, the controllable adjustment of communication resource consumption by the agent's pricing strategy can be directly and clearly achieved; wherein These are all engineering parameters that can be flexibly selected based on the deployment conditions of the communication network. The data are obtained by quantizing load changes and voltage changes at corresponding resolutions, and those skilled in the art can set them conventionally according to the scale of the charging station network and the bandwidth of the communication link. For price broadcast, load reporting, and voltage feedback product pairs, corresponding components are counted according to their information types, and uncorresponding information components are set to 0; wherein, the same price broadcast by the same receiving object is only counted once in its corresponding price broadcast product pair, thereby avoiding the same information interaction being counted repeatedly.

[0049] Step S3: Establish a linear power flow model for the distribution network, and perform power flow calculations by combining photovoltaic power output and electric vehicle charging load to obtain node voltage and line power distribution.

[0050] In a preferred embodiment, the linearized power flow model is constructed based on the steady-state operating parameters of the radial distribution network, including: This represents a radial distribution network, as shown in the diagram. ,in, Indicates a collection of power distribution lines. Represents the set of busbars; The linearized power flow model, i.e., the PDN model, is constructed as follows: (26) (27); among which, For power distribution network Central route reactive power, For power distribution network Central route reactive power, For power distribution network Central route active power, For power distribution network Middle node The corresponding active power output of the photovoltaic unit, Indicates power distribution network The reactive power exchanged between the substation and the upper-level power grid; Indicates power distribution network The active power exchanged between the substation and the upper-level power grid; Represents a node The set of child nodes, i.e., all nodes that satisfy the path nodes A set; This represents the set of numbers for the power distribution network, i.e. Indicates the first A power distribution network. Indicates power distribution network Middle node Whether it is a reference bus, i.e. a balancing node. If yes, the value is 1; otherwise, the value is 0.

[0051] (28); in, Represents a node voltage amplitude, Represents a node voltage amplitude, express The resistance, Indicates the line The reactance; (29); in, For power distribution network Middle node The maximum value of the active power output corresponding to the photovoltaic unit; (30); in, Represents network Middle node The power factor corresponding to the photovoltaic unit; (31); in, Indicates parameters Follows probability distribution ; (32); in, Represents a node In time Voltage amplitude at that time Represents a node The minimum value of the voltage amplitude, Represents a node The maximum value of the voltage amplitude; (33) (34) Among them, and They represent the power distribution network. Central route At any moment The active and reactive power flow.

[0052] Step S4: Construct a demand response mechanism, set grid operation constraints and reward / penalty rules, and realize dynamic interaction between charging service operators and the grid.

[0053] In a preferred embodiment of this invention, the total active power load is represented as follows: (35); in, For power distribution network Middle node In time Total active power load at time For power distribution network Middle node In time The base load at that time This represents the load demand of fast charging stations obtained from the traffic distribution model of the extended transportation network. This represents the load demand of slow charging stations obtained from the traffic distribution model of the extended transportation network. To determine the charging station Is it connected to the power distribution network? nodes The corresponding mapping coefficients, when Time indicates charging station Connected to the power distribution network nodes , A collection of electric vehicle charging stations. Represents a set of time sections.

[0054] Based on the ETAP-UE and PDN models, a coordinated Demand Response (DR) framework is established to couple the operation of the transportation and power systems. The ETAP-UE model provides the spatiotemporal charging demand of electric vehicles, while the PDN model reflects power flow constraints. Its task is to determine the operating state of the entire system based on given operating conditions and network structure, such as voltage (amplitude and phase angle) on each bus, power distribution in the network, and power loss. By integrating these two layers, each electric vehicle charging system acts as an intelligent agent, capable of dynamically adjusting its pricing strategy based on DR signals and network status. This embodiment constructs a DR mechanism and models the decision-making process of the Electric Vehicle Charging System (EVCS), i.e., the electric vehicle charging stations connected to the Internet, as a Markov process (MDP) to learn the optimal joint pricing strategy.

[0055] Step S5: Treat the charging station as an intelligent agent, establish a pricing optimization model based on Markov decision process, and use deep reinforcement learning methods to obtain the optimal pricing strategy for fast and slow charging.

[0056] Step S6: Through a three-layer feedback loop of transportation, electricity, and price, the system converges until it achieves a balance between user travel, safe operation of the power grid, and optimal coordination of revenue.

[0057] In a preferred embodiment of this example, under a regulatory framework, the power grid sets upper and lower limits for charging station operators (CSOs) based on their charging demand (DR). When total charging demand exceeds the upper limit, a penalty is imposed; when demand remains within the specified range, the CSO receives a reward; and when demand falls below the lower limit, no compensation is provided. The amount of the reward or penalty is determined by the DR pricing mechanism set by the power grid. Upon receiving these regulatory signals, the CSO dynamically adjusts charging prices based on the changing trends of electric vehicle travel demand and charging behavior. It should be noted that this embodiment does not explicitly model the internal decision-making mechanism of the power grid operator, but rather treats DR-related parameters as external random inputs. Under this setting, the CSO's goal is to maximize its own revenue while satisfying power grid constraints. Subsequently, under new price signals, EV users will re-plan their travel and charging schedules based on their own travel arrangements, charging station accessibility, and current traffic conditions. This hierarchical and sequential interaction process accurately depicts the dynamic decision-making behavior of the power grid, CSO, and EV users under DR-based regulation.

[0058] Each EVCS operates as an independent intelligent agent, simultaneously managing both fast and slow charging facilities. Charging stations must collaboratively determine the optimal pricing strategy for both charging methods to participate in the DR (Radio Demand) program. The electricity purchase cost, determined by the distribution network operator, serves as the input, while the retail charging price for EV users is the decision variable. After obtaining the optimal pricing, the EVCS broadcasts the price to all connected users. The EVCS uses environmental information such as historical load, electricity price records, and current DR parameters to determine the charging price that maximizes expected profits.

[0059] In the DR framework proposed in this embodiment, each EVCS is modeled as an autonomous agent following a Markov decision process (MDP). The agent's decisions depend on the current observable state, which contains sufficient information to predict future transitions, thus satisfying the Markov property assumption. The MDP consists of the following: 1) State space: At time step Electric vehicle charging station The state vector integrates local operating variables, fast and slow charging demands from the ETAP-UE model, and the grid-side power state. The state vector is defined as: (36); in, This represents the total charging load at the previous moment. and To meet the current needs of fast charging and slow charging, and The prices for fast charging and slow charging at the previous moment, The electricity purchase price at the previous moment. and DR threshold This is the current DR signal. and These represent the voltage and active power flow of the connected bus, respectively. This state design enables the agent to simultaneously perceive the traffic and power network states, thereby capturing their spatiotemporal coupling characteristics. This represents the total number of charging stations included in a charging system.

[0060] 2) Motion space: EVCS The action corresponds to the time. Price comparison between fast charging and slow charging: (37) The agent's policy network outputs the aforementioned price to achieve an optimal balance between revenue and DR participation.

[0061] 3) Reward function: The reward function in the Markov decision process, based on the energy trading revenue and demand response reward, further introduces the total communication cost of the communication network to achieve a synergistic trade-off between revenue and communication resource consumption.

[0062] The single-step reward function at time... Defined as: (38) (39) (40)

[0063] in, Indicates charging station At any moment Energy trading revenue; Indicates charging station At any moment Rewards or penalties obtained based on demand response mechanisms; Indicates time The total communication cost of a communication network; This is a communication cost weighting coefficient used to adjust the degree of trade-off between revenue and communication resource consumption.

[0064] By maximizing the cumulative reward in equation (33), as described in equation (36), the EVCS agent can learn an adaptive joint pricing strategy, thereby maximizing its own profit while ensuring network stability and meeting the system-level DR objective.

[0065] (41) In a preferred embodiment of this invention, the energy trading revenue for fast charging and slow charging are respectively: (42) in, and Charging stations At any moment The prices for fast charging and slow charging are set; and For the corresponding fast and slow charging loads from the ETAP-UE model; and These represent the electricity purchase price and the reward coefficient for demand response, respectively; wherein, in a preferred embodiment, the reward or penalty based on DR is expressed as: (43) in, for or , and Charging stations At any moment Fast charging and slow charging are based on demand-responsive rewards or penalties. It is a binary variable, representing the charging station. Is it connected to the distribution network node? , for , This is the demand response reward coefficient. Indicates charging station At any moment The threshold power for demand response of type fast charging or slow charging; Indicates charging station At any moment Threshold power for demand response of type fast charging or slow charging; when hour, and These are the upper and lower thresholds for fast charging power, respectively; when hour, and These are the upper and lower thresholds for slow charging power, respectively; where, Together, they constitute the demand response power range for either fast charging or slow charging.

[0066] In a preferred embodiment of this invention, the decision variables are subject to the following boundary constraints: (44) in, DR load thresholds for each mode; binary variables Indicates charging station Is it connected to the distribution network node? , This refers to the charging load from the ETAP-UE model, specifically the fast or slow charging load. Therefore, the total charging load coupled with the distribution network is expressed as... This is to ensure consistency between systems.

[0067] In electric vehicle charging networks, the multi-level marketing (MDP) modeling of the charging and discharging (DR) scheduling problem involves both fast and slow charging facilities. Due to the existence of multiple types of charging resources, the system has a large and continuous action space, requiring coordination of charging and discharging power levels across multiple charging stations and time periods. This complexity demands that reinforcement learning (RL) algorithms possess strong exploratory capabilities. Traditional actor-critic methods are typically highly sensitive to neural network architecture and hyperparameter selection. For example, in the DDPG algorithm, inappropriate learning rate or exploration strategy settings can severely impact convergence, leading to unstable or suboptimal results.

[0068] To alleviate the aforementioned problems, this embodiment employs the Soft Actor-Critic (SAC) algorithm based on the maximum entropy reinforcement learning principle to construct a demand response optimization framework. By maximizing policy entropy, SAC encourages stochastic exploration and avoids premature convergence, making it particularly suitable for continuous high-dimensional control spaces in electric vehicle charging coordination.

[0069] In this modeling, and These represent the system state and actions, respectively. For instant rewards, The maximum number of time steps within the decision-making timeframe. As a discount factor, This is the experience replay buffer. For the policy function, Represents policy entropy. It represents the temperature coefficient, used to balance exploration and utilization. and These represent the action-value function and the state-value function, respectively.

[0070] This is a normalization constant to ensure the probability distribution. The integral is 1.

[0071] The goal of SAC is to maximize the weighted combination of expected return and policy entropy: (45) In a demand response scenario, state This displays real-time system information, such as total load, available charging capacity, electricity price, and charging station occupancy rate; actions. Indicates the allocation of charging and discharging power between fast charging and slow charging stations; Rewards It reflects economic and operational performance, including energy cost reduction, peak shaving and valley filling, and user satisfaction.

[0072] The action value function and the state value function are defined as follows: Therefore, the optimal demand response strategy can be expressed as: And it can be approximated by minimizing the KL divergence: (49) in, For strategy The action value function below, This is the normalization factor for the probability distribution. Both the policy network and the Q network are parameterized using deep neural networks. To mitigate the overestimation bias of action values, SAC introduces a double Q-learning mechanism, employing two Q networks and their corresponding target networks, denoted as […]. and its target network The Q-network loss function is defined as: in, and These represent the online Q-network and the target Q-network, respectively. During training, the smaller value of their outputs is taken to reduce estimation bias. Parameters Updated via gradient descent, while the target parameter... A soft update mechanism is adopted.

[0073] Strategy parameters By maximizing the desired state value function The optimization, in gradient form, is as follows: In addition, the algorithm integrates an adaptive entropy adjustment mechanism, which can automatically adjust the temperature coefficient. The value of allows the agent to dynamically control the exploration intensity based on the system's learning state.

[0074] This framework enables SAC agents to jointly optimize fast and slow charging behaviors under a unified stochastic demand response model, achieving a dual improvement in system-level operational efficiency and user-level flexibility.

[0075] In this example, to verify the effectiveness of the proposed electric vehicle charging coordinated control method based on spatiotemporally extended user balance and demand response, a comprehensive simulation environment is constructed, including a traffic network, a radial distribution network, and a charging station communication structure. On the traffic side, a spatiotemporally extended traffic network model is used to model the origin-destination pairs of electric vehicle trips. Based on the user balance principle, the optimal travel and charging paths for each time period are solved, obtaining the fast-charging and slow-charging load demand distributions of each charging station at different time sections. On the power side, a linearized power flow model is used to model the radial distribution network, mapping the charging load output from the traffic model to the corresponding distribution network nodes. This load, along with the base load and photovoltaic output, participates in the power flow calculation, thereby obtaining the node voltage and line power distribution. Subsequently, in this coupled environment, each charging station is modeled as a Markov decision process agent, and a joint pricing strategy is trained using the SAC algorithm based on the maximum entropy reinforcement learning principle.

[0076] like Figure 2 As shown in the figure, this example sets up two comparative scenarios. In the first scenario, the reward function only considers the energy trading revenue of the charging station, i.e., charging revenue minus electricity purchase cost, without introducing demand response reward or penalty mechanism. In the second scenario, a demand response threshold constraint and corresponding reward and penalty terms are further introduced on the basis of energy trading revenue, while comprehensively considering the information interaction cost in the communication network. Both scenarios are trained multiple times under the same traffic demand, the same photovoltaic power output distribution, and the same reinforcement learning parameters to compare the policy convergence characteristics and long-term revenue level.

[0077] As can be seen from the reinforcement learning convergence curve, without the introduction of a demand response mechanism, the agent quickly obtains high positive returns in the early stages of training. This is because its strategy tends to continuously increase charging prices and expand charging power during periods of high demand to maximize immediate profits. However, as training progresses, the electric vehicle charging load is concentrated at a few charging stations during peak hours, leading to voltage drops at corresponding distribution network nodes and some lines approaching their capacity limits. Although the model does not directly impose hard constraints on voltage exceeding limits, the increased travel costs and waiting times on the transportation side cause changes in user route choices and exacerbate load fluctuations, ultimately resulting in significant fluctuations and a gradual decline in returns. The average cumulative reward during the convergence phase is close to zero or even slightly negative, with a large fluctuation range, indicating that in the absence of demand response constraints, simply pursuing profit maximization is unlikely to form a stable strategy in a transportation-electricity coupled system.

[0078] In scenarios incorporating demand response mechanisms, the reward function includes reward and penalty terms based on load ranges. When the total load of a charging station exceeds an upper threshold, an economic penalty is imposed; when the load remains within a reasonable range, an additional reward is awarded; and when it falls below a lower threshold, no compensation is provided. Initially, due to frequent penalties triggered by exploratory behavior, the agent's cumulative reward rapidly declines into negative territory during the early training phase. However, as training progresses, the strategy gradually learns to adjust fast and slow charging prices, guiding the redistribution of some electric vehicles across different time periods or charging stations, thus achieving peak shaving and valley filling. Power flow calculations show that the range of node voltage fluctuations significantly converges, line load rates decrease, and the system operates more smoothly. During the convergence phase, although the absolute value of the average cumulative reward is lower than the initial level in scenarios without demand response, the curve fluctuation amplitude is significantly reduced, demonstrating better stability and repeatability.

[0079] Further comparison of the operational results in the two scenarios reveals that the introduction of the demand response mechanism significantly reduces the peak-to-valley difference in the system, lowers the maximum voltage deviation of the distribution network, and reduces the risk of line overload. Simultaneously, although individual charging station profits are somewhat constrained, long-term cumulative revenue is more stable, avoiding revenue losses due to system oscillations. Furthermore, by incorporating communication cost weights into the reward function, the agent avoids frequent and significant price adjustments during the decision-making process, reducing the burden of information interaction and improving overall operational efficiency.

[0080] Regarding communication costs, comparing the total amount of information exchange before and after control can further verify the effectiveness of the method of this invention. Simulation statistics show that, without the introduction of a collaborative control mechanism, the total amount of communication data generated by the system during training and operation is approximately 1.96 × 10⁻⁶. 9 However, after introducing a collaborative pricing strategy based on demand response and reinforcement learning, the total communication volume was reduced to approximately 1.68 × 10^6 bits; 7 Compared to the former, the scale of communication data has decreased by more than two orders of magnitude, a reduction of over 99%.

[0081] Furthermore, the reduction of one or two orders of magnitude or more objectively verifies the improvement in communication costs. This is indeed a pricing action. A variable function rather than a constant: under different strategies If the values ​​are the same, such a difference is impossible. The experimental results are completely consistent with the mechanism of "price threshold triggering broadcast + load / voltage deviation quantification reporting" mentioned above in this specification. See formulas (18)-(20) and their causal chain descriptions to prove that the reward function design proposed in this application can truly guide the agent to make a cooperative trade-off between benefits and communication resource consumption during the policy learning process.

[0082] The results show that, without coordinated control, charging stations frequently adjust prices and continuously report their operating status in pursuit of local optimal returns. This leads to a significant increase in the number of load uploads, voltage feedbacks, and demand response signal transmissions between charging stations and distribution network nodes, resulting in a large amount of redundant communication traffic. However, in the transportation-power-communication multi-network coupling framework proposed in this embodiment, the reinforcement learning agent achieves a synergistic trade-off between revenue and information interaction overhead by introducing communication cost weights into the reward function. This makes price adjustments more stable and load changes more orderly, thereby significantly reducing the total amount of data transmission on the communication links.

[0083] Therefore, this embodiment can not only improve the operational stability of transportation and power systems, but also effectively reduce the occupation of information network resources, improve the scalability and economy of the system in large-scale charging scenarios, and further demonstrate the technical advantages of incorporating communication costs into the optimization objectives.

[0084] like Figure 4 The present invention also provides an electric vehicle charging coordinated control system, the system comprising: The spatiotemporal load distribution determination module is used to calculate the optimal travel and charging routes for electric vehicles based on the traffic allocation model of the extended transportation network, thereby obtaining the spatiotemporal load distribution of each charging station in different time periods. The communication network construction module is used to construct a communication network model between electric vehicle charging stations, determine the total amount of information transmitted on each communication edge based on the communication network model, and then obtain the communication cost of all communication links. The decision-making module models the decision-making of the electric vehicle charging system as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. During the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; a reward is given when the total charging demand remains within a specified range; and no compensation is given when the total charging demand falls below a set lower limit. The reward function is configured to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on a radial distribution network, thus coupling the operation of the transportation system and the power system.

[0085] The other technical features of the electric vehicle charging coordinated control system based on spatiotemporal extended user balance and demand response described in this invention are similar to those of the corresponding electric vehicle charging coordinated control method based on spatiotemporal extended user balance and demand response, and will not be repeated here.

[0086] This embodiment also provides a computer system, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the electric vehicle charging coordinated control method described above.

[0087] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0088] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0089] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for coordinated control of electric vehicle charging, characterized in that, The method includes: The optimal travel and charging routes for electric vehicles are calculated based on the traffic distribution model of the extended transportation network, thereby obtaining the spatiotemporal load distribution of each charging station at different times. A communication network model between electric vehicle charging stations is constructed. Based on the communication network model, the total amount of information transmitted on each communication edge is determined, and then the communication cost of all communication links is obtained. The decision-making process of the electric vehicle charging system is modeled as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. During the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; when the total charging demand remains within a specified range, the charging service operator receives a reward. The reward function is configured to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on a radial distribution network, thus coupling the operation of the transportation system and the power system.

2. The electric vehicle charging coordinated control method according to claim 1, characterized in that, The construction of the communication network model between electric vehicle charging stations includes: The information network between electric vehicle charging stations is modeled as a communication network CN. It is assumed that each controllable switch is equipped with a communication exchange CS for sending and receiving information. Each CS is considered a node in the communication network, and the set of all communication nodes is denoted as . ; Considering that each communication node needs to both broadcast local operating status information and receive control signals from neighboring nodes or the upper-level power grid, the communication link between any two adjacent nodes must support bidirectional communication. The set of bidirectional edges in the communication network is denoted as […]. any one of the edges Represents a node With nodes There is a two-way communication link between them, therefore, ; Since information exchange originates from load transmission, voltage feedback, and demand response signal transmission between charging stations and distribution network nodes, the set of commodity pairs is defined as follows: ; in, Indicates a collection of charging stations. Represents the set of nodes in the distribution network. For charging stations With distribution network nodes The mapping coefficients, when =1 indicates a charging station Access Node , Represents a pair of goods. As the source node for sending information, For the destination node of information reception, a set This represents all node pairs that actually exchange information during the coupled operation of transportation and power systems; In terms of communication topology, a depth-first search (DFS) algorithm is used for each product pair. Search for its shortest communication path, let... Indicates the product pair The set of communication edges contained in the corresponding shortest communication path, and path edge association variables defined based on path information: ; Among them, if the communication side Belongs to path If the edge carries 1 bit of information, then the edge carries no information flow for the item pair; otherwise, the edge does not carry information flow for the item pair.

3. The electric vehicle charging coordinated control method according to claim 2, characterized in that, The step of determining the total amount of information transmitted on each communication edge based on the communication network model, and then obtaining the communication cost of all communication links, includes: The total amount of information transmitted on each communication edge is represented as: ; in, Indicates communication edge At any moment The total number of bits transmitted cumulatively; Indicates the product pair At any moment The number of bits transmitted; Therefore, the overall communication cost of a communication network can be expressed as the sum of the amount of information transmitted on all communication links: ; in, Indicates time The total communication cost of a communication network, an indicator used to quantify the communication resource consumption resulting from information exchange during the coupling and control of transportation and power; commodity pairs At any moment Number of transmitted bits Based on current pricing actions Pricing action compared to the previous moment The differences, as well as the changes in distribution network node load and voltage caused by the pricing action, are jointly determined, specifically as follows: ;in, These represent the basic number of bits per instance for price broadcasting, load reporting, and voltage feedback, respectively. The preset relative price change threshold, This indicates that the relative changes of each price component for each charging station and each charging type in the price action vector are calculated item by item. This is an indicator function that triggers a price broadcast only when the relative price change reaches the threshold; otherwise, this term is zero. express The load change quantification level after rounding up; express The quantization level of voltage change after rounding up; For information quantization resolution; Indicates charging station At any moment Load changes; Indicates distribution network node At any moment Voltage changes; and Each product pair The charging stations and distribution network nodes associated with the corresponding information interaction; for non-price broadcast product pairs, the price broadcast item is 0; for price broadcast product pairs, the same price broadcast for the same recipient is only counted once in its corresponding product pair to avoid the same price broadcast being counted repeatedly in all product pairs.

4. The electric vehicle charging coordinated control method according to claim 3, characterized in that, The method of dynamically adjusting the pricing strategy based on the spatiotemporal load distribution of the charging station, treating each electric vehicle charging system as an intelligent agent, includes: The agent's decisions depend on the current observable state, which contains enough information to predict future transitions, thus satisfying the Markov property assumption. The Markov decision process consists of: State space: at time step Electric vehicle charging station The state vector integrates local operating variables, fast and slow charging demand from the traffic allocation model of the extended transportation network, and grid-side power status. The state vector is defined as follows: ; in, This represents the total charging load at the previous moment. and To meet the current needs of fast charging and slow charging, and The prices for fast charging and slow charging at the previous moment, The electricity purchase price at the previous moment. Electric vehicle charging stations At any moment The upper limit of the demand response load threshold is set in fast charging mode. Electric vehicle charging stations At any moment The lower limit of the demand response load threshold is set in fast charging mode. Electric vehicle charging stations At any moment The upper limit of the demand response load threshold is set in slow charging mode. Electric vehicle charging stations At any moment The lower limit of the demand response load threshold is set in slow charging mode. As a current demand response signal, express Always connected to the power distribution network nodes voltage, express Always connected to the power distribution network The lines in The corresponding active current, This refers to the total number of electric vehicle charging stations in an electric vehicle charging system. Total number of time steps; Motion space: Electric vehicle charging station The action corresponds to the time. The price pair for fast charging and slow charging is represented as follows: ； in, and Charging stations At any moment The prices for fast charging and slow charging are set; therefore, the agent's policy network outputs the prices for fast charging and slow charging to achieve an optimal balance between revenue and demand response participation.

5. The electric vehicle charging coordinated control method according to claim 4, characterized in that, The reward function is configured to take into account energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links, including: The single-step reward function at time 1 Defined as: in, Indicates charging station At any moment Energy trading revenue; Indicates charging station At any moment Rewards or penalties obtained based on demand response mechanisms; Indicates time The total communication cost of a communication network; The communication cost weighting coefficient is used to adjust the trade-off between revenue and communication resource consumption. Therefore, by maximizing the cumulative reward function, the agent can learn an adaptive joint pricing strategy, thereby maximizing its own profit while ensuring network stability and meeting system-level demand response objectives.

6. The electric vehicle charging coordinated control method according to claim 5, characterized in that, The energy trading revenue is expressed as follows: ; in, Electric vehicle charging stations At any moment The revenue from fast charging energy trading Electric vehicle charging stations At any moment The revenue from slow-charging energy trading and These represent the purchase price of electricity and the incentive coefficient for demand response, respectively. and For fast and slow charging loads from the traffic distribution model of the extended traffic network; The rewards or penalties obtained based on the demand response mechanism are represented as follows: ; Among them, binary variables Indicates electric vehicle charging station Is it connected to the distribution network node? , , The upper and lower limits of the demand response load threshold for each mode are defined. , For fast charging mode, Slow charging mode For fast or slow charging loads from traffic allocation models of extended transportation networks.

7. The electric vehicle charging coordinated control method according to claim 1, characterized in that, The steady-state operating parameters of the radial distribution network are achieved by constructing a linearized power flow model, including: This represents a radial distribution network, as shown in the diagram. ,in, Indicates a collection of power distribution lines. Represents the set of busbars; A linear power flow model is constructed, described as follows: ;in, Indicates power distribution network Middle node Is it a reference bus, i.e., a balancing node? If yes, the value is 1; otherwise, the value is 0. Indicates power distribution network The reactive power exchanged between the substation and the upper-level power grid; For power distribution network Central route reactive power, For power distribution network Middle node Corresponding to the reactive power output of the photovoltaic unit, For power distribution network Central route reactive power, For power distribution network Middle node Total reactive load, Indicates the terminal node; Represents a node The set of child nodes, i.e., all nodes that satisfy the path nodes A set; This represents the set of numbers for the power distribution network, i.e. Indicates the first A power distribution network; Indicates power distribution network The active power exchanged between the substation and the upper-level power grid; For power distribution network Central route active power, For power distribution network Middle node The corresponding active power output of the photovoltaic unit, For power distribution network Middle node Total active load; ; in, Indicates the first Nodes in a power distribution network voltage amplitude, Represents a node voltage amplitude, Indicates the first Lines in a power distribution network The resistance, Indicates the first Lines in a power distribution network The reactance; ; in, For power distribution network Middle node The maximum value of the active power output corresponding to the photovoltaic unit; ; in, Represents network Middle node The power factor corresponding to the photovoltaic unit; ; in, Indicates parameters Follows probability distribution ; ; in, Represents a node In time Voltage amplitude at that time Represents a node The minimum value of the voltage amplitude, Represents a node The maximum value of the voltage amplitude; in, and They represent the power distribution network. Central route At any moment The active and reactive power flow.

8. The electric vehicle charging coordinated control method according to claim 7, characterized in that, The demand response mechanism is obtained by integrating the flow allocation model of the extended transportation network and the linearized power flow model based on the radial distribution network, including: The total active power load of the distribution network is expressed as: ; in, For power distribution network Middle node In time Total active power load at time For power distribution network Middle node In time The base load at that time This represents the load demand of fast charging stations obtained from the traffic distribution model of the extended transportation network. This represents the load demand of slow charging stations obtained from the traffic distribution model of the extended transportation network. To determine the charging station Is it connected to the power distribution network? nodes The corresponding mapping coefficients, when Time indicates charging station Connected to the power distribution network nodes , A collection of electric vehicle charging stations. Represents a set of time sections; This integrates the traffic distribution model and the linearized flow model of the extended transportation network to form a coordinated demand response framework, thereby treating each electric vehicle charging station as an intelligent agent that can dynamically adjust its pricing strategy based on demand impact signals.

9. A coordinated control system for electric vehicle charging, characterized in that, The system includes: The spatiotemporal load distribution determination module is used to calculate the optimal travel and charging routes for electric vehicles based on the traffic allocation model of the extended transportation network, thereby obtaining the spatiotemporal load distribution of each charging station in different time periods. The communication network construction module is used to construct a communication network model between electric vehicle charging stations, determine the total amount of information transmitted on each communication edge based on the communication network model, and then obtain the communication cost of all communication links. The decision-making module models the decision-making of the electric vehicle charging system as a Markov decision process to learn the optimal joint pricing strategy. Each electric vehicle charging system acts as an agent, dynamically adjusting its pricing strategy based on the spatiotemporal load distribution of the charging stations. After obtaining the optimal price, the electric vehicle charging system broadcasts the price to all connected users. During the Markov decision process, a penalty is incurred when the total charging demand exceeds a set upper limit; when the total charging demand remains within a specified range, the charging service operator receives a reward. The reward function is configured to consider energy trading revenue, rewards or penalties obtained based on the demand response mechanism, and the communication costs of all communication links. The demand response mechanism is obtained by integrating the traffic allocation model of the extended transportation network and the linearized power flow model based on a radial distribution network, thus coupling the operation of the transportation system and the power system.

10. A computer system comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to perform the steps of the method as described in any one of claims 1-8.

Images