Cooperative control method and system for main and auxiliary micro multi-level vehicle network interaction and medium

Abstract

The present application relates to the technical field of power system regulation, and more particularly to a coordinated regulation method and system for main-distribution-micro multi-level vehicle-grid interaction, and a medium, the method comprising: constructing an electric vehicle aggregation model corresponding to a day-ahead time scale and an intra-day time scale according to historical operation data, real-time access state and state of charge of multiple types of electric vehicles; constructing a main-distribution-micro multi-level electric vehicle aggregation body coordinated regulation model based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, the distribution grid, the micro grid and the charging station; generating coordinated regulation instructions for each level based on the coordinated regulation model, and distributing charging and discharging power to electric vehicles connected in the charging station to achieve orderly charging and discharging regulation. Through the present application, the problem that the prior art cannot perform multi-time scale coordinated aggregation and real-time orderly regulation of multiple types of electric vehicles in the main-distribution-micro multi-level vehicle-grid interaction scenario is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a coordinated control method, system and medium for multi-level vehicle-grid interaction of main, distribution and micro-level systems. Background Technology

[0002] With the large-scale grid connection of new energy sources and the continuous advancement of transportation electrification, electric vehicles are not only an important component of transportation load but are also gradually becoming a distributed and flexible resource that can participate in grid regulation. Through orderly charging, discharging, and vehicle-grid interaction, they can play a role in peak shaving and valley filling, promoting the consumption of renewable energy, and improving the flexibility of system operation. Due to the characteristics of electric vehicles, such as small individual capacity, large number, dispersed distribution, and rapid changes in access status, their participation in grid regulation usually requires the reliance on aggregation modeling and hierarchical organization to gradually aggregate dispersed vehicles into dispatchable resource units and coordinate with the operational needs of different levels such as the main grid, distribution network, microgrid, and charging stations.

[0003] In existing technologies, most research on electric vehicles' participation in power grid regulation focuses on single-level optimization scheduling or single-time-scale control. Different types of electric vehicles are typically treated as homogeneous resources, failing to fully consider the significant differences in operating patterns, access behaviors, spatial distribution, response capabilities, and user needs among electric buses, electric taxis, and electric private cars. Furthermore, there is a lack of systematic coordination mechanisms for multi-level vehicle-grid interaction scenarios involving main, distribution, and micro-level systems. Simultaneously, existing solutions often struggle to balance the connection between different time scales, including day-ahead and intraday, and lack effective support for the layer-by-layer decomposition, hierarchical transmission, and end-point execution of regulation commands. When facing high-dimensional dynamic optimization problems under multi-level, multi-type, and multi-time-scale coupling conditions, they are prone to insufficient real-time performance, weak adaptability, and limited regulation effectiveness.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a collaborative control method, system, and medium for multi-level vehicle-to-grid interaction between main, auxiliary, and micro-level systems, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A collaborative control method for multi-level vehicle-to-grid (V2G) interaction, comprising: Based on the historical operating data, real-time access status and state of charge of various types of electric vehicles, an electric vehicle aggregation model corresponding to the day-ahead time scale and intraday time scale is constructed to obtain the adjustable power and adjustable capacity of various types of electric vehicles. Based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging station, a multi-level coordinated control model of the main grid, distribution network and microgrid is constructed to obtain the control constraint relationship between the levels. Based on the aforementioned multi-level electric vehicle aggregation collaborative control model, collaborative control commands for each level are generated according to the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. According to the coordinated control command, the adjustable power, and the adjustable capacity, the charging and discharging power is allocated to the electric vehicles connected to the charging station to achieve orderly charging and discharging control under vehicle-to-grid interaction.

[0007] Furthermore, an electric vehicle aggregation model corresponding to the day-ahead and intraday time scales is constructed, including: Based on the battery capacity, charging and discharging power, connection status and state of charge of various types of electric vehicles, a single-vehicle adjustable capacity model is established to determine the adjustable capacity of various types of electric vehicles in the target time period. Based on the adjustable capacity, the number of electric vehicles of each type, the maximum charging power, and the reliability coefficient, a type aggregation power model is established to determine the upper limit of the total adjustable power of each type of electric vehicle during the target period. Based on the historical access data sequences and time characteristics of various types of electric vehicles, an access probability prediction model is established to determine the access probability of each type of electric vehicle under the stated day-ahead time scale and the stated intraday time scale.

[0008] Furthermore, the spatial distribution and temporal type of the access probability are corrected, including: The control area is spatially divided based on the location distribution information of various types of electric vehicles, and the aggregate power of various types of electric vehicles in each spatial area is determined based on the location of each vehicle and the actual control power. The corresponding weekday or holiday time type is determined based on the time feature information, and the access probability and the reliability coefficient are corrected according to the time type. Based on the corrected access probability and the reliability coefficient, the aggregation characteristics of multiple types of electric vehicles in each spatial region under different time types are determined.

[0009] Furthermore, constructing an electric vehicle aggregation model corresponding to the day-ahead and intraday timescales also includes: Based on the access probability of various types of electric vehicles in each hour, an hourly aggregated power model is established at the day-ahead time scale to determine the day-ahead predicted aggregated power of various types of electric vehicles. Based on the real-time access status and actual adjustable power of various electric vehicles at each minute, a minute-level aggregated power model is established at the intraday time scale to determine the intraday real-time aggregated power of various electric vehicles. Based on the time cycle characteristics and response characteristics of various types of electric vehicles, electric buses are identified as the basic control resource under the day-ahead time scale, and electric taxis and electric private cars are identified as supplementary control resources under the day-ahead time scale and flexible control resources under the intraday time scale.

[0010] Furthermore, a multi-level coordinated control model for electric vehicle aggregates (main, auxiliary, and micro levels) is constructed, including: Based on the hierarchical topology between the main network layer, distribution network layer, microgrid layer and charging station layer, establish the hierarchical relationships between each layer; Based on the aforementioned hierarchical relationship, multiple types of electric vehicles are aggregated layer by layer from the charging station layer to the microgrid layer, from the microgrid layer to the distribution network layer, and from the distribution network layer to the main grid layer, in order to form electric vehicle aggregates at each level; Based on the aggregation results of electric vehicle aggregates at each level, the power external characteristics, capacity external characteristics, response speed external characteristics, and reliability external characteristics of electric vehicle aggregates at each level are determined, and power balance constraints, capacity constraints, and response speed constraints between levels are established.

[0011] Furthermore, the construction of a multi-level coordinated control model for electric vehicle aggregates, including main, auxiliary, and micro-level systems, also includes: Based on the number of electric vehicle aggregates connected at each level, the state of charge distribution, time characteristics, and external input information, a prediction model for the external characteristics of electric vehicle aggregates at each level is established. The external characteristics of electric vehicle aggregates at each level under the daytime scale are predicted based on the external characteristic prediction model to obtain the daytime external characteristic prediction results. Based on real-time measurement information and the daytime external characteristic prediction results, the external characteristics of electric vehicle aggregates at each level under the intraday time scale are corrected for errors to obtain intraday real-time external characteristic update results.

[0012] Furthermore, coordinated control instructions are generated at each level, including: The state space of the main grid layer intelligent agent is constructed based on the load power, renewable energy output power, electricity price and historical status information of the main grid layer, and the action space of the main grid layer intelligent agent is constructed based on the control power instructions issued by the main grid layer to each distribution network area. The state space of the distribution network layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate in the distribution network layer, the control power command issued by the main grid layer, and the node voltage. The action space of the distribution network layer intelligent agent is constructed based on the control power command issued by the distribution network layer to each microgrid. The state space of the microgrid layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate and the local load power of the microgrid, and the action space of the microgrid layer intelligent agent is constructed based on the power control instructions issued by the microgrid layer to each charging station. Based on the state and action spaces of the main network layer agent, the distribution network layer agent, and the microgrid layer agent, a multi-agent reinforcement learning algorithm is used to generate the coordinated control instructions that are transmitted step by step from the main network layer to the distribution network layer, from the distribution network layer to the microgrid layer, and from the microgrid layer to the charging station layer.

[0013] Furthermore, the charging and discharging power is allocated to the electric vehicles connected to the charging station, including: Based on the coordinated control command received by the charging station layer, the target control power of the charging station layer in the current time period is determined; Based on the target control power and the requested power of the connected vehicles, the electric vehicles connected to the charging station layer are proportionally allocated to determine the actual charging and discharging power of each connected vehicle. The actual charging and discharging power is adjusted according to the charging state priority of the access vehicle, and the charging state of the access vehicle is updated according to the adjusted actual charging and discharging power to meet the off-site charging state requirements of the access vehicle.

[0014] A collaborative control system for multi-level vehicle-to-grid (V2G) interaction, comprising: The time-scale aggregation module constructs electric vehicle aggregation models corresponding to the day-ahead and intraday time scales based on historical operating data, real-time access status, and state of charge of various types of electric vehicles, thereby obtaining the adjustable power and adjustable capacity of various types of electric vehicles. The power grid hierarchical aggregation module constructs a multi-level coordinated control model for electric vehicle aggregation based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging stations, and obtains the control constraint relationship between the levels; The hierarchical collaborative control module, based on the multi-level collaborative control model of electric vehicle aggregates (main, distribution, and micro), generates collaborative control commands for each level according to the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. The charging and discharging power distribution module allocates charging and discharging power to electric vehicles connected to the charging station based on coordinated control instructions and adjustable power and capacity, so as to achieve orderly charging and discharging control under vehicle-to-grid interaction.

[0015] A computer-readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, can implement the aforementioned collaborative control method for multi-level vehicle-to-grid interaction oriented towards primary, secondary, and micro-level systems.

[0016] The technical solution of this invention can achieve the following technical effects: Based on the operation and status data of various types of electric vehicles, a multi-timescale aggregation model covering the day-ahead and intraday periods is constructed. Then, a collaborative control model and its constraint relationships are constructed by aggregating the main grid, distribution network, microgrid, and charging stations layer by layer. Subsequently, collaborative control instructions are generated according to the hierarchical transmission path from the main grid to the charging station, and the charging and discharging power allocation is completed at the charging station. This achieves orderly collaborative control under the multi-level vehicle-grid interaction of the main grid, distribution network, and microgrid, effectively solving the problem that existing technologies are difficult to use for multi-timescale collaborative aggregation and real-time orderly control of various types of electric vehicles in the multi-level vehicle-grid interaction scenario.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a collaborative control method for multi-level vehicle-to-grid interaction between main, auxiliary, and micro-level systems; Figure 2 The diagram shows the charging load distribution characteristics of different types of electric vehicles on a typical weekday over 24 hours and the control effect on multiple time scales. Figure 3 This diagram illustrates the performance of different methods in terms of multi-level synergistic regulation effects. Figure 4 The image shows the effect of multi-objective collaborative optimization based on the SAC reinforcement learning algorithm. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1; like Figure 1 As shown, this application provides a collaborative control method for multi-level vehicle-to-grid interaction, including: S10: Based on the historical operating data, real-time access status and state of charge of various types of electric vehicles, construct an electric vehicle aggregation model corresponding to the daytime and intraday time scales to obtain the adjustable power and adjustable capacity of various types of electric vehicles. S20: Based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging station, a multi-level coordinated control model of the electric vehicle aggregation system is constructed to obtain the control constraint relationship between the levels. S30: Based on the multi-level coordinated control model of electric vehicle aggregates, the coordinated control instructions of each level are generated according to the transmission path from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. S40: Based on the coordinated control instructions and adjustable power and capacity, it allocates charging and discharging power to electric vehicles connected to the charging station to achieve orderly charging and discharging control under vehicle-grid interaction.

[0023] Specifically, this solution is applied to a multi-level vehicle-to-grid interaction scenario consisting of the main grid, distribution grid, microgrid, and charging stations, and treats electric buses, electric taxis, and electric private cars connected to the system as controllable objects. First, historical operating data, real-time access status, state of charge, location distribution information, and corresponding time characteristics of various types of electric vehicles are collected. Based on the differences in charging behavior, operating patterns, and response capabilities of different types of electric vehicles, an electric vehicle aggregation model covering the day-ahead and intraday time scales is constructed. At the individual vehicle level, the adjustable capacity corresponding to battery capacity, charging and discharging power, access status, and state of charge is determined. At the type level, the adjustable power upper limit of each type of electric vehicle within the target time period is determined. The access probability of each type of electric vehicle at different time scales is determined by combining historical access data and time characteristic information to obtain the adjustable power and adjustable capacity of each type of electric vehicle. To improve the modeling accuracy, a long short-term memory neural network can be used in this embodiment to predict the access probability, and the aggregation characteristics of each type of electric vehicle can be corrected by combining weekdays, holidays, and spatial regional distribution, thereby forming an electric vehicle aggregation model for multiple types and multiple time scales. After completing the construction of the electric vehicle aggregation model, a multi-level coordinated control model for electric vehicle aggregations is constructed based on the hierarchical topology relationship between the main grid, distribution network, microgrid, and charging stations. Specifically, the hierarchical relationship between the main grid layer, distribution network layer, microgrid layer, and charging station layer is first established. Then, the various types of electric vehicle resources in the charging station layer are aggregated to the microgrid layer, from the microgrid layer to the distribution network layer, and from the distribution network layer to the main grid layer to form electric vehicle aggregations at each level. Subsequently, the power external characteristics, capacity external characteristics, response speed external characteristics, and reliability external characteristics of each level of electric vehicle aggregation are determined, and power balance constraints, capacity constraints, and response speed constraints between levels are established accordingly to obtain the control constraint relationship between levels. In this embodiment, a deep neural network can also be used to quickly predict the external characteristics of each level of aggregation, and Kalman filtering can be used to correct the prediction error during the intraday real-time operation to improve the prediction accuracy and real-time update capability of the external characteristics of multi-level aggregations. After establishing a multi-level coordinated control model for electric vehicle aggregates (main, distribution, and microgrid layers), coordinated control instructions for each level are generated based on the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. Specifically, the main grid layer can be set as the central agent, and the distribution grid layer and microgrid layer can be set as local agents. The state space and action space corresponding to each level of agent are constructed respectively. The main grid layer agent generates control instructions for the distribution grid layer by comprehensively considering load power, renewable energy output power, electricity price, and historical state information. The distribution grid layer agent generates control instructions for the microgrid layer based on the instructions issued by the main grid layer, the external characteristics of the electric vehicle aggregate, and the node voltage. The microgrid layer agent generates control instructions for the charging station layer based on the local load power and the external characteristics of the aggregate. On this basis, a multi-agent deep reinforcement learning algorithm can be used to train the agents at each level collaboratively. The Soft Actor-Critic algorithm is preferred to realize the control decision in the continuous action space, so that the generated coordinated control instructions take into account the requirements of economy, balance, and reliability. After receiving the coordinated control command at the charging station level, the charging and discharging power of the electric vehicles connected to the charging station is allocated according to the coordinated control command and the aforementioned adjustable power and adjustable capacity, so as to realize orderly charging and discharging control under vehicle-to-grid interaction. Specifically, the target control power of the charging station for the current time period is first determined, and then the power requested by the vehicles connected to the station is allocated proportionally to obtain the actual charging and discharging power of each connected vehicle. The allocation result is further corrected by combining the priority of the state of charge of each connected vehicle, and the state of charge of each connected vehicle is updated according to the actual charging and discharging power to ensure that the vehicle meets the minimum state of charge requirement set by the user when leaving the station. The above scheme enables hierarchical and coordinated control among the main grid, distribution grid, microgrid and charging stations, and fully leverages the fundamental role of electric buses in daytime dispatching and the flexible role of electric taxis and electric private cars in intraday regulation, thereby achieving orderly and coordinated regulation of multiple types of electric vehicles at multiple time scales.

[0024] The technical solution of this invention constructs a multi-timescale aggregation model covering the day and day based on the operation and status data of multiple types of electric vehicles. Then, it constructs a collaborative control model and its constraint relationships that aggregates the main network, distribution network, microgrid, and charging station layer by layer. Subsequently, it generates collaborative control instructions according to the hierarchical transmission path from the main network to the charging station and completes the charging and discharging power allocation at the charging station. This achieves orderly collaborative control under the interaction of the main network, distribution network, and microgrid, effectively solving the problem that existing technologies are unable to perform multi-timescale collaborative aggregation and real-time orderly control of multiple types of electric vehicles in the scenario of multi-level vehicle-network interaction.

[0025] Furthermore, constructing an electric vehicle aggregation model corresponding to the day-ahead and intraday time scales includes: Based on the battery capacity, charging and discharging power, connection status and state of charge of various types of electric vehicles, a single-vehicle adjustable capacity model is established to determine the adjustable capacity of various types of electric vehicles in the target time period. Based on adjustable capacity, the number of electric vehicles of various types, maximum charging power and reliability coefficient, a type aggregated power model is established to determine the upper limit of the total adjustable power of various types of electric vehicles in the target period. Based on the historical access data sequences and time characteristics of various types of electric vehicles, an access probability prediction model is established to determine the access probability of various types of electric vehicles at the day-ahead and intraday time scales.

[0026] As a preferred embodiment of the above, a multi-type electric vehicle aggregation model oriented towards multiple time scales is established. Furthermore, various types of electric vehicles include electric taxis, electric private cars, electric buses, and other different models, and these models differ significantly in terms of time cycle characteristics, charging behavior patterns, and control response capabilities.

[0027] Furthermore, a vehicle-to-grid (V2G) capacity model for electric vehicles is established, taking into account factors such as battery capacity, charging and discharging power, and access status; Electric vehicles vehicle during the time period The adjustable capacity is: ; in, For the first Electric vehicles vehicle during the time period Adjustable capacity (kWh); For charging efficiency; For the first Class 1 Vehicle battery capacity (kWh); This represents the upper limit of the state of charge. For time period The current state of charge; For access status indicator variables (0 or 1); Furthermore, the first electric vehicles during the period The maximum total adjustable power is: ; in, For the first electric vehicles during the period Total adjustable power limit (kW); For the first Number of electric vehicles; For the first Class 1 The vehicle's maximum charging power (kW); The reliability coefficient (between 0 and 1) represents the degree of reliability in responding to control commands. Furthermore, a time distribution prediction model for electric vehicles based on historical data and artificial intelligence algorithms is established. A Long Short-Term Memory (LSTM) neural network is used to capture temporal patterns and predict the time distribution of electric vehicles. electric vehicles during the period Access probability: ; in, For the first electric vehicles during the period The probability of access; This refers to the access data sequence for historical time periods; This is a time feature vector (including hours, days of the week, holidays, etc.); For LSTM network parameters; The length of the history window; Furthermore, electric buses exhibit a clear hourly periodicity, with their temporal distribution displaying a bimodal pattern. The access probability of electric buses can be modeled as follows: ; in, The probability of electric buses being connected on weekdays is a small random fluctuation term, indicating a relatively high reliability coefficient. Holidays ; Furthermore, electric taxis operate for extended periods, around the clock, but have a higher probability of charging during shift changes (06:00-08:00 in the morning and 18:00-20:00 in the evening). Their access probability is modeled as follows: ; in, The term represents random fluctuations; the reliability of electric taxis is moderate; and the probability of electric taxis joining the service is [not specified]. ; Furthermore, the charging time for electric private vehicles is mainly concentrated at night (19:00-07:00 the next day), but user behavior is highly random, and its access probability model is as follows: ; in, The probability of electric private vehicles connecting to the network on weekdays is relatively low due to the large random fluctuations in the data. ,weekend .

[0028] Furthermore, spatial distribution and temporal type corrections are applied to the access probability, including: The control area is spatially divided based on the location distribution information of various types of electric vehicles, and the aggregate power of various types of electric vehicles in each spatial area is determined based on the location of each vehicle and the actual control power. The corresponding weekday or holiday time type is determined based on the time characteristic information, and the access probability and reliability coefficient are corrected according to the time type. Based on the corrected access probability and reliability coefficient, the aggregation characteristics of multiple types of electric vehicles in each spatial region under different time types are determined.

[0029] As a preferred embodiment of the above, a further step is to establish an electric vehicle aggregation model that considers spatial distribution, and divide it into... The spatial region, the first Region 1 electric vehicles during the period The adjustable power is: ; in, For the first Region 1 electric vehicles during the period Adjustable power (kW); For the first Region 1 A collection of electric vehicles; For the first vehicle during the time period The actual control power; Assign a position coefficient (0 or 1) to represent the position. Is the vehicle located at the ? area; Furthermore, a graph neural network (GNN-Attention) based on an attention mechanism is used to model the spatial distribution features: ; in, For the first Region 1 electric vehicles during the period Aggregated feature vectors; For the first The topological structure of the region; The node feature matrix; The edge feature matrix is ​​used; the attention mechanism weights are calculated as follows: ; in, For the region For adjacent areas Attention weights; area For adjacent areas The score is the attention score; For the region The neighborhood set; For all adjacent regions Attention score; It is an exponential function; Furthermore, a differentiated model of the interactive characteristics between holidays and weekdays is established, and a time type indicator function is defined. : ; The access probability and reliability of different types of electric vehicles adjust over time: ; ; in, and For the adjusted access probability and reliability; and For adjustment coefficients; For the first electric vehicles during the period The probability of access; The reliability coefficient is between 0 and 1; therefore, the adjustment coefficient for electric buses is... The adjustment factor for electric private vehicles is: The adjustment factor for electric taxis is: .

[0030] Furthermore, constructing an electric vehicle aggregation model corresponding to the day-ahead and intraday timescales also includes: Based on the access probability of various types of electric vehicles in each hour, an hourly aggregated power model is established at the day-ahead time scale to determine the day-ahead predicted aggregated power of various types of electric vehicles. Based on the real-time access status and actual adjustable power of various electric vehicles at each minute, a minute-level aggregated power model is established at the intraday time scale to determine the intraday real-time aggregated power of various electric vehicles. Based on the time cycle characteristics and response characteristics of various types of electric vehicles, electric buses are identified as the basic control resource on the day-ahead time scale, while electric taxis and electric private cars are identified as supplementary control resources on the day-ahead time scale and flexible control resources on the intraday time scale.

[0031] As a preferred embodiment of the above, a multi-timescale aggregation model is further established. The day-ahead timescale (hour-level) aggregation model is as follows: ; in, For the first Electric vehicles were recently... Predicted hourly polymerization power (kW); For the first Number of electric vehicles; For the first The car in Hourly access probability ; For the first Class 1 The vehicle's maximum charging power or maximum adjustable power; For the first Class 1 The car in Reliability coefficient per hour; The intraday timescale (minute-level) aggregation model is as follows: ; in, For the first Electric vehicles were the first in the day Real-time aggregated power (kW) per minute; Real-time access status (0 or 1); This refers to the actual adjustable power. Indexed by minutes For the first Class 1 The car in Reliability coefficient per minute; Furthermore, due to their hourly periodicity, electric buses primarily participate in day-ahead scheduling, serving as a fundamental regulatory resource. ; in, This represents the day-ahead scheduling participation coefficient. For electric buses, the first day Hourly predicted polymerization power; Electric private cars and electric taxis serve as a supplement to daytime dispatch, but will play a key role in intraday dispatch: ; ; in, For electric taxis, the first day of The supplementary control power available during each time period, This represents the actual controlled power of the electric taxi during the m-th minute of the day. For electric taxis, the first day of Hourly predicted polymerization power The day-ahead dispatch participation coefficient of electric taxis is the real-time aggregated power of electric taxis at minute m within a day. Daily dispatch participation coefficient of electric taxis ; For electric private cars, the first The supplementary control power available during each time period, This refers to the actual controlled power of an electric private vehicle in the m-th minute of the day. For electric private cars, the first Hourly predicted polymerization power The real-time aggregated power of electric private vehicles in the m-th minute of the day, and the day-ahead dispatch participation coefficient of electric private vehicles. Daily dispatch participation coefficient of electric private vehicles ; Furthermore, the dynamic update equation for the state of charge is established: ; in, For the first Class 1 The vehicle's state of charge for the next time period; For the first Class 1 vehicle during the time period The current state of charge; For the first Class 1 vehicle time period The charging power (kW); The time interval is (h). For the first Class 1 vehicle time period Energy consumption for travel (kWh); For the first Class 1 The vehicle's battery capacity; Furthermore, the state-of-charge constraints are as follows: ; ; in, and These are the lower and upper limits of the state of charge, respectively; The state of charge at the moment of departure; The minimum state of charge required by the user.

[0032] Furthermore, constructing a multi-level coordinated control model for electric vehicle aggregates, including: Based on the hierarchical topology between the main network layer, distribution network layer, microgrid layer and charging station layer, establish the hierarchical relationships between each layer; Based on their affiliation, various types of electric vehicles are aggregated layer by layer from the charging station layer to the microgrid layer, from the microgrid layer to the distribution network layer, and from the distribution network layer to the main grid layer, in order to form electric vehicle aggregates at each level. Based on the aggregation results of electric vehicle aggregates at each level, the power external characteristics, capacity external characteristics, response speed external characteristics, and reliability external characteristics of electric vehicle aggregates at each level are determined, and power balance constraints, capacity constraints, and response speed constraints between levels are established.

[0033] Furthermore, constructing a multi-level coordinated control model for electric vehicle aggregates, including: Based on the number of electric vehicle aggregates connected at each level, the state of charge distribution, time characteristics, and external input information, a prediction model for the external characteristics of electric vehicle aggregates at each level is established. The external characteristics of electric vehicle aggregates at various levels under the daytime timescale are predicted based on the external characteristic prediction model to obtain the daytime external characteristic prediction results. Based on real-time measurement information and day-ahead external characteristic prediction results, error correction is performed on the external characteristics of electric vehicle aggregates at each level on the intraday timescale to obtain intraday real-time external characteristic update results.

[0034] As a preferred embodiment of the above, a collaborative control model for a multi-level electric vehicle aggregate, including main and auxiliary components, is established: Furthermore, a three-tiered topology structure is constructed: main network - distribution network - microgrid. The main network layer includes... One main network node, the distribution network layer includes Each distribution network area, the microgrid layer includes Each microgrid unit, the charging station layer includes For each charging station, define a hierarchical membership matrix: ; in Represents a node Belonging to node .

[0035] Furthermore, an external characteristic model of the polymer in a microgrid layer electric vehicle was established. The aggregation of individual microgrid units in time period The power external characteristics are: ; in, For the first Total controllable power (kW) of each microgrid aggregate; For the first Within the micronet A collection of electric vehicles; Number of electric vehicle types; For the first The first type of electric vehicle vehicle during the time period Adjustable power; External characteristics of capacity are: ; in, For the first Adjustable capacity (kWh) of each microgrid aggregate; For the first The first type of electric vehicle vehicle during the time period Adjustable capacity; The external characteristics of response speed are characterized by a weighted average time constant: ; in, For the first The response time constant (min) of a microgrid aggregate; For the first Class 1 The vehicle's response time constant; For the first The first type of electric vehicle Maximum charging power of the vehicle; response time constant of electric buses min, response time constant of electric taxi min, response time constant of electric private vehicles min; Furthermore, an external characteristic model of electric vehicle aggregates at the distribution network layer is established. The aggregate power of each distribution network area is: ; in, For the first Total control power (kW) of each distribution network aggregate; For the first A distribution network area contains a set of microgrid units; For the first Total controllable power (kW) of each microgrid aggregate; The reliability characteristics of the distribution network layer aggregate take into account the aggregation effect of the underlying microgrid: ; in, For the first Overall reliability of individual power distribution network aggregates; For the first The reliability of each microgrid aggregate is calculated as follows: ; in, For the first Within the micronet A collection of electric vehicles; Number of electric vehicle types; For the first Class 1 The car in Reliability coefficient for a given time period; For the first The first type of electric vehicle The vehicle's maximum charging power; Furthermore, an external characteristic model of the electric vehicle aggregate at the main network layer is established, with main network nodes... The polymer power is: ; in, mainnet node Electric vehicle aggregate power (kW); mainnet node The set of distribution network areas under its jurisdiction; For the first Total control power (kW) of each distribution network aggregate; The response bandwidth characteristics of the main network layer aggregate are: ; in, The response bandwidth (kW / min) of the main network aggregate; and These are the upper and lower limits of power, respectively; For the response time interval; Furthermore, a rapid prediction model for aggregate in vitro properties is established using deep neural networks (DNNs): ; in, is the prediction vector for the external characteristics of the main network aggregate, where External characteristics of the main network layer aggregate capacity External characteristics related to the reliability of the main network layer aggregate; These are state variables (number of electric vehicles connected to the network for each type, SOC distribution, etc.). External input variables (time characteristics, meteorological information, etc.); It is a historical external characteristic sequence; For neural network parameters; The DNN uses a three-layer fully connected structure with ReLU as the activation function. ; ; ; in, The input vector; The number of fully connected layers; and The first Layer weight matrix and bias vector; Furthermore, a multi-level collaborative constraint model is established, including power balance constraints: ; ; ; in, mainnet node The combined power (kW) of electric vehicles. mainnet node The set of distribution network areas under its jurisdiction; For the first Total control power (kW) of each distribution network aggregate. For the first A distribution network area contains a set of microgrid units; For the first Total controllable power (kW) of each microgrid aggregate; For the first The power regulation of each charging station; For the first A microgrid is a collection of charging stations; Capacity constraints: ; ; ; in, mainnet node Maximum power output of electric vehicle aggregate (kW); For the first The maximum total control power (kW) of each distribution network aggregate. For the first The maximum total controllable power (kW) of each microgrid aggregate. Response speed constraints: ; in, The upper limit of the power change rate of the main network aggregate is related to the response bandwidth. , The response bandwidth (kW / min) of the main network aggregate; Furthermore, a multi-timescale extrinsic characteristic model is established; extrinsic characteristic prediction at the current timescale: ; in, For the day before Hourly external characteristic prediction; For expectation operators; No. The node at the th The unpredictable external characteristics of the hour; This is the set of information available at present; Updates of external characteristics on intraday timescales: ; in, For the first time in a day Real-time external characteristic values ​​per minute; For the first The hour corresponding to the minute For the first The minute corresponding to the day before yesterday Hourly external characteristic prediction; The prediction error correction term is estimated in real time using the Kalman filter algorithm: ; in, Kalman gain; These are measured values; The observation matrix; These are predicted values.

[0036] Furthermore, generating coordinated control instructions at various levels includes: The state space of the main grid layer intelligent agent is constructed based on the load power, renewable energy output power, electricity price and historical status information of the main grid layer, and the action space of the main grid layer intelligent agent is constructed based on the control power instructions issued by the main grid layer to each distribution network area. The state space of the distribution network layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate in the distribution network layer, the control power command issued by the main grid layer, and the node voltage. The action space of the distribution network layer intelligent agent is constructed based on the control power command issued by the distribution network layer to each microgrid. The state space of the microgrid layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate and the local load power of the microgrid, and the action space of the microgrid layer intelligent agent is constructed based on the power control instructions issued by the microgrid layer to each charging station. Based on the state and action spaces of the main network layer agents, distribution network layer agents, and microgrid layer agents, a multi-agent reinforcement learning algorithm is used to generate coordinated control commands that are passed down step by step from the main network layer to the distribution network layer, from the distribution network layer to the microgrid layer, and from the microgrid layer to the charging station layer.

[0037] Furthermore, the charging and discharging power allocation for electric vehicles connected to the charging station includes: Based on the coordinated control instructions received by the charging station layer, the target control power of the charging station layer in the current time period is determined; Based on the target power control and the power requested by the connected vehicles, the electric vehicles connected to the charging station are proportionally allocated to determine the actual charging and discharging power of each connected vehicle. The actual charging and discharging power is adjusted according to the charging status priority of the access vehicle, and the charging status of the access vehicle is updated based on the adjusted actual charging and discharging power to meet the off-site charging status requirements of the access vehicle.

[0038] As a preferred embodiment of the above, a multi-level electric vehicle aggregate collaborative control strategy based on reinforcement learning algorithm is designed; Furthermore, a multi-agent deep reinforcement learning (MADRL) algorithm is used to address the multi-level collaborative regulation problem; the main network layer acts as the central agent, while the distribution network layer and micro-network layer act as local agents, achieving collaborative regulation through hierarchical decision-making. Furthermore, the state space of the central agent in the main network layer is defined as follows: ; in, mainnet node During the period The state vector; Load power (kW); Power output from renewable energy sources (kW); Electricity price (RMB / kWh); This is historical state information; Action space is defined as the control commands issued from the main network layer to the distribution network layer: ; in, mainnet node The action vector; To be issued to the first Control power command (kW) for each distribution network area; mainnet node Number of distribution network areas under its jurisdiction; The reward function comprehensively considers economy, balance, and reliability: ; in, Total reward; These are the weighting coefficients; The awards are based on economic efficiency, balance, and reliability. The economic incentives are: ; in, mainnet node Electric vehicle aggregate power (kW); Network loss cost (RMB); The balance reward is: ; in, This is the balance penalty coefficient; Load power (kW); mainnet node Electric vehicle aggregate power (kW); Power output from renewable energy sources (kW); The exchange power (kW) with the upstream power grid; The reliability bonus is: ; in, This is a reliability penalty coefficient, applied when the control command exceeds the reliable power range; External characteristics related to the reliability of the main network layer aggregate; mainnet node Maximum power output of electric vehicle aggregate (kW); Furthermore, the state space of the local intelligent agent in the distribution network layer is defined as follows: ; in, For distribution network area The state vector; External properties of the power distribution network polymer; Control instructions issued by the main network; For distribution network area Load power (kW); Node voltage (pu); This is historical state information; The distribution network layer's action space consists of the control commands allocated to each microgrid. ; in, To be assigned to the Control power command (kW) for each microgrid; For distribution network area The number of microgrids included; The reward function for the distribution network layer is: ; in, This is the penalty coefficient; To be issued to the first Control power command (kW) for each distribution network area; Node voltage (pu); Reference voltage (typically 1.0 pu); Furthermore, the state space of the local agents in the microgrid layer is defined as follows: ; in, For micro-network The state vector; For micro-network The external properties of the distribution network polymer; Local load power of the microgrid (kW); For micro-network Historical status information; The microgrid layer's action space consists of control commands allocated to each charging station: ; in, To be assigned to the Control power command (kW) for each charging station; For micro-network The number of charging stations included; The reward function for the microgrid layer is: ; in, The second term, representing the penalty coefficient, penalizes the degree to which the user's SOC (System-Oriented Content) requirements are not met. For the first Class 1 The state of charge of an electric vehicle at the moment of departure; For the first Class 1 Users of electric vehicles are required to have a minimum state of charge. Furthermore, the Soft Actor-Critic (SAC) algorithm is used to train agents at each level; SAC combines the actor-critic architecture and maximum entropy reinforcement learning, which can handle continuous action space and improve exploration efficiency. The probability distribution of the actions output by the Actor network (policy network): ; in, For parameters The strategy function; For power control commands; It is a state vector; and These are functions of the mean and standard deviation, respectively. It follows a normal distribution; Critic networks (value networks) contain two Q-functions to mitigate overestimation problems: ; in, For parameters The Q function; As a discount factor, To sum the time index, For the current moment, This is the time when the round ends; In strategy The expected value of the following; The entropy temperature parameter; It is the entropy function; In order to be in The state vector at any given time; The Critic network update objective is to minimize the Bellman error: ; in, The reward for the current state; This serves as a buffer for experience replay. For the target Q value: in, For target network parameters; Discount factor; The entropy temperature parameter; Actions for sampling the next state; The Actor network update objective is to maximize expected return and policy entropy: ; Temperature parameters Automatic adjustment, with the following optimization goals: ; in, The target entropy is typically set to the negative of the action dimension. Furthermore, a multi-level collaborative communication mechanism is established; the main network layer sends control commands and status information to the distribution network layer: ; in, To be issued to the first Control power command (kW) for each distribution network area; Electricity price (RMB / kWh); Load power (kW); Power output from renewable energy sources (kW); The distribution network layer sends control commands and constraint information to the microgrid layer: ; in, To be assigned to the Control power command (kW) for each microgrid; Node voltage (pu); For distribution network Load power (kW); The microgrid layer sends power allocation commands and SOC targets to the charging station: ; in, To be assigned to the Control power command (kW) for each charging station; For future time periods SOC objectives; Furthermore, a multi-timescale coordinated control process is established from day-ahead to intraday; during the day-ahead phase (T-1 day), hourly scheduling is performed based on forecast information: ; in, For the first The first mainnet node was recently... Hourly control instructions; This is the day-ahead scheduling strategy function; For the first The first mainnet node was recently... The state vector for each hour; During the intraday phase (Day T), adjustments are made minute-by-minute based on real-time information: ; in, For the first The first mainnet node within the day Minute-level control instructions; This is the intraday scheduling strategy function; For the first The first mainnet node within the day The state vector for each minute; Intraday adjustments are constrained by the previous day's plan: ; in, For the first The first mainnet node within the day The corresponding minute of the day Hourly dispatch instructions; The daily adjustment limit; Furthermore, electric buses primarily operate under a day-ahead scheduling plan: ; in, For electric buses in the first One charging station, the first of the day Intraday control instructions at specific times; For electric buses in the first One charging station, the first of the day The date corresponding to the time is the The daytime control order for the hour; For minor adjustments, ; Electric taxis and electric private cars will primarily be subject to intraday adjustments: in, For electric taxis in the first One charging station, the first of the day Intraday control instructions at specific times; For electric taxis in the first One charging station, the first of the day The date corresponding to the time is the The daytime control order for the hour; For electric private cars in the first One charging station, the first of the day Intraday control instructions at specific times; For electric private cars in the first One charging station, the first of the day The date corresponding to the time is the The daytime control order for the hour; , ; Furthermore, the charging station layer executes a power allocation algorithm; A charging station received a microgrid instruction. Then, a proportional allocation strategy is adopted: ; in, For the first Class 1 The actual power distribution of the vehicle; To request power; For charging stations The current collection of electric vehicles connected, Index for electric vehicle types, Index for the number of electric vehicles; Simultaneously consider SOC priority adjustment: ; in, This is the SOC priority coefficient (usually taken as 0.2). For the first Class 1 The vehicle's user requires a state of charge; For the first Class 1 The car is The current state of charge for the current time period; For the first Class 1 The minimum state of charge of a vehicle.

[0039] Example 2; Based on the same inventive concept as the collaborative control method for multi-level vehicle-to-grid interaction of main, service, and micro-level vehicles in the foregoing embodiments, the present invention also provides a collaborative control system for multi-level vehicle-to-grid interaction of main, service, and micro-level vehicles, the system comprising: The time-scale aggregation module constructs electric vehicle aggregation models corresponding to the day-ahead and intraday time scales based on historical operating data, real-time access status, and state of charge of various types of electric vehicles, thereby obtaining the adjustable power and adjustable capacity of various types of electric vehicles. The power grid hierarchical aggregation module constructs a multi-level coordinated control model for electric vehicle aggregation based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging stations, and obtains the control constraint relationship between the levels; The hierarchical collaborative control module, based on the multi-level collaborative control model of electric vehicle aggregates (main, distribution, and micro), generates collaborative control commands for each level according to the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. The charging and discharging power distribution module allocates charging and discharging power to electric vehicles connected to the charging station based on coordinated control instructions and adjustable power and capacity, so as to achieve orderly charging and discharging control under vehicle-to-grid interaction.

[0040] The system described above in this invention can effectively realize a collaborative control method for multi-level vehicle-to-grid interaction, and the technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0041] Example 3; Based on the same inventive concept as the collaborative control method for multi-level vehicle-to-network interaction of main, auxiliary, and micro-level vehicles in the foregoing embodiments, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, can realize the collaborative control method for multi-level vehicle-to-network interaction of main, auxiliary, and micro-level vehicles.

[0042] Example 4; To verify the effectiveness of this solution, actual operating data from a municipal power grid was selected for case analysis. The city's main grid comprises 2 major nodes, its distribution network comprises 8 areas, its microgrid comprises 32 units, and it has a total of 128 charging stations. The total number of electric vehicles is 150,000, including 100,000 electric private cars, 35,000 electric taxis, and 15,000 electric buses. Data collection spanned from January 1, 2024 to December 31, 2024, with a time resolution of 15 minutes. The experimental environment consisted of an Intel Xeon Platinum 8358 processor, 256GB of memory, and was implemented using the PyTorch 2.0 deep learning framework and the RayRLlib reinforcement learning library. Reinforcement learning algorithm parameter settings: discount factor Initial entropy temperature (Adaptive adjustment), both the actor and critic networks use a three-layer fully connected structure (hidden layer dimensions are 256-256-128), with a learning rate of... The experience replay buffer size is Batch size is 256, target network soft update coefficient The training process uses a parallel environment simulation, with a total of 500 epochs, each containing a 24-hour (1440 time steps) adjustment process. Electric vehicle parameter settings: Electric buses have a battery capacity of 200 kWh, a maximum charging power of 60 kW, a weekday reliability of 0.90, a holiday reliability of 0.85, and a response time of 10 minutes; electric taxis have a battery capacity of 70 kWh, a maximum charging power of 40 kW, a reliability of 0.75, and a response time of 5 minutes; electric private cars have a battery capacity of 60 kWh, a maximum charging power of 7 kW, a weekday reliability of 0.60, a weekend reliability of 0.55, and a response time of 15 minutes; the charging efficiency is uniformly set to 0.90, the SOC constraint range is [0.20, 0.95], and the minimum departure SOC required by the user is 0.80. Economic parameter settings: Time-of-use electricity price: Peak period (10:00-15:00, 18:00-21:00) 1.2 yuan / kWh, Side period (07:00-10:00, 15:00-18:00, 21:00-23:00) 0.7 yuan / kWh, Off-peak period (23:00-07:00 the next day) 0.4 yuan / kWh; Network loss cost coefficient: 0.05 yuan / (kW·h); Reward function weight coefficient: ; ; ; ; The comparison methods include: (1) traditional optimization scheduling method (based on mixed integer linear programming); (2) single-level centralized reinforcement learning method; (3) unified control method that does not consider the differences of multiple types; the evaluation indicators include: load peak-valley difference, renewable energy consumption rate, operating cost, voltage deviation, user satisfaction, and computation time; Simulation results are as follows Figure 2-4 As shown; Figure 2This study demonstrates the charging load distribution characteristics of different types of electric vehicles over a typical weekday within 24 hours and its control effect across multiple time scales. It shows that the charging load of electric buses exhibits a clear bimodal distribution, mainly concentrated between 12:00-14:00 and 21:00-06:00 the next day, exhibiting strong temporal regularity. In daytime scheduling, it serves as a fundamental control resource with a tracking error of less than 5%. Electric private vehicles primarily charge between 19:00 and 07:00 the next day, with a relatively dispersed load distribution, allowing for flexible adjustment in daytime scheduling. Resources can be adjusted by 40%-60%; electric taxis are distributed throughout the day, with higher charging loads during shift change periods (06:00-08:00 and 18:00-20:00), and participate in day-ahead and intraday dispatching, resulting in a fast response time (within 5 minutes); the method of this invention, through multi-type coordinated regulation, guides a large number of electric vehicles to charge during off-peak hours (23:00-06:00) to absorb wind power, and reduces charging load during peak hours (18:00-21:00) by utilizing some electric taxis for V2G discharge, significantly reducing the peak-valley load difference; Figure 3 The performance of different methods in terms of multi-level collaborative control was compared. The results show that the method of the present invention can achieve effective decomposition of the main grid, distribution network, microgrid, and charging station: after the main grid layer issues the overall control command, the distribution network layer reasonably allocates sub-commands according to the aggregate characteristics (adjustable capacity, reliability, etc.) of each distribution network area, with an allocation deviation of less than 3%; the microgrid layer further decomposes the command to each charging station according to the local load and voltage constraints, with the voltage deviation controlled within ±0.03 pu; the charging station layer allocates power according to the SOC status and charging demand of the connected electric vehicles, and the user SOC demand satisfaction rate reaches 97.5%; compared with the single-level centralized method, the multi-level collaborative method of the present invention reduces the calculation time by 73% (from an average of 18.2 seconds to 4.9 seconds), while improving the executability of the control command by 26%, fully demonstrating the advantages of the layered architecture; Figure 4 The study demonstrates the effectiveness of multi-objective collaborative optimization based on the SAC reinforcement learning algorithm. The final trained control strategy performed well on the test set, resulting in a 19.7% reduction in operating costs (compared to traditional optimization methods) and a 15.3% reduction (compared to methods that do not consider the differences between different types of loads), a 34.6% reduction in load peak-valley difference, a 22.4% increase in renewable energy absorption rate, and a 41.2% reduction in voltage deviation. At the same time, it ensured a user satisfaction rate of 91.3%, achieving multi-objective collaborative optimization of economy, balance, and reliability.

[0043] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A collaborative control method for multi-level vehicle-to-grid interaction, characterized in that, The method includes: Based on the historical operating data, real-time access status and state of charge of various types of electric vehicles, an electric vehicle aggregation model corresponding to the day-ahead time scale and intraday time scale is constructed to obtain the adjustable power and adjustable capacity of various types of electric vehicles. Based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging station, a multi-level coordinated control model of the main grid, distribution network and microgrid is constructed to obtain the control constraint relationship between the levels. Based on the aforementioned multi-level electric vehicle aggregation collaborative control model, collaborative control commands for each level are generated according to the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. According to the coordinated control command, the adjustable power, and the adjustable capacity, the charging and discharging power is allocated to the electric vehicles connected to the charging station to achieve orderly charging and discharging control under vehicle-to-grid interaction.

2. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 1, characterized in that, Construct an electric vehicle aggregation model corresponding to the day-ahead and intraday time scales, including: Based on the battery capacity, charging and discharging power, connection status and state of charge of various types of electric vehicles, a single-vehicle adjustable capacity model is established to determine the adjustable capacity of various types of electric vehicles in the target time period. Based on the adjustable capacity, the number of electric vehicles of each type, the maximum charging power, and the reliability coefficient, a type aggregation power model is established to determine the upper limit of the total adjustable power of each type of electric vehicle during the target period. Based on the historical access data sequences and time characteristics of various types of electric vehicles, an access probability prediction model is established to determine the access probability of each type of electric vehicle under the stated day-ahead time scale and the stated intraday time scale.

3. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 2, characterized in that, The spatial distribution and temporal type of the access probability are corrected, including: The control area is spatially divided based on the location distribution information of various types of electric vehicles, and the aggregate power of various types of electric vehicles in each spatial area is determined based on the location of each vehicle and the actual control power. The corresponding weekday or holiday time type is determined based on the time feature information, and the access probability and the reliability coefficient are corrected according to the time type. Based on the corrected access probability and the reliability coefficient, the aggregation characteristics of multiple types of electric vehicles in each spatial region under different time types are determined.

4. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 3, characterized in that, Constructing an electric vehicle aggregation model corresponding to the day-ahead and intraday time scales also includes: Based on the access probability of various types of electric vehicles in each hour, an hourly aggregated power model is established at the day-ahead time scale to determine the day-ahead predicted aggregated power of various types of electric vehicles. Based on the real-time access status and actual adjustable power of various electric vehicles at each minute, a minute-level aggregated power model is established at the intraday time scale to determine the intraday real-time aggregated power of various electric vehicles. Based on the time cycle characteristics and response characteristics of various types of electric vehicles, electric buses are identified as the basic control resource under the day-ahead time scale, and electric taxis and electric private cars are identified as supplementary control resources under the day-ahead time scale and flexible control resources under the intraday time scale.

5. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 1, characterized in that, Constructing a multi-level coordinated control model for electric vehicle aggregates, including: Based on the hierarchical topology between the main network layer, distribution network layer, microgrid layer and charging station layer, establish the hierarchical relationships between each layer; Based on the aforementioned hierarchical relationship, multiple types of electric vehicles are aggregated layer by layer from the charging station layer to the microgrid layer, from the microgrid layer to the distribution network layer, and from the distribution network layer to the main grid layer, in order to form electric vehicle aggregates at each level; Based on the aggregation results of electric vehicle aggregates at each level, the power external characteristics, capacity external characteristics, response speed external characteristics, and reliability external characteristics of electric vehicle aggregates at each level are determined, and power balance constraints, capacity constraints, and response speed constraints between levels are established.

6. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 5, characterized in that, The construction of a multi-level coordinated control model for electric vehicle aggregates, including main, auxiliary, and micro levels, also includes: Based on the number of electric vehicle aggregates connected at each level, the state of charge distribution, time characteristics, and external input information, a prediction model for the external characteristics of electric vehicle aggregates at each level is established. The external characteristics of electric vehicle aggregates at each level under the daytime scale are predicted based on the external characteristic prediction model to obtain the daytime external characteristic prediction results. Based on real-time measurement information and the daytime external characteristic prediction results, the external characteristics of electric vehicle aggregates at each level under the intraday time scale are corrected for errors to obtain intraday real-time external characteristic update results.

7. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 1, characterized in that, Generate coordinated control instructions at each level, including: The state space of the main grid layer intelligent agent is constructed based on the load power, renewable energy output power, electricity price and historical status information of the main grid layer, and the action space of the main grid layer intelligent agent is constructed based on the control power instructions issued by the main grid layer to each distribution network area. The state space of the distribution network layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate in the distribution network layer, the control power command issued by the main grid layer, and the node voltage. The action space of the distribution network layer intelligent agent is constructed based on the control power command issued by the distribution network layer to each microgrid. The state space of the microgrid layer intelligent agent is constructed based on the external characteristics of the electric vehicle aggregate and the local load power of the microgrid, and the action space of the microgrid layer intelligent agent is constructed based on the power control instructions issued by the microgrid layer to each charging station. Based on the state and action spaces of the main network layer agent, the distribution network layer agent, and the microgrid layer agent, a multi-agent reinforcement learning algorithm is used to generate the coordinated control instructions that are transmitted step by step from the main network layer to the distribution network layer, from the distribution network layer to the microgrid layer, and from the microgrid layer to the charging station layer.

8. The collaborative control method for multi-level vehicle-to-grid interaction based on main, auxiliary, and micro-level systems according to claim 1, characterized in that, Distributing charging and discharging power to electric vehicles connected to the charging station, including: Based on the coordinated control command received by the charging station layer, the target control power of the charging station layer in the current time period is determined; Based on the target control power and the requested power of the connected vehicles, the electric vehicles connected to the charging station layer are proportionally allocated to determine the actual charging and discharging power of each connected vehicle. The actual charging and discharging power is adjusted according to the charging state priority of the access vehicle, and the charging state of the access vehicle is updated according to the adjusted actual charging and discharging power to meet the off-site charging state requirements of the access vehicle.

9. A collaborative control system for multi-level vehicle-to-grid interaction, characterized in that: The system includes: The time-scale aggregation module constructs electric vehicle aggregation models corresponding to the day-ahead and intraday time scales based on historical operating data, real-time access status, and state of charge of various types of electric vehicles, thereby obtaining the adjustable power and adjustable capacity of various types of electric vehicles. The power grid hierarchical aggregation module constructs a multi-level coordinated control model for electric vehicle aggregation based on the electric vehicle aggregation model and the hierarchical relationship between the main grid, distribution network, microgrid and charging stations, and obtains the control constraint relationship between the levels; The hierarchical collaborative control module, based on the multi-level collaborative control model of electric vehicle aggregates (main, distribution, and micro), generates collaborative control commands for each level according to the transmission paths from the main grid layer to the distribution grid layer, from the distribution grid layer to the microgrid layer, and from the microgrid layer to the charging station layer. The charging and discharging power distribution module allocates charging and discharging power to electric vehicles connected to the charging station based on coordinated control instructions and adjustable power and capacity, so as to achieve orderly charging and discharging control under vehicle-to-grid interaction.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the collaborative control method for multi-level vehicle-to-grid interaction as described in any one of claims 1-8.

Images