Cooperative scheduling method and system for communication, electric power and traffic coupling network resources

Abstract

The invention relates to a communication, electric power and traffic coupling network resource cooperative scheduling method and system, and belongs to the technical field of telecommunication. The method comprises the following steps: firstly, establishing a communication-electric power-traffic coupling network model, fully considering EV user heterogeneity, and constructing a demand response model considering EV user finite rationality; furthermore, power grid load characteristics are analyzed from two dimensions of time and space, EV demand response time and space guide electricity price are constructed on the basis, and a space-time guide electricity price model is provided. Secondly, an EV demand response scheduling algorithm based on space-time electricity price guidance and migration collaboration is provided, and comprehensive optimization of a communication-electric power-traffic coupling network is realized through space-time electricity price guidance and migration collaboration of demand response strategy knowledge between EVs.

Description

Technical Field

[0001] The present invention relates to a method and system for collaborative scheduling of communication, power, and transportation coupled network resources, and belongs to the field of electric communication technology. Background Art

[0002] Currently, with the continuous innovative development of technologies such as autonomous driving and electrified transportation, the connection between communication, power, and transportation systems has become increasingly close. Structurally, the power grid and the communication network are coupled through data interaction of base stations, the transportation network and the power grid are coupled through load scheduling of fast charging stations (FCS), and the communication network and the transportation network are coupled through task data transmission and path selection of electric vehicles (EV). The charging stations in the transportation network, the base stations on the communication network side, and the edge servers are deployed along traffic lines to provide real-time information support for EVs; the power grid provides power supply for the charging stations arranged along traffic lines, enabling the three networks to achieve deep resource coupling in the physical structure. Functionally, the transportation network combines the FCS model to simulate the path selection and charging station selection behaviors of EVs, optimize travel paths and charging strategies, and improve traffic efficiency; the power grid quantifies the impact of EV participation in demand response from two dimensions of time and space through a distribution network model including FCS, analyzes its optimization effect on the power flow, network loss, and energy efficiency of the distribution network using load stability and load balance models, and relies on the communication network to obtain EV data in the transportation network, thereby optimizing power distribution and achieving balanced and stable load; the communication network relies on 4G / 5G / 6G technologies to build base stations and edge servers to provide low-latency and high-efficiency communication services for EVs, support two-way information interaction between FCS and the power grid, and promote the efficient interconnection and interoperability of EVs with the power grid and the transportation network.

[0003] However, the charging and discharging and travel characteristics of large-scale EVs will cause fluctuations in the power grid load and traffic flow in the transportation network in terms of time, and will cause uneven distribution of the power grid load and road congestion in the transportation network in terms of space, seriously affecting the safe and stable operation of the power-transportation coupled network. Therefore, adopting effective demand response strategies to provide charging and discharging guidance and travel path planning for EVs in both time and space dimensions can effectively reduce the adverse effects brought by the charging and discharging and travel of large-scale EVs on the communication-power-transportation coupled network, and ensure the safe and stable operation of the communication-power-transportation coupled network.

[0004] The implementation principles of traditional EV demand response methods are all to optimize the operation of the power grid and meet the needs of the transportation network by managing the charging and discharging behaviors of EVs. For example, the existing patent "EV Charging Scheduling Method under Smart Grid" proposes a demand response method based on dynamic electricity prices; the existing patent "EV Charging and Discharging Optimization Method in Urban Transportation Network" proposes an EV charging and discharging optimization method considering the influence of traffic flow.

[0005] However, the existing methods still have the following problems: First, when considering EV demand response, the existing methods ignore the coupling of the communication-power-transportation network. Moreover, the existing EV demand response-guided electricity price only considers the time-dimensional distribution characteristics of the grid load, guiding EVs in the time dimension and ignoring the spatial-dimensional distribution characteristics of the grid load as well as the electricity price guidance for EVs in both the spatial and temporal dimensions, resulting in poor stability and balance of the grid load. Second, the existing methods ignore the migration and cooperation of demand response strategy knowledge among EVs in the communication-power-transportation coupling network. When the environment where EVs are located changes, a large amount of interaction with the environment needs to be carried out again, failing to make full use of the knowledge learned by other EVs, generating a large amount of repeated learning, reducing the overall learning efficiency, and resulting in a poor final demand response effect. Therefore, there is an urgent need to design a method and system for collaborative scheduling of communication, power, and transportation coupling network resources.

[0006] In view of the above defects, the present invention aims to create a method and system for collaborative scheduling of communication, power, and transportation coupling network resources, making it more valuable for industrial utilization. Summary of the Invention

[0007] To solve the above technical problems, the object of the present invention is to provide a method and system for collaborative scheduling of communication, power, and transportation coupling network resources.

[0008] A method for collaborative scheduling of communication, power, and transportation coupling network resources according to the present invention specifically includes the following steps:

[0009] First, construct a communication-power-transportation coupling network model including a communication network-power-transportation coupling network architecture, a transportation network model with FCS, a distribution network model with FCS, and a communication network model, and fully consider the heterogeneity of EV users to construct an EV bounded rationality demand response model;

[0010] Analyze the grid load characteristics from both time and space dimensions, and on this basis, construct an EV demand response time and space-guided electricity price, and propose a spatio-temporal guided electricity price model;

[0011] By optimizing the EV travel path, charging and discharging nodes, and charging and discharging power, minimize the weighted sum of the EV road travel time, demand response revenue, travel anxiety, grid load stability, balance degree, and network loss to construct an objective function for the optimization problem;

[0012] Based on the above models and the objective function of the optimization problem, model the EV demand response as a partially observable Markov cooperative game model, and solve the partially observable Markov cooperative game model based on the spatio-temporal electricity price guidance and migration and cooperation-based EV demand response scheduling algorithm to achieve the comprehensive optimization of the communication-power-transportation coupling network.

[0013] Furthermore, the specific architecture of the communication network - power - transportation coupling network is as follows: It is defined that there are a total of M EVs and N edge servers; the sets of EVs and edge servers are denoted as and The total optimization duration is divided into continuous T, indexed by t = 1, 2, …, T; the duration of one time period is denoted as τ; the route selection variable of the EV is defined as μ m,a (t), where μ m,a (t) = 1 indicates that EV m selects route a to travel in the t-th time period; the FCS selection variable of the EV is defined as and indicates that EV m selects FCS q for charging, indicates that EVm selects FCS q for discharging; the decision variable for autonomous driving data offloading of the EV is defined as y m,n (t), y m,n (t) = 1 indicates that EV m offloads its autonomous driving data to edge server n for processing through the base station closest to it.

[0014] Furthermore, the transportation network model with FCS is defined as an undirected graph G T =(N T , A T , Q), where N T represents the set of road network nodes; A T represents the set of road segments, and road segment a ∈ A T ; Q represents the set of FCSs in the road network, and FCS q ∈ Q;

[0015] The semi - dynamic traffic flow assignment model is specifically expressed as:

[0016]

[0017] In the formula, Ω od is the set of EV travel demands; K od is the set of paths that satisfy travel demand od; is the traffic flow of travel demand od on path k; δ k,a (t) is a binary variable characterizing the relationship between the path and the road segment, and δ k,a (t) = 1 indicates that road segment a is on path k, and conversely, δ k,a (t) = 0; is the residual flow of demand od, τ is the time period length, r od , r od,mod are the original travel demand flow and the travel demand flow after considering the residual flow correction respectively; τ a (t), and They are the passing time of section a and the charging and discharging queuing time of FCS q respectively.

[0018] Furthermore, the distribution network model containing FCS is represented as an undirected graph G P =(N P , A P , Q), where N P represents the set of distribution network nodes; A P represents the set of distribution network line branches, and branch ij ∈ A P ;

[0019] The distribution network power flow model is specifically represented as:

[0020]

[0021] In the formula, P i,j (t) and Q i,j (t) are the active power and reactive power flowing from node i to node j on branch ij respectively, X i,j is the reactance of branch ij, v i (t) is the voltage of node i;

[0022] The distribution network power loss C DN,loss (t) is expressed as:

[0023]

[0024] In the formula, C loss (t) is the unit cost of power loss, R i,j is the resistance of branch ij, I i,j (t) is the current of branch ij.

[0025] Furthermore, the communication network model is specifically: the total power consumption of the base station is expressed by the following formula (8), including data processing power consumption and refrigeration power consumption

[0026]

[0027] Among them, the data processing power consumption of the edge server is expressed as:

[0028]

[0029] In the formula, P peak , P ideal are the peak power and standby power of a single server of the edge server respectively, κ m is the average data service rate of a single server, which is related to the operating frequency of the server CPU, N server is the number of data center front-end processors working, H n (t) is the task data processed by the data center;

[0030] The operating power of the edge server refrigeration equipment is expressed as:

[0031]

[0032] In the formula, is the air supply temperature of the refrigeration equipment, ξ n,2 , ξ n,1 and ξ n,0 are the refrigeration coefficients respectively;

[0033] The queuing delay of the task queue on the EV side of the communication network is:

[0034]

[0035] In the formula, Q m (t) represents the backlog of the task queue of EV m, K m (t), α m represent the number of tasks and the size of a single task respectively;

[0036] The queuing delay of the task queue on the edge server side is:

[0037]

[0038] Furthermore, the specific EV bounded rationality demand response model is as follows: The costs mainly considered in EV demand response include four aspects: the demand response cost C m,1 (t) on the power grid side; the travel time cost C m,2 (t) and travel anxiety cost C m,3 (t) on the traffic network side; the service data processing time cost C m,4 (t) on the communication network side; which are respectively expressed as:

[0039]

[0040] In the formula, μ m,a (t) is the road segment selection indication variable of EV m, μ m,a (t) = 1 indicates that EV m has selected road segment a; are the charge and discharge indication variables of EV m in FCS q respectively, indicates charging, indicates discharging; are the spatio-temporal guiding charge and discharge prices respectively; are the travel time anxiety coefficient and SoC anxiety coefficient respectively; are the expected arrival time and expected SoC respectively; is the indication variable of whether the destination has been reached, indicates that the destination has not been reached;

[0041] Based on the positive and negative expectations of EV as reference points respectively, the positive and negative prospect values of EV are

[0042]

[0043] In the formula, C m,z (t) (z = 1, 2, 3, 4) respectively represent the demand response cost, travel time cost, travel anxiety cost and data processing time cost of EV users; v + [C m,z (t)], v - [C m,z (t)] are respectively the positive and negative prospect values of EV based on its positive and negative expectations as the benchmark;

[0044] The comprehensive prospect value considering the bounded rationality of EV is expressed as:

[0045]

[0046] In the formula, π + [C m,z (t)], π - [C m,z (t)] are respectively the weights of EV facing gains and losses.

[0047] Furthermore, the spatio-temporal guiding electricity price model is specifically as follows: The load time distribution characteristic is expressed as:

[0048]

[0049] The time-guiding electricity price is expressed as:

[0050]

[0051] In the formula, μ time is the sensitivity coefficient of the time-guiding electricity price to the load time distribution characteristic of the distribution network;

[0052] The load distribution characteristic of distribution network FCS q is expressed as:

[0053]

[0054] The space-guiding electricity price is expressed as:

[0055]

[0056] In the formula, η space is the sensitivity coefficient of the space-guiding electricity price to the load distribution characteristic, and r-q represents the base station located at the same distribution network node as FCS q;

[0057] The spatio-temporal guiding electricity price is expressed as:

[0058]

[0059] Wherein, are respectively the basic charging and discharging electricity prices obtained according to the power flow tracking method.

[0060] Furthermore, the objective function of the optimization problem is

[0061]

[0062] Wherein, μ = {μ m,a (t)}, y = {ym,n(t)}; C1 is the EV section selection constraint; C2 is the EV server selection variable constraint; C3 is the EV charging station selection constraint; C4 is the charging and discharging state constraint; C5 - C6 are the upper and lower limits of the EV charging and discharging power constraints, where, and are respectively the upper and lower limits of the charging and discharging power of EVm; C7 - C8 are respectively the distribution network node voltage and safety current constraints, where, v i,min and v i,max are respectively the minimum and maximum values allowed for the voltage of node i under the safe condition of the distribution network, and I i,j,min and I i,j,max are respectively the minimum and maximum values allowed for the current of branch ij; C9 is the EV end-to-end delay probability constraint, where, is the upper limit of the EV end-to-end delay, and τ P is the communication reliability threshold, that is, the probability that the EV end-to-end delay is lower than the upper limit should be higher than τ P .

[0063] Furthermore, the partially observable Markov cooperative game model is:

[0064] Agent state and action: The state S m (t) and action a m (t) of agent m are respectively:

[0065]

[0066] Wherein, the state space includes the spatio-temporal guiding charging and discharging electricity prices of each FCS, the average charging and discharging queuing delays, the free passage times of each section, the SoC, the target arrival time, the target SoC, the backlog of the local data storage queue, and the backlog information of the data processing queue of the edge server of the EV; the action space includes the section selection variable, the charging and discharging selection variable, the charging and discharging power selection variable, and the autonomous driving task offloading server selection indication variable;

[0067] Global reward for agent cooperation game: Based on line loss, grid load stability, and balance degree, the global reward for agent cooperation game is defined, and the security constraints of the distribution network are defined in it in the form of penalty functions, expressed as

[0068] R Global (t) = -V F ΔL F (t) - V G ΔL G (t) - V loss C DN,loss (t)

[0069] -V U B U (t) - V I B I (t)(29)

[0070]

[0071] In the formula, V U 、V I are the penalty coefficients for grid node voltage violation and branch current overload respectively;

[0072] Aggregate reward for agents: Set the global optimization goal as the global reward, convert the constraint conditions into penalty terms and add them to the global reward. The VCG cost of agent m is expressed as:

[0073]

[0074] In the formula, represents the global reward when agent m does not participate in demand response;

[0075] The aggregate reward r m (t) of agent m is expressed as

[0076]

[0077] In the formula, V C represents the weight of the electric vehicle experience quality, and V tau represents the weight of exceeding the time limit for autonomous driving task data processing.

[0078] Furthermore, the EV demand response scheduling algorithm based on spatio-temporal electricity price guidance and migration cooperation includes four steps: spatio-temporal electricity price guidance, demand response action decision and execution, weighted experience perception learning, and migration cooperation:

[0079] Step 1: Spatio-temporal electricity price guidance

[0080] The distribution network control center first calculates the time distribution characteristics and spatial distribution characteristics of the grid load respectively. Secondly, it calculates the time-guided electricity price and the spatial-guided electricity price. Finally, it calculates the spatio-temporal guided electricity price and publishes this electricity price to each EV;

[0081] Step 2: Action decision and execution

[0082] Based on the published spatio-temporal guided electricity price, the EV agent m inputs the state S m (t) into the policy network π m to obtain the policy π m (·|S m (t)), and selects an action a based on this policy m ; Secondly, the EV executes the action to obtain the immediate reward r m , observes the state S m (t + 1) at the next moment, and forms an experience sample Then, calculate the TD-error of the two critic evaluation networks, expressed as:

[0083]

[0084] In the formula, is the discount factor, is the action selected according to the policy π m (·|S m (t + 1)); Set the sample TD-error judgment threshold TD m,max . When the experience sample meets the following conditions, discard the sample and do not put it into the experience replay pool;

[0085]

[0086] Otherwise, put it into the experience replay pool for subsequent model training; Assume that this experience sample is the h-th sample in the experience replay pool, then the priority of this sample can be expressed as:

[0087]

[0088] In the formula, v is the experience sample priority supplement coefficient, which is used to avoid samples with too small TD-error being ignored;

[0089] Step 3: Weighted experience perception learning

[0090] According to the priority of the samples in the experience replay pool, the probability of each sample being drawn can be obtained as

[0091]

[0092] In the formula, H is the capacity of the experience replay pool, is the priority sampling index;

[0093] Sample a group of empirical samples from the experience replay pool based on the probability of the empirical samples being drawn and calculate the critic network loss function as

[0094]

[0095] where ε h (t) is the dynamic weight of the empirical sample, which is used to balance the deviation of the loss function calculation caused by introducing the empirical sample drawing probability;

[0096] Update the two critic networks using the stochastic gradient descent method based on the loss function, and periodically update the two target critic networks using the soft update method, which is specifically expressed as

[0097]

[0098] where λ m is the gradient descent step size; κ m is the soft update coefficient;

[0099] The loss function of the EV intelligent agent m actor policy network can be characterized by the Kullback-Leibler divergence between the policy output by the policy network and the Q value, which is specifically expressed as

[0100]

[0101] Update the actor network using the stochastic gradient descent method, and update the entropy regularization temperature coefficient in an adaptive manner, which is expressed as:

[0102]

[0103] where dim(a m (t)) is the action dimension of the intelligent agent m;

[0104] Step 4: Knowledge transfer and collaboration

[0105] When the decrease amplitude of the EV intelligent agent reward function exceeds the set threshold, that is then it is determined that a large change has occurred in the EV environment, and the migration and collaboration mechanism is triggered; First, the EV that triggers the migration and collaboration mechanism sends a model migration request to other EVs within its communication coverage range, and the EV that receives the request transmits its own model to the EV that sends the request; Second, the EV that sends the request puts the received model into the set Extract a set of state sampling sets S samp ={s1, s2,..., s D} from its own state space, and respectively according to The model to be migrated in obtains the action set corresponding to the state sampling set Then, according to the action set, the corresponding reward set can be obtained Finally, calculate the cumulative reward of each model to be migrated, and select the model to be migrated with the optimal performance according to the cumulative reward, which is specifically expressed as:

[0106]

[0107] In the formula, is the cumulative reward of the model; b * is the optimal model to be migrated in; r0 is the cumulative reward of the EV's own model that issues the migration request. When μ > 1, it means that using the optimal model to be migrated can obtain higher rewards in the current environment, and the model to be migrated can be migrated to its own agent; otherwise, migration is not suitable.

[0108] A communication, power, and transportation coupled network resource collaborative scheduling system includes:

[0109] The demand response module of the distribution network control center: used to calculate and publish the spatio-temporal guiding electricity price by monitoring the spatio-temporal distribution characteristics of the grid load, guide the EV to reasonably select the charging and discharging behavior in the time and space dimensions, and optimize the grid load distribution; based on the grid load status, EV charging and discharging power feedback, and base station power consumption feedback, schedule the distribution network power flow to reduce the peak load and grid loss;

[0110] The demand response module of the base station: used to receive the EV computing task offloading, provide low-latency and high-reliability communication services for the EV, cooperate with the edge server to perform the offloading and computing processing of the EV's autonomous driving task data, ensure the minimum backlog of the task queue, meet the latency requirements, and complete the transmission of the computed data; monitor and calculate its own communication data processing power consumption and cooling power consumption, and feedback the power consumption data to the demand response module of the distribution network control center for power grid scheduling and energy management;

[0111] The EV demand response module: used to determine the computing offloading target of the autonomous driving data according to the service quality of the base station and the edge server, ensure low-latency data processing, and feedback the autonomous driving task offloading decision to the demand response module of the base station; under the guidance of the spatio-temporal guiding electricity price published by the distribution network control center, select the charging and discharging time period and charging and discharging power, and feedback the charging and discharging decision information to the demand response module of the distribution network control center to participate in the power grid demand response, realize peak shaving and valley filling, and balance the grid load; select the optimal travel path and charging station according to the congestion situation and charging demand of the transportation network to relieve traffic congestion and optimize the charging efficiency.

[0112] Furthermore, the EV demand response module includes the following sub-modules:

[0113] Information collection module: It is used to collect the EV battery status, charging station status, grid and traffic network path status information in real time, and transmit the information to the decision-making module;

[0114] Experience perception learning module: It is used to optimize the demand response strategy of EVs by learning from historical decision-making experiences, improve the decision-making efficiency and response effect of EVs in the dynamic power-transportation-communication coupling network environment; based on weighted experience replay and priority sampling, learn and update different experience samples according to their importance, and strengthen the demand response strategy of EVs;

[0115] Decision-making module: It is used to generate charging and discharging decisions, path decisions and task offloading decisions based on the cumulative prospect theory, integrating factors such as electricity price, path status and task delay;

[0116] Knowledge transfer collaborative control module: It is used to solve problems such as poor initial model performance, long learning time and poor convergence effect of EVs in a high-speed and dynamically changing environment, improve the decision-making efficiency and demand response effect of EVs, realize the strategy knowledge transfer and collaborative optimization between EVs, and quickly adapt to the new system state when the environment changes;

[0117] Task queue management module: It is used to dynamically maintain the task data queue and track the task backlog and processing delay;

[0118] Data transmission module: It is used to send the EV charging and discharging behaviors to the demand response module of the distribution network control center, including the selected charging / discharging node and power information; the traffic network status, including the travel path of the EV and the current road congestion information; send the data task transmission demand and calculation task offloading decision to the demand response module of the base station; the backlog of the local task queue and the task data size information; receive the spatio-temporal guiding electricity price and the grid node load status fed back by the demand response module of the distribution network control center; receive the edge computing results and data processing delay fed back by the demand response module of the base station.

[0119] Furthermore, the base station demand response module includes the following sub-modules:

[0120] Data receiving module: It is used to receive the data task transmission demand and calculation task offloading decision from the EV demand response module; receive the backlog of the EV local task queue and the task data size information;

[0121] Edge server data calculation module: It is used to provide computing task processing services for EVs by collaborating with edge servers, such as autonomous driving data analysis, path planning, obstacle recognition, etc.; dynamically allocate computing resources to process the data tasks unloaded by EVs;

[0122] Result Return Module: It is used to send edge computing results and data processing delays to the EV Demand Response Module, support the autonomous driving decision-making or demand response scheduling of EVs, ensure the real-time performance and reliability of result return, and reduce the delay in the communication link;

[0123] Base Station Energy Consumption Management Module: It is used to send the real-time power consumption information of the base station to the Demand Response Module of the Distribution Network Control Center.

[0124] Furthermore, the Demand Response Module of the Distribution Network Control Center includes the following sub-modules:

[0125] Power Acquisition Module: It is used to acquire the power data of each node and line, including information such as node voltage, current, and power flow distribution;

[0126] Power Monitoring Module: It is used to monitor the power flow conditions of each node and line in the distribution network in real time, including active power and reactive power; monitor the charging and discharging power and energy consumption of FCS and base stations to ensure the stable operation of the power grid; monitor whether the voltage and current of the power grid exceed the limit, and obtain node load distribution and line power flow data;

[0127] Electricity Price Calculation Module: It is used to calculate time-guided electricity prices, space-guided electricity prices, and form spatio-temporal guided electricity prices based on the time load distribution and space load distribution characteristics of the distribution network. Guide EVs to make charging and discharging decisions through electricity prices, achieve peak shaving and valley filling, balance the grid load, and improve the operational stability and economy of the power grid;

[0128] Load Analysis and Stability Evaluation Module: It is used to evaluate the load stability and load balance of the distribution network, and analyze the impact of EV charging and discharging on the operation of the power grid;

[0129] Demand Response Scheduling Module: It is used to coordinate EV charging and discharging and travel path selection based on the demand response behavior of EVs and the game optimization model, and improve the overall performance of the power grid and transportation network; use spatio-temporal electricity prices to guide EVs to perform demand response, optimize the load stability, line loss, and traffic congestion of the power grid; update the electricity price strategy in combination with EV feedback information (such as charging and discharging power, charging station selection, path selection, etc.), and dynamically regulate demand response behavior;

[0130] Data Interaction and Communication Module: It is used to receive the real-time power consumption information of the base station transmitted by the Base Station Demand Response Module; receive the charging and discharging behavior of the EV Demand Response Module, including the selected charging / discharging node and power information; send spatio-temporal guided electricity prices and the load status of power grid nodes to the EV Demand Response Module.

[0131] With the above solution, the present invention has at least the following advantages:

[0132] (1) The present invention proposes a network structure for the integration of communication, power, and transportation networks, which improves the stability and balance of the power grid load through multi-resource collaborative optimization. The transportation network accurately simulates the path selection, charging station selection, and charging and discharging behaviors of electric vehicles (EVs) by combining the FCS model, optimizes the travel path and charging strategy, and significantly improves transportation efficiency; the distribution network model containing FCS quantifies the impact of EV participation in demand response on the distribution network from both time and space dimensions, and analyzes the impact of EV participation in demand response on the power flow, network loss, and energy efficiency optimization of the distribution network through load stability and load balance models; the communication network relies on 4G / 5G / 6G technologies to build base stations and edge servers, provides low-latency and high-efficiency communication services for EVs, supports the two-way interaction between FCS and the power grid, and realizes the effective interconnection and interoperability of EVs with the power grid and transportation network; a spatio-temporal guiding electricity price model is proposed to guide EVs to reasonably select charging and discharging time periods and charging stations, optimize the spatial and temporal distribution of the power grid load, further improve the load balance and stability of the power grid, reduce the congestion degree of the transportation network, reduce the travel time and anxiety of users, and provides an efficient communication-power-transportation coupled network resource scheduling scheme.

[0133] (2) The present invention proposes an EV demand response scheduling algorithm based on spatio-temporal electricity price guidance and migration collaboration, which optimizes the stability and balance of the power grid load and reduces the user's road travel time. This algorithm conducts demand response guidance for EVs in both time and space dimensions by constructing a spatio-temporal guiding electricity price, optimizes the stability and balance of the power grid load, reduces the road congestion degree of the transportation network, and reduces the user's road travel time. At the same time, the algorithm supports the migration and collaboration of demand response strategy knowledge among EVs, realizes the interactive optimization between the EV demand response strategy and the communication-power-transportation coupled network, and improves the EV demand response effect. In addition, the present invention constructs a demand response model based on the cumulative prospect theory, accurately describes the decision-making behavior of EV users when facing uncertainties or random events, reflects the costs and benefits of EVs in different charging and discharging decisions, path selections, and travel time periods, and realizes a demand response scheduling that is more in line with actual needs. Based on the communication network model, the present invention studies the power consumption of base stations and the energy efficiency of edge servers, and proposes a delay optimization strategy applicable to the information transmission between EVs and base stations and edge servers, accurately constructs the task queue queuing delay of EVs and the task queue queuing delay of edge servers, ensures the low-latency performance of the system under high-efficiency communication, provides strong support for key applications such as autonomous driving, and ensures the high availability and efficient circulation of data.

[0134] The above description is only an overview of the technical solution of the present invention. In order to be able to more clearly understand the technical means of the present invention and implement it in accordance with the content of the specification, the following describes the preferred embodiments of the present invention in detail in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0135] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0136] Figure 1 is the flowchart of the EV demand response scheduling method based on the coordination of time-space electricity price guidance and migration of the present invention;

[0137] Figure 2 is the structure diagram of the communication-electricity-transportation triple-network coupling demand response system of the present invention;

[0138] Figure 3 is the composition diagram of the EV demand response module in the communication-electricity-transportation triple-network coupling demand response system of the present invention;

[0139] Figure 4 is the composition diagram of the base station demand response module in the communication-electricity-transportation triple-network coupling demand response system of the present invention;

[0140] Figure 5 is the composition diagram of the distribution network control center demand response module in the communication-electricity-transportation triple-network coupling demand response system of the present invention. Specific Embodiments

[0141] The following will further describe in detail the specific embodiments of the present invention in conjunction with the drawings and embodiments. The following embodiments are used to illustrate the present invention but are not used to limit the scope of the present invention.

[0142] First, establish a communication-electricity-transportation coupling network model, fully consider the heterogeneity of EV users, and construct a demand response model considering the bounded rationality of EV users. Further, analyze the grid load characteristics from two dimensions of time and space, and on this basis, construct the time and space guiding electricity prices for EV demand response, and propose a time-space guiding electricity price model. Secondly, propose an EV demand response scheduling algorithm based on the coordination of time-space electricity price guidance and migration, and realize the comprehensive optimization of the communication-electricity-transportation coupling network through the coordination of time-space electricity price guidance and the migration of demand response strategy knowledge between EVs.

[0143] I. EV Demand Response Model Based on Communication Network-Electricity-Transportation Coupling Network

[0144] (1) Communication Network-Electricity-Transportation Coupling Network Architecture

[0145] The present invention proposes a communication network - power - transportation coupled network architecture, including five types of resources such as EVs, FCSs, base stations, and edge servers. The base station and the edge server are located on the communication network side, and they are distributed in the vicinity of each other, providing high - speed data transmission and computing support for EVs in the transportation network, and being able to carry complex services such as autonomous driving and demand response. As the coupling point between the power grid and the communication network, the base station conducts information interaction with the charging stations in the transportation network and the power grid through high - speed communication links. On the one hand, it provides support for power grid - side monitoring and control; on the other hand, it obtains the location information and task requests of EVs on the transportation network side, providing data transmission support for EV autonomous driving, path optimization, and task offloading. The FCS is located on the transportation network side and is the coupling node between the power grid and the transportation network. On the one hand, the FCS provides charging and discharging services for EVs on the transportation network side, supporting the needs of EVs during path selection and energy replenishment; on the other hand, as a load on the power grid side, the FCS feeds back real - time load information to the power grid through its connection with the communication network, and guides the optimization of EV charging and discharging behaviors based on power grid price signals to mitigate the impact of large - scale new energy access on the stability of the power grid. As the core support of the system energy, the power grid provides power support for the base station through real - time connection with the communication network, and realizes dynamic optimization by monitoring the energy consumption of the base station, improving energy utilization efficiency. The base station in the communication network also provides two - way information interaction for the power grid side, supporting the power grid to dynamically regulate the demand response behaviors of EVs in the transportation network. As an important coupling node between the communication network and the transportation network, the EV makes comprehensive travel decisions based on the power grid price and power load information, the distribution of charging stations and road congestion information in the transportation network, and the distribution of computing resources of edge servers and the channel state information of base stations in the communication network: First, at the beginning of each time period, the EV generates autonomous driving task data, and at the same time collects information such as power price and power load from the power grid side, collects information such as the distribution of charging stations and road congestion from the transportation network side, and obtains information such as the distribution of computing resources of available edge servers and the channel state of base stations from the communication network side. Second, based on the collected multi - source information, the in - vehicle intelligent agent uses built - in algorithms to make comprehensive decisions, determining whether charging and discharging are needed and the optimal path, and at the same time selecting the most suitable edge server for task processing. Subsequently, the EV offloads the generated task data to the nearest base station, and the base station transmits the task data to the selected edge server for processing through a high - speed communication link. In the above process, the power grid price is a spatio - temporal guiding price constructed based on the distribution characteristics of the power grid load in time and space. Based on this price, on the one hand, it guides EVs to discharge during peak hours and charge during off - peak hours, thus realizing load peak shaving and valley filling and improving the stability of the power grid load curve; on the other hand, it guides EVs to choose charging stations with lower loads for charging, reducing the congestion degree of charging stations and roads. At the same time, as the EV's position changes, it realizes the dynamic transfer of communication load between different base stations, improving the balance of the communication network load, ensuring low - latency performance, and providing strong support for autonomous driving.

[0146] Suppose there are a total of M EVs and N edge servers. The sets of EVs and edge servers are denoted as and The total optimization duration is divided into continuous T, indexed by t = 1, 2, …, T. The duration of a time period is denoted as τ. The route selection variable of an EV is defined as μ m,a (t), where μ m,a (t) = 1 indicates that EV m selects route a to travel in the t-th time period. The FCS selection variable of an EV is defined as and indicates that EV m selects FCS q for charging, indicates that EV m selects FCS q for discharging. The autonomous driving data offloading decision variable of an EV is defined as y m,n (t), y m,n (t) = 1 indicates that EV m offloads its autonomous driving data to edge server n for processing through the base station closest to it.

[0147] (2) Traffic network model with FCS

[0148] The traffic network model with FCS can be represented as an undirected graph G T =(N T , A T , Q), where N T represents the set of road network nodes; A T represents the set of road segments, and road segment a ∈ A T ; Q represents the set of FCSs in the road network, and FCS q ∈ Q. The traffic road network has strong time-varying characteristics, and the traffic flows in different time periods are coupled with each other. The semi-dynamic traffic flow model considers the residual flow phenomenon in the road network and can well reflect the coupling between traffic flows in different time periods. The semi-dynamic traffic flow assignment model is specifically expressed as

[0149]

[0150] In the formula, Ω od is the set of EV travel demands; K od is the set of paths that satisfy travel demand od; is the traffic flow of travel demand od on path k; δ k,a (t) is a binary variable representing the relationship between the path and the road segment. δ k,a (t) = 1 indicates that road segment a is located on path k. Conversely, δ k,a (t) = 0; is the residual flow of demand od, τ is the time period length, r od , r od,modThey are the original travel demand flow and the travel demand flow after considering the residual flow correction; τ a (t), and are the travel time of section a and the charging / discharging queuing time of FCS q respectively.

[0151] (3) Distribution network model with FCS

[0152] The distribution network model with FCS can be represented as an undirected graph G P =(N P , A P , Q), where N P represents the set of distribution network nodes; A P represents the set of distribution network line branches, and branch ij ∈ A P .

[0153] To reflect the impact of EV participation in demand response on the distribution network, load stability and load balance models are constructed respectively to quantify the impact of EV participation in demand response on the distribution network from the two dimensions of time and space. The distribution network power flow model is specifically expressed as

[0154]

[0155] In the formula, P i,j (t) and Q i,j (t) are the active power and reactive power flowing from node i to node j on branch ij respectively, X i,j is the reactance of branch ij, and v i (t) is the voltage of node i.

[0156] The distribution network power loss C DN,loss (t) is expressed as

[0157]

[0158] In the formula, C loss (t) is the unit cost of power loss, R i,j is the resistance of branch ij, and I i,j (t) is the current of branch ij.

[0159] (4) Communication network model

[0160] The total power consumption of the base station is expressed by the following formula (8), which mainly includes data processing power consumption and refrigeration power consumption

[0161]

[0162] Among them, the data processing power consumption of the edge server is expressed as

[0163]

[0164] In the formula, P peak , P ideal are respectively the peak power and standby power of a single edge server, κ m is the average data service rate of a single server, which is related to the CPU operating frequency of the server, N server is the number of front-end processors working in the data center, H n (t) is the task data processed by the data center.

[0165] The operating power of the edge server refrigeration equipment is expressed as

[0166]

[0167] In the formula, is the air supply temperature of the refrigeration equipment, ξ n,2 , ξ n,1 and ξ n,0 are respectively the refrigeration coefficients.

[0168] The queuing delay of the task queue on the EV side of the communication network is

[0169]

[0170] In the formula, Q m (t) represents the backlog of the task queue of EV m, K m (t), α m respectively represent the number of tasks and the size of a single task.

[0171] The queuing delay of the task queue on the edge server side is

[0172]

[0173] (5) EV Bounded Rationality Demand Response Model

[0174] Different EVs have strong heterogeneity in their subjective feelings about charging and discharging decisions and travel routes, and have the characteristics of bounded rationality when making decisions. Cumulative prospect theory can describe the decision-making behavior of EV users with bounded rationality due to the differences in the perception of decision-making information in the face of uncertainty or random events, and is applicable to the modeling of EV bounded rationality demand response.

[0175] The costs mainly considered in EV demand response include four aspects: the demand response cost C m,1 (t) on the power grid side; the travel time cost C m,2 (t) and travel anxiety cost C m,3 (t) on the transportation network side; the service data processing time cost C m,4 (t) on the communication network side. They are respectively expressed as

[0176]

[0177]

[0178] Wherein, μ m,a (t) is the route selection indication variable of EV m, and μ m,a (t) = 1 indicates that EV m selects route a; respectively represent the charge and discharge indication variables of EV m at FCS q, represents charging, represents discharging; are the spatio-temporal guiding charge and discharge prices respectively; are the travel time anxiety coefficient and the SoC anxiety coefficient respectively; are the expected arrival time and the expected SoC respectively; is the indication variable of whether the destination is reached, indicates that the destination has not been reached.

[0179] Based on the prospect utility value function, the objective cost / benefit of EV users can be transformed into the subjective cost / benefit considering bounded rationality. Based on the positive and negative expectations of EV as reference points respectively, the positive and negative prospect values of EV can be obtained as

[0180]

[0181] Wherein, C m,z (t) (z = 1, 2, 3, 4) respectively represent the demand response cost, travel time cost, travel anxiety cost and data processing time cost of EV users; v + [C m,z (t)], v - [C m,z (t)] are the positive and negative prospect values of EV based on its positive and negative expectations as the benchmark.

[0182] The comprehensive prospect value considering the bounded rationality of EV is expressed as

[0183]

[0184] Wherein, π + [C m,z (t)], π - [C m,z (t)] are the weights of EV facing benefits and losses respectively.

[0185] (6) Spatio-temporal guiding electricity price model

[0186] The time-guided electricity price guides EVs to reasonably select charge and discharge periods on the time scale based on the time distribution characteristics of the power grid load, so as to realize peak shaving and valley filling. The time distribution characteristics of the load are expressed as

[0187]

[0188] When ΔG(t) > 0, it indicates that the current load of the distribution network is higher than the average load. At this time, EVs should be guided to discharge; on the contrary, EVs are guided to charge. Therefore, the time-guided electricity price can be expressed as

[0189]

[0190] In the formula, μ time is the sensitivity coefficient of the time-guided electricity price to the time distribution characteristics of the distribution network load.

[0191] The space-guided electricity price aims to guide the balanced distribution of EVs in space. On the one hand, it guides the balanced distribution of EVs in the road network to reduce road network congestion; on the other hand, it guides the balanced distribution of EV loads among FCSs to improve the load balance of grid nodes.

[0192] The load distribution characteristics of the distribution network FCS q are expressed as

[0193]

[0194] When ΔL q (t) > 0, it indicates that the load of FCS q is higher than the average load at this time. EVs should be guided to charge at other FCSs or discharge at this FCS. On the contrary, EVs are guided to discharge at other FCSs or charge at this FCS, and ΔL r (t) is similar. Therefore, the space-guided electricity price is expressed as

[0195]

[0196] In the formula, η space is the sensitivity coefficient of the space-guided electricity price to the load distribution characteristics, and r-q represents the base station located at the same distribution network node as FCS q.

[0197] Therefore, the spatio-temporal guided electricity price is expressed as

[0198]

[0199] In the formula, are the basic charging and discharging electricity prices obtained according to the power flow tracking method respectively.

[0200] II. EV Demand Response Scheduling Strategy Based on the Collaboration of Spatio-Temporal Electricity Price Guidance and Migration

[0201] (1) Optimization Problem

[0202] By optimizing the EV travel path, charging and discharging nodes, and charging and discharging power, minimizing the weighted sum of EV road travel time, demand response revenue, travel anxiety, grid load stability, balance degree, and network loss, the objective function is constructed as

[0203]

[0204] where μ = {μ m,a (t)}, y = {y m,n (t)}. C1 is the EV section selection constraint; C2 is the EV server selection variable constraint; C3 is the EV charging station selection constraint; C4 is the charging and discharging state constraint; C5 - C6 are the upper and lower limits of the EV charging and discharging power constraints, where and are the upper and lower limits of the charging and discharging power of EV m respectively; C7 - C8 are the distribution network node voltage and safety current constraints respectively, where v i,min and v i,max are the minimum and maximum values allowed for the voltage of node i under the safe condition of the distribution network respectively, and I i,j,min and I i,j,max are the minimum and maximum values allowed for the current of branch ij respectively; C9 is the EV end - to - end delay probability constraint, where is the upper limit of the EV end - to - end delay, and τ P is the communication reliability threshold, that is, the probability that the EV end - to - end delay is lower than the upper limit should be higher than τ P .

[0205] (2) EV Demand Response Scheduling Algorithm Based on the Collaboration of Spatiotemporal Electricity Price Guidance and Migration

[0206] 1) Partially Observable Markov Cooperative Game Model

[0207] The communication - power - transportation coupling network has strong real - time performance and a large scale. During the driving process, EVs can only make demand response decisions based on the observed partial environmental information, and there is a certain game relationship among EVs. Therefore, the EV demand response can be modeled as a partially observable Markov cooperative games (POMCG) model, and the specific introduction is as follows.

[0208] Agent State and Action: The state S m (t) and action a m (t) of agent m are respectively

[0209]

[0210]

[0211] In the formula, the state space includes the spatio-temporal guiding charging and discharging electricity prices of each FCS, the average charging and discharging queuing delays, the free passage times of each road section, the SoC of the EV, the target arrival time, the target SoC, the backlog of the local data storage queue, and the backlog information of the data processing queue of the edge server. The action space includes the road section selection variable, the charging and discharging selection variable, the charging and discharging power selection variable, and the autonomous driving task offloading server selection indication variable.

[0212] Global reward for agent cooperation game: All agents cooperate to participate in demand response to reduce network losses, improve the stability and balance of the power grid load. Therefore, the global reward for agent cooperation game is defined based on network losses, power grid load stability, and balance, and the security constraints of the distribution network are considered in the form of penalty functions, expressed as

[0213]

[0214] In the formula, V U , V I are the penalty coefficients for voltage violation at the power grid nodes and branch current overload respectively.

[0215] Aggregate reward for agents: The optimization goal constructed in this paper is jointly affected by the behaviors of global EVs. Therefore, the global optimization goal is set as the global reward. To promote the cooperation relationship among EVs, this paper adopts the Vickrey-Clarke-Groves (VCG) auction mechanism to allocate rewards according to the differential contributions of EVs to the global reward, so as to improve the global revenue. At the same time, in order to make the decisions of EVs meet the constraint conditions, the constraint conditions are converted into penalty terms and added to the global reward, and penalties are given when the decisions of EVs exceed the limits, ensuring that the decisions of EVs meet the constraint conditions. The VCG cost of agent m is expressed as

[0216]

[0217] In the formula, represents the global reward when agent m does not participate in demand response.

[0218] Therefore, the aggregate reward r m (t) of agent m is expressed as

[0219]

[0220] In the formula, V C represents the weight of the Quality of Experience (QoE) of the electric vehicle, and V tau represents the weight of exceeding the data processing time limit of the autonomous driving task.

[0221] 2) EV Demand Response Scheduling Algorithm Based on Spatiotemporal Electricity Price Guidance and Migration Collaboration

[0222] Traditional EV demand response scheduling algorithms ignore the spatiotemporal characteristics of electricity prices and cannot guide the charging, discharging, and travel of EVs in both the time and space dimensions in the power-transportation coupling network, resulting in poor stability and balance of the power grid load and long travel times for EVs on the road. They also ignore the migration and collaboration of demand response strategy knowledge among EVs, leading to poor initial performance, long learning times, and poor convergence effects of their agents when the environment in which the EVs are located changes, and are not suitable for the dynamically changing communication-power-transportation coupling network.

[0223] To address the above problems, a demand response scheduling algorithm based on spatiotemporal electricity price guidance and migration collaboration is proposed. This algorithm constructs a spatiotemporal guidance electricity price to guide EVs to perform demand response in both the time and space dimensions. In the time dimension, the spatiotemporal guidance electricity price guides EVs to discharge during peak power grid loads and charge during valley loads according to the fluctuations of the power grid load curve, achieving peak shaving and valley filling of the load curve. In the space dimension, the electricity price guides EVs to charge at FCSs with lower loads and discharge at FCSs with higher loads according to the load distribution of each FCS, achieving spatial load balance. This method effectively improves the stability and balance of the power grid load, alleviates traffic congestion, and reduces user travel time. At the same time, the proposed algorithm supports the migration and collaboration of demand response strategy knowledge among EVs. When the decline in the reward function of the EV agent exceeds a set threshold, it is determined that the environment has changed significantly, and the migration collaboration mechanism is triggered. The EV that triggers the mechanism sends a model migration request to other EVs within the communication coverage range, and the receiving EV transmits its own model to the requesting party. The requesting party uses the received model to generate action and reward sets, calculates the cumulative reward, and then selects the model with the optimal performance for update. This mechanism effectively solves problems such as poor initial performance, long learning time, and poor convergence effect of agents caused by the dynamically changing communication-power-transportation coupling network environment, realizes the interactive optimization among the EV demand response strategy, the power grid, and the transportation network, and improves the EV demand response effect.

[0224] The demand response scheduling algorithm based on spatiotemporal electricity price guidance and migration collaboration mainly includes four steps: spatiotemporal electricity price guidance, demand response action decision and execution, weighted experience perception learning, and migration collaboration, which are introduced as follows.

[0225] Step 1: Spatiotemporal Electricity Price Guidance

[0226] The distribution network control center first calculates the time distribution characteristics and space distribution characteristics of the power grid load respectively. Secondly, it calculates the time guidance electricity price and the space guidance electricity price. Finally, it calculates the spatiotemporal guidance electricity price and publishes this electricity price to each EV.

[0227] Step 2: Action Decision and Execution

[0228] Based on the published spatio-temporal guided electricity price, EV agent m takes the state S m (t) as the input to the policy network π m to obtain the policy π m (·|S m (t)), and selects an action a based on this policy m . Secondly, the EV executes the action to obtain the immediate reward r m , observes the state S m (t + 1) at the next moment, and forms an experience sample Then, calculate the TD-error of the two critic evaluation networks, denoted as

[0229]

[0230] In the formula, is the discount factor, is the action selected according to the policy π m (·|S m (t + 1)). The TD-error can measure the degree of correction of the experience sample to the agent model. The larger the absolute value of the TD-error, the higher the degree of correction of the experience sample to the agent model, that is, the higher the learning value of this sample. However, when the TD-error is negative, it indicates that the agent has selected a poor-performing action. At this time, the experience sample with too large an absolute value of the TD-error is likely to correct the agent model in the reverse direction and reduce the model training effect. Therefore, set the sample TD-error decision threshold TD m,max . When the experience sample meets the following conditions, discard this sample and do not put it into the experience replay pool

[0231]

[0232] Conversely, put it into the experience replay pool for subsequent model training. Assume that this experience sample is the h-th sample in the experience replay pool, then the priority of this sample can be expressed as

[0233]

[0234] In the formula, v is the experience sample priority supplement coefficient, which is used to avoid ignoring samples with too small TD-errors

[0235] Step 3: Weighted experience-aware learning

[0236] According to the priorities of the samples in the experience replay pool, the probability of each sample being drawn can be obtained as

[0237]

[0238] Wherein, H is the capacity of the experience replay pool, is the prioritized sampling exponent.

[0239] Based on the probability of the experience sample being drawn, a group of experience samples are drawn from the experience replay pool, and the loss function of the critic network is calculated as When

[0240]

[0241] In the formula, ε h (t) is the dynamic weight of the experience sample, which is used to balance the deviation of the loss function calculation caused by introducing the experience sample drawing probability.

[0242] Based on the loss function, the two critic networks are updated using the stochastic gradient descent method, and the two target critic networks are updated regularly using the soft update method, which can be specifically expressed as

[0243]

[0244] In the formula, λ m is the gradient descent step size; κ m is the soft update coefficient.

[0245] The loss function of the EV intelligent agent m actor policy network can be characterized by the Kullback-Leibler (KL) divergence between the policy output by the policy network and the Q value, and is specifically expressed as

[0246]

[0247] Similarly, the actor network is updated using the stochastic gradient descent method, and the entropy regularization temperature coefficient is updated in an adaptive manner, which is expressed as

[0248]

[0249] In the formula, dim(a m (t)) is the action dimension of the intelligent agent m.

[0250] Step Four: Knowledge Transfer Collaboration

[0251] When the reduction amplitude of the EV intelligent agent reward function exceeds the set threshold, that is at this time, it is determined that a large change has occurred in the EV environment, and the migration collaboration mechanism is triggered. First, the EV that triggers the migration collaboration mechanism sends a model migration request to other EVs within its communication coverage range, and the EV that receives the request transmits its own model to the EV that sends the request. Second, the EV that sends the request puts the received model into the set A group of state sampling sets S are drawn from its own state spacesamp = {s1, s2, …, s D}, and respectively obtain the action set corresponding to the state sampling set according to the pending migration model in Then, the corresponding reward set can be obtained according to the action set Finally, calculate the cumulative reward of each pending migration model, and select the pending migration model with the optimal performance according to the cumulative reward, which is specifically expressed as

[0252]

[0253] In the formula, is the model cumulative reward; b * is the optimal pending migration model in

[0254] III. Structure of Communication-Power-Transportation Three-Network Coupled Demand Response System

[0255] As Figure 2 shown, the communication-power-transportation three-network coupled demand response system includes a distribution network control center demand response module, a base station demand response module, and an EV demand response module.

[0256] Distribution network control center demand response module: By monitoring the spatio-temporal distribution characteristics of the grid load, calculate and publish spatio-temporal guiding electricity prices to guide EVs to reasonably select charging and discharging behaviors in the time and space dimensions, and optimize the grid load distribution; Based on the grid load status, EV charging and discharging power feedback, and base station power consumption feedback, schedule the power flow of the distribution network to reduce peak load and grid losses.

[0257] Base station demand response module: Receive EV computing task offloading, provide low-latency and high-reliability communication services for EVs, cooperate with edge servers to perform data offloading and computing processing of EV autonomous driving tasks, ensure the minimum backlog of task queues, meet the latency requirements, and complete the transmission of computing data back; Monitor and calculate its own communication data processing power consumption and cooling power consumption, and feedback the power consumption data to the distribution network control center demand response module for grid scheduling and energy management.

[0258] EV Demand Response Module: Determine the computing offloading target of autonomous driving data according to the service quality of the base station and the edge server, ensure low-latency data processing, and feedback the autonomous driving task offloading decision to the base station demand response module; Under the guidance of the spatio-temporal guiding electricity price issued by the distribution network control center, select the charging and discharging time periods and charging and discharging power, and feedback the charging and discharging decision information to the distribution network control center demand response module to participate in the power grid demand response, achieve peak shaving and valley filling, and balance the power grid load; According to the congestion situation of the transportation network and the charging demand, select the optimal travel route and charging station to relieve traffic congestion and optimize the charging efficiency.

[0259] The composition of the EV demand response module in the communication-electricity-transportation triple-network coupling demand response system is as Figure 3 shown, and the introduction of each functional sub-module is as follows:

[0260] Information Acquisition Module: Real-time collect the EV battery status, charging station status, power grid and transportation network path status information, and transmit the information to the decision-making module.

[0261] Experience Perception Learning Module: Optimize the demand response strategy of EVs by learning from historical decision-making experiences, improve the decision-making efficiency and response effect of EVs in the dynamic power-transportation-communication coupling network environment; Based on weighted experience replay and priority sampling, learn and update different experience samples according to their importance to strengthen the demand response strategy of EVs.

[0262] Decision-making Module: Based on the cumulative prospect theory, comprehensively consider factors such as electricity price, path status and task delay to generate charging and discharging decisions, path decisions and task offloading decisions.

[0263] Knowledge Transfer and Cooperative Control Module: Solve the problems of EVs in the high-speed dynamic environment, such as poor initial model performance, long learning time, and poor convergence effect, improve the EV decision-making efficiency and demand response effect, realize the strategy knowledge transfer and cooperative optimization among EVs, and quickly adapt to the new system state when the environment changes.

[0264] Task Queue Management Module: Dynamically maintain the task data queue and track the task backlog and processing delay.

[0265] Data Transmission Module: Send the EV charging and discharging behavior to the distribution network control center demand response module, including the selected charging / discharging node and power information; The transportation network status, including the travel route of the EV and the current road congestion information. Send the data task transmission demand and computing task offloading decision to the base station demand response module; The backlog of the local task queue and the task data size information. Receive the spatio-temporal guiding electricity price and the power grid node load status feedback by the distribution network control center demand response module; Receive the edge computing result and data processing delay feedback by the base station demand response module.

[0266] Energy management module: Optimize the charging and discharging operations of the battery in real time and participate in vehicle-to-grid interaction.

[0267] The base station demand response module in the communication-electricity-transportation triple-network coupling demand response system consists of Figure 4 as shown below. The introduction of each functional sub-module is as follows:

[0268] Data reception module: Receive the data task transmission requirements and computing task offloading decisions from the EV demand response module; receive the backlog volume and task data size information of the EV local task queue.

[0269] Edge server data computing module: Collaborate with the edge server to provide computing task processing services for EVs, such as autonomous driving data analysis, path planning, obstacle recognition, etc.; dynamically allocate computing resources to process the data tasks offloaded by EVs.

[0270] Result return module: Send the edge computing results and data processing delays to the EV demand response module, support the autonomous driving decision-making or demand response scheduling of EVs, ensure the real-time and reliability of result return, and reduce the delay in the communication link.

[0271] Base station energy consumption management module: Send the real-time power consumption information of the base station to the demand response module of the distribution network control center.

[0272] The demand response module of the distribution network control center in the communication-electricity-transportation triple-network coupling demand response system consists of Figure 5 as shown below. The introduction of each functional sub-module is as follows:

[0273] Power acquisition module: Collect the power data of each node and line, including information such as node voltage, current, and power flow distribution.

[0274] Power monitoring module: Monitor the power flow conditions of each node and line in the distribution network in real time, including active power and reactive power; monitor the charging and discharging power and energy consumption of FCS and base stations to ensure the stable operation of the power grid; monitor whether the voltage and current of the power grid exceed the limit, and obtain the node load distribution and line power flow data.

[0275] Electricity price calculation module: Based on the time load distribution and space load distribution characteristics of the distribution network, calculate the time-guided electricity price, space-guided electricity price, and form a spatio-temporal guided electricity price. Guide EVs to make charging and discharging decisions through electricity prices to achieve peak shaving and valley filling, balance the grid load, and improve the operation stability and economy of the power grid.

[0276] Load analysis and stability assessment module: Evaluate the load stability and load balance of the distribution network, and analyze the impact of EV charging and discharging on the operation of the power grid.

[0277] Demand Response Scheduling Module: Based on the demand response behavior and game optimization model of EVs, it coordinates the charging / discharging of EVs and the selection of travel routes to improve the overall performance of the power grid and transportation network; uses time-space electricity prices to guide EVs to perform demand response, optimizing the load stability of the power grid, network losses, and traffic congestion; combines EV feedback information (such as charging / discharging power, charging station selection, route selection, etc.) to update the electricity price strategy and dynamically regulate demand response behavior.

[0278] Data Interaction and Communication Module: Receives the real-time power consumption information of the base station transmitted from the base station demand response module; receives the charging / discharging behavior of the EV demand response module, including the selected charging / discharging node and power information; sends the time-space guiding electricity price and the load status of the power grid nodes to the EV demand response module.

[0279] 4. Method Flow

[0280] The specific steps for the implementation of the present invention are as Figure 1 shown.

[0281] Step 1 Time-Space Electricity Price Guidance: Initialize the communication-power-transportation triple-network coupling network model, including components such as EVs, FCSs, distribution networks, communication networks, and edge servers; the distribution network control center calculates the time-guiding electricity price and the space-guiding electricity price respectively according to the time and space distribution characteristics of the power grid load, synthesizes the time-space guiding electricity price, and publishes it to each EV.

[0282] Step 2 EV Demand Response Action Decision and Execution: The EV agent determines the charging / discharging strategy and route selection through the policy network according to the received time-space guiding electricity price and its own state, and executes the corresponding actions, including selecting the charging / discharging time period and charging station; after executing the action, the EV agent observes the new state, and forms an experience sample based on the obtained immediate reward and the new state for subsequent model training and optimization.

[0283] Step 3 Weighted Experience Perception Learning: Extract samples from the experience replay pool based on the priority of the experience samples, calculate the dynamic weights of the samples, and then update the evaluation network (critic network) and the policy network (actor network) to improve the response efficiency.

[0284] Step 4 Migration and Collaboration: When the EV agent detects that the environmental change causes a significant decrease in the reward function, trigger the knowledge migration and collaboration mechanism to perform the migration of demand response strategy knowledge to adapt to the environmental change and improve the learning efficiency; through the above steps, realize the joint optimization among the EV demand response strategy, the power grid, and the transportation network, improve the load stability and balance of the power grid, reduce the road congestion degree of the transportation network, and reduce the user's road travel time.

[0285] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for collaborative scheduling of communication, power, and transportation coupled network resources, characterized in that: The specific steps are: Firstly, a communication-power-transportation coupling network model is constructed, including a communication network-power-transportation coupling network architecture, a transportation network model including FCS, a distribution network model including FCS, and a communication network model. The EV limited rationality demand response model is constructed by fully considering the heterogeneity of EV users. Analyze the grid load characteristics from the time and space dimensions, and on this basis, construct the EV demand response time and space guided electricity price, and propose a time and space guided electricity price model; By optimizing EV travel paths, charging and discharging nodes, and charging and discharging power, the objective function of the optimization problem is constructed by minimizing EV road travel time, demand response benefits, travel anxiety, grid load stability, balance, and weighted sum of network losses; Based on the above model and the objective function of the optimization problem, the EV demand response is modeled as a partially observable Markov cooperative game model. The EV demand response scheduling algorithm based on spatiotemporal electricity price guidance and migration coordination is used to solve the partially observable Markov cooperative game model to achieve comprehensive optimization of the communication-power-transportation coupled network.

2. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The specific communication network - power - transportation coupled network architecture is as follows: It is defined that there are a total of M EVs and N edge servers; the sets of EVs and edge servers are respectively denoted as and The total optimization duration is divided into consecutive T, indexed by t = 1, 2, …, T; the duration of one time period is denoted as τ; the route selection variable of the EV is defined as μ m,a (t), where μ m,a (t) = 1 indicates that EV m selects route a to travel in the t-th time period; the FCS selection variable of the EV is defined as and indicating that EV m selects FCS q for charging, indicating that EV m selects FCS q for discharging; the autonomous driving data offloading decision variable of the EV is defined as y m,n (t), y m,n (t) = 1 indicates that EV m offloads its autonomous driving data to edge server n for processing through the base station closest to it.

3. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The traffic network model containing FCS is defined as an undirected graph G T =(N T , A T , Q), where N T represents the set of road network nodes; A T represents the set of road segments, and the road segment a ∈ A T ; Q represents the set of FCSs in the road network, and the FCS q ∈ Q; The semi-dynamic traffic flow allocation model is specifically expressed as: where, Ω od is the set of EV travel demands; K od is the set of paths to meet the travel demand od; is the traffic flow of travel demand od on path k; δ k,a (t) is a binary variable representing the relationship between the path and the road section. δ k,a (t) = 1 indicates that road section a is on path k. Conversely, δ k,a (t) = 0; is the residual flow of demand od, τ is the time period length, r od , r od,mod are the original travel demand flow and the travel demand flow after considering the residual flow correction, respectively; τ a (t), and are the travel time of road section a and the charging and discharging queuing time of FCS q, respectively.

4. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The distribution network model containing FCS is represented as an undirected graph G P =(N P , A P , Q), where N P represents the set of distribution network nodes; A P represents the set of distribution network line branches, and branch ij ∈ A P ; The distribution network power flow model is specifically expressed as: where P i,j (t) and Q i,j (t) are the active power and reactive power flowing from node i to node j on branch ij respectively, X i,j is the reactance of branch ij, and v i (t) is the voltage of node i; Distribution network power loss C DN,loss is expressed as: Where C loss (t) is the unit cost of network loss, and R i,j is the resistance of branch ij, and I i,j (t) is the current of branch ij.

5. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The communication network model is specifically as follows: the total power consumption of the base station is expressed as the following formula (8), including data processing power consumption and cooling power consumption Among them, the edge server data processing power consumption is expressed as: Wherein, P peak , P ideal are respectively the peak power and standby power of a single server of the edge server, κ m is the average data service rate of a single server, which is related to the operating frequency of the server CPU, N server is the number of front-end processors working in the data center, H n (t) is the task data processed by the data center; The operating power of the edge server cooling equipment is expressed as: In the formula, is the air supply temperature of the refrigeration equipment, and ξ n,2 , ξ n,1 and ξ n,0 are the refrigeration coefficients respectively; The queuing delay of the task queue on the EV side of the communication network is: where Q m (t) represents the task queue backlog of EV m, K m (t), and α m represent the number of tasks and the size of a single task respectively; The queuing delay of the task queue on the edge server side is:

6. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The specific EV bounded rationality demand response model is as follows: The costs mainly considered in EV demand response include four aspects: the demand response cost C m,1 (t) on the power grid side; the travel time cost C m,2 (t) and travel anxiety cost C m,3 (t) on the transportation network side; the service data processing time cost C m,4 (t) on the communication network side; which are respectively expressed as: where μ m,a (t) is the route selection indicator variable of EV m, and μ m,a (t) = 1 indicates that EV m has selected route a; respectively represent the charge and discharge indicator variables of EV m at FCS q, represents charging, represents discharging; are the spatio-temporal guiding charge and discharge prices respectively; are the travel time anxiety coefficient and the SoC anxiety coefficient respectively; are the expected arrival time and the expected SoC respectively; is the destination arrival indicator variable, indicates not arriving at the destination; Based on the positive and negative expectations of EV as reference points, the positive and negative prospect values ​​of EV are: Where, C m,z (t) (z = 1, 2, 3, 4) respectively represent the demand response cost, travel time cost, travel anxiety cost, and data processing time cost of EV users; v + [C m,z (t)], v - [C m,z (t)] are respectively the positive and negative prospect values of the EV based on its positive and negative expectations as the benchmark; The comprehensive prospect value taking into account the bounded rationality of EV is expressed as: where, π + [C m,z (t)], π - [C m,z (t)] are the weights of the EV facing gains and losses respectively.

7. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The time-space guided electricity price model is specifically as follows: the load time distribution characteristic is expressed as: The time-guided electricity price is expressed as: where μ time is the sensitivity coefficient of time-of-use electricity price to the time distribution characteristics of distribution network load; The load distribution characteristics of the distribution network FCS q are expressed as: The space guidance electricity price is expressed as: where η space is the sensitivity coefficient of the spatially-guided electricity price to the load distribution characteristics, and r-q represents the base station located at the same distribution network node as FCS q; The time-space guided electricity price is expressed as: In the formula, are the basic charging and discharging prices obtained according to the power flow tracing method, respectively.

8. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The objective function of the optimization problem is where μ = {μ m,a (t)}, C1 is the EV section selection constraint; C2 is the EV server selection variable constraint; C3 is the EV charging station selection constraint; C4 is the charge and discharge state constraint; C5 - C6 are the upper and lower limits of the EV charge and discharge power, where, and are the upper and lower limits of the charge and discharge power of EV m respectively; C7 - C8 are the distribution network node voltage and safety current constraints respectively, where v i,min and v i,max are the minimum and maximum values allowed for the voltage of node i under the safe condition of the distribution network respectively, and I i,j,min and I i,j,max are the minimum and maximum values allowed for the current of branch ij respectively; C9 is the EV end - to - end delay probability constraint, where, is the upper limit of the EV end - to - end delay, and τ P is the communication reliability threshold, that is, the probability that the EV end - to - end delay is lower than the upper limit should be higher than τ P .

9. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The partially observable Markov cooperative game model is: Agent State and Action: The state S of agent m m (t) and action a m (t) are respectively: In the formula, the state space includes the spatiotemporal guided charging and discharging electricity prices of each FCS, the average charging and discharging queue delay, the free travel time of each road section, the SoC of EV, the target arrival time, the target SoC, the local data storage queue backlog, and the data processing queue backlog information of the edge server; the action space includes the road section selection variable, the charging and discharging selection variable, the charging and discharging power selection variable, and the automatic driving task offloading server selection indicator variable; Global reward for cooperative game of intelligent agents: The global reward for cooperative game of intelligent agents is defined based on network loss, grid load stability and balance, and the safety constraint of distribution network is defined in the form of penalty function, which is expressed as Where, V U and V I are the penalty coefficients for the over-limit of the grid node voltage and the over-load of the branch current, respectively; Agent aggregate reward: Set the global optimization target as the global reward, convert the constraints into penalty items and add them to the global reward. The VCG cost of agent m is expressed as: In the formula, represents the global reward when the agent m does not participate in demand response; The aggregated reward r of agent m m (t) is expressed as Where, V C represents the weight of the quality of experience of electric vehicles, and V tau represents the weight of exceeding the time limit for processing autonomous driving task data.

10. A method for collaborative scheduling of communication, power, and transportation coupled network resources according to claim 1, characterized in that: The EV demand response scheduling algorithm based on spatiotemporal electricity price guidance and migration coordination includes four steps: spatiotemporal electricity price guidance, demand response action decision and execution, empowered experience perception learning and migration coordination: Step 1: Time and space electricity price guidance The distribution network control center first calculates the time distribution characteristics and spatial distribution characteristics of the grid load respectively. Secondly, it calculates the time-guided electricity price and the spatial-guided electricity price. Finally, it calculates the spatio-temporal guided electricity price and publishes this electricity price to each EV; Step 2: Action decision and execution Based on the published spatio-temporal guiding electricity price, EV agent m inputs the state S m (t) into the policy network π m to obtain the policy π m (·|S m (t)), and selects the action a based on this policy m ; Secondly, the EV executes the action to obtain the immediate reward r m , observes the state S m (t + 1) at the next moment, and forms an experience sample Then, calculate the TD-error of the two critic evaluation networks, expressed as: wherein, is the discount factor, is the action selected according to the policy π m (·|S m (t + 1)); set the sample TD-error decision threshold TD m,max , when the empirical sample satisfies the following conditions, discard the sample and do not put it into the empirical replay pool; Otherwise, it is put into the experience replay pool for subsequent model training. Assuming that this experience sample is the h-th sample in the experience replay pool, then the priority of this sample can be expressed as: In the formula, v is the experience sample priority supplement coefficient, which is used to avoid samples with too small TD-error being ignored; Step 3: Weighted experience perception learning According to the priority of the samples in the experience replay pool, the probability of each sample being drawn can be obtained as where H is the capacity of the experience replay pool, is the prioritized sampling exponent; Sample a group of empirical samples from the experience replay pool based on the probability of the empirical samples being drawn and calculate the loss function of the critic network as where ε h (t) is the dynamic weight of the empirical sample, which is used to balance the calculation deviation of the loss function caused by introducing the extraction probability of the empirical sample; Based on the loss function, the two critic networks are updated using the stochastic gradient descent method, and the two target critic networks are updated regularly using the soft update method, which is specifically expressed as where λ m is the gradient descent step size; κ m is the soft update coefficient; The loss function of the EV intelligent agent m actor policy network can be characterized by the Kullback-Leibler divergence between the policy output by the policy network and the Q value, which is specifically expressed as The actor network is updated using the stochastic gradient descent method, and the entropy regularization temperature coefficient is updated adaptively, which is expressed as: where dim(a m (t)) is the action dimension of agent m; Step 4: Knowledge transfer and collaboration When the reduction of the EV agent reward function exceeds the set threshold, that is When it is determined that a large change has occurred in the EV environment, the migration cooperation mechanism is triggered; first, the EV that triggers the migration cooperation mechanism sends a model migration request to other EVs within its communication coverage range, and the EV that receives the request transmits its own model to the EV that sends the request; second, the EV that sends the request puts the received model into the set Extract a set of state sampling sets S samp ={s1, s2, …, s D}, and respectively according to Obtain the action set corresponding to the state sampling set according to the model to be migrated in Then, the corresponding reward set can be obtained according to the action set Finally, calculate the cumulative reward of each model to be migrated, and select the model to be migrated with the optimal performance according to the cumulative reward, which is specifically expressed as: Wherein, is the cumulative reward of the model; b * is the optimal model to be migrated in ; r0 is the cumulative reward of the EV that sends the migration request for its own model. When μ > 1, it indicates that using the optimal model to be migrated can obtain higher rewards in the current environment, and the model to be migrated can be migrated to its own agent; otherwise, migration is not suitable.

11. A communication, power, and transportation coupled network resource collaborative scheduling system, characterized in that, Including: Distribution network control center demand response module: used to calculate and publish the spatio-temporal guided electricity price by monitoring the spatio-temporal distribution characteristics of the grid load, guide EVs to reasonably select charging and discharging behaviors in the time and space dimensions, and optimize the grid load distribution; based on the grid load status, EV charging and discharging power feedback, and base station power consumption feedback, schedule the power flow of the distribution network to reduce the peak load and grid losses; Base station demand response module: used to receive the EV computing task offloading, provide low-latency and high-reliability communication services for EVs, cooperate with the edge server to perform data offloading and computing processing of the EV's autonomous driving tasks, ensure the minimum backlog of the task queue, meet the latency requirements, and complete the transmission of the computed data back; Monitor and calculate its own communication data processing power consumption and cooling power consumption, and feedback the power consumption data to the distribution network control center demand response module for power grid scheduling and energy management; EV demand response module: used to determine the computing offloading target of the autonomous driving data according to the service quality of the base station and the edge server, ensure low-latency data processing, and feedback the autonomous driving task offloading decision to the base station demand response module; under the guidance of the spatio-temporal guided electricity price published by the distribution network control center, select the charging and discharging time periods and charging and discharging power, and feedback the charging and discharging decision information to the distribution network control center demand response module to participate in the power grid demand response, realize peak shaving and valley filling, and balance the grid load; according to the congestion situation of the transportation network and the charging demand, select the optimal travel route and charging station to relieve traffic congestion and optimize the charging efficiency.

12. A communication, power, and transportation coupled network resource collaborative scheduling system according to claim 11, characterized in that: The said EV demand response module Includes the following sub-modules: Information acquisition module: used to collect real-time information on the EV battery status, charging station status, grid and transportation network path status, and transmit the information to the decision-making module; Experience Perception Learning Module: It is used to optimize the demand response strategy of EVs by learning from historical decision-making experiences, improve the decision-making efficiency and response effect of EVs in the dynamic power-transportation-communication coupling network environment; based on weighted experience replay and priority sampling, learn and update different experience samples according to their importance, and strengthen the demand response strategy of EVs; Decision-making Module: It is used to generate charging and discharging decisions, path decisions and task offloading decisions based on the cumulative prospect theory, considering factors such as electricity price, path status and task delay; Knowledge Transfer and Cooperative Control Module: It is used to solve problems such as poor initial model performance, long learning time and poor convergence effect of EVs in a highly dynamic environment, improve the decision-making efficiency and demand response effect of EVs, realize the transfer of policy knowledge and cooperative optimization among EVs, and quickly adapt to the new system state when the environment changes; Task Queue Management Module: It is used to dynamically maintain the task data queue and track the task backlog and processing delay; Data Transmission Module: It is used to send the charging and discharging behaviors of EVs to the demand response module of the distribution network control center, including the selected charging / discharging nodes and power information; the traffic network status, including the travel path of the EV and the current road congestion information; Send the data task transmission requirements and computing task offloading decisions to the base station demand response module; the backlog of the local task queue and the size information of the task data; receive the spatio-temporal guiding electricity price and the grid node load status fed back by the demand response module of the distribution network control center; receive the edge computing results and data processing delay fed back by the base station demand response module.

13. A communication, power, and transportation coupled network resource collaborative scheduling system according to claim 11, characterized in that: The base station demand response module includes the following sub-modules: Data Reception Module: It is used to receive the data task transmission requirements and computing task offloading decisions from the EV demand response module; receive the backlog of the EV local task queue and the size information of the task data; Edge Server Data Computing Module: It is used to provide computing task processing services for EVs by collaborating with edge servers, such as autonomous driving data analysis, path planning, obstacle recognition, etc.; dynamically allocate computing resources to process the data tasks offloaded by EVs; Result Return Module: It is used to send the edge computing results and data processing delay to the EV demand response module, support the autonomous driving decision-making or demand response scheduling of EVs, ensure the real-time and reliability of the result return, and reduce the delay in the communication link; Base Station Energy Consumption Management Module: It is used to send the real-time power consumption information of the base station to the demand response module of the distribution network control center.

14. A communication, power, and transportation coupled network resource collaborative scheduling system according to claim 11, characterized in that: The demand response module of the distribution network control center includes the following sub-modules: Power Acquisition Module: It is used to collect the power data of each node and line, including information such as node voltage, current, and power flow distribution; Power Monitoring Module: It is used to monitor the power flow conditions of each node and line in the distribution network in real time, including active power and reactive power; monitor the charging and discharging power and energy consumption of FCS and base stations to ensure the stable operation of the power grid; monitor whether the voltage and current of the power grid exceed the limit, and obtain the node load distribution and line power flow data; Electricity price calculation module: It is used to calculate the time-guided electricity price, space-guided electricity price based on the time load distribution and space load distribution characteristics of the distribution network, and form the spatio-temporal guided electricity price; guide EVs to make charge and discharge decisions through electricity prices, achieve peak shaving and valley filling, balance the grid load, and improve the operation stability and economy of the grid; Load analysis and stability assessment module: It is used to evaluate the load stability and load balance of the distribution network, and analyze the impact of EV charge and discharge on grid operation; Demand response scheduling module: It is used to coordinate the charge and discharge of EVs and the selection of travel routes based on the demand response behavior of EVs and the game optimization model, and improve the overall performance of the power grid and transportation network; use spatio-temporal electricity prices to guide EVs to perform demand response, optimize the grid load stability, line loss and traffic congestion; update the electricity price strategy in combination with EV feedback information (such as charge and discharge power, charging station selection, route selection, etc.) to dynamically regulate the demand response behavior; Data interaction and communication module: It is used to receive the real-time power consumption information of the base station transmitted by the base station demand response module; Receive the charge and discharge behavior from the EV demand response module, including the selected charging / discharging node and power information; send the spatio-temporal guided electricity price and the load status of the grid node to the EV demand response module.

Images