Virtual power plant vehicle network resource optimization control method and system based on digital twinning

Abstract

The invention discloses a virtual power plant vehicle network resource optimization control method and system based on digital twinning, and belongs to the technical field of intelligent power grid resource optimizing.The method comprises the steps that video stream data and battery state data are collected, physical connection stability features, thermal stress features and energy state features are extracted, dynamic clustering is conducted on electric vehicles, and the dynamic clustering result is obtained. And a resource cluster is formed, the resource cluster is used as an agent, a digital twinborn collaborative optimization model based on a multi-agent deep reinforcement learning algorithm is constructed, and a charging and discharging control instruction is generated. According to the method, the physical connection reliability is evaluated by introducing computer vision, the battery health state is evaluated in combination with a gating circulation unit network, the vehicle state is comprehensively perceived, and through collaborative optimization of dynamic clustering and multi-agent deep reinforcement learning, the real-time performance of the vehicle is improved. The problems of scheduling execution failure and battery life acceleration attenuation caused by incomplete state evaluation in the prior art are solved, and the safety, accuracy and reliability of vehicle network resource optimization control are improved.

Description

Technical Field

[0001] This invention belongs to the field of smart grid resource optimization technology, specifically relating to a method and system for optimizing and controlling virtual power plant vehicle-grid resources based on digital twins. Background Technology

[0002] Virtual power plant (V2P) vehicle-to-grid (V2G) resources refer to the aggregation of geographically dispersed electric vehicles capable of bidirectional interaction with the power grid into a unified and dispatchable collective through advanced communication and control technologies. The core component of V2P V2G resources is the energy storage system of electric vehicle batteries. V2P V2G resources have adjustable charging load and controllable discharge power. Through centralized monitoring and optimized scheduling of the aggregated electric vehicle cluster, it participates as a whole in the electricity market and grid ancillary services, providing capacity support, peak shaving and frequency regulation services, and reserve capacity for the power system, thereby improving the flexibility and reliability of grid operation.

[0003] To address the challenges of precise control and safe operation when large-scale electric vehicle clusters participate in grid dispatch, existing technologies employ a unified dispatching approach by establishing an aggregation optimization model based on battery electrical state parameters obtained through vehicle-to-grid communication. However, this approach still suffers from limitations in quantifying the physical connection status of charging interfaces and neglecting the differences in battery health status among electric vehicles within the cluster. Consequently, dispatching commands are issued to vehicles with unreliable physical connections, and indiscriminate charging and discharging control is applied to batteries with varying health statuses, ultimately leading to dispatching failures and accelerated battery life degradation. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing and controlling vehicle-to-grid resources in a virtual power plant based on digital twins, thereby improving the safety, accuracy, and reliability of vehicle-to-grid resource optimization and control.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A virtual power plant vehicle-to-grid resource optimization and control method based on digital twins includes the following steps: The video stream data of the charging port area and the battery status data of the electric vehicle are collected and preprocessed to obtain the preprocessed video stream data and battery status data. The preprocessed video stream data is input into a three-dimensional convolutional neural network model to extract physical connection stability features and establish an early warning mechanism. Thermal stress features and energy state features are extracted from the preprocessed battery state data. The thermal stress characteristics and energy state characteristics are input into the gated cyclic unit network model, and the battery health state parameters are output. Based on the physical connection stability characteristics and health status parameters, a density-based noise-based spatial clustering algorithm is used to divide the connected electric vehicles into different resource clusters. Using the aggregation parameters of resource clusters as input to a digital twin collaborative optimization model based on a multi-agent deep reinforcement learning algorithm, a distributed collaborative solution is performed to obtain the total power adjustment command for each resource cluster, and a secondary power allocation is executed to generate charging and discharging control commands. Visualize physical connection stability characteristics, health status parameters, resource clusters, and charge / discharge control commands.

[0006] Preferably, the process of obtaining the preprocessed video stream data and battery status data includes: Video acquisition devices are deployed on each charging pile, with the devices positioned directly opposite the charging port area of ​​the connected electric vehicle, to acquire continuous video stream data of the charging port area. A data acquisition unit is integrated into the control system of the charging pile. The data acquisition unit establishes a communication connection with the battery management system of the connected electric vehicle through the charging communication protocol to acquire battery status data of the connected electric vehicle, including battery state of charge data and battery surface temperature data. Frame sampling is performed on the video stream data to obtain a video sequence; Perform image transformation on each frame of the video sequence to convert multi-channel color images into single-channel images; The pixel values ​​of a single-channel image are normalized to scale them to a standardized range, resulting in preprocessed video stream data. The collected battery status data is cleaned to remove missing and outlier values. A filtering algorithm based on a state estimation model is used to denoise the cleaned battery state data; The denoised battery state data is standardized to eliminate the influence of different physical quantities, resulting in preprocessed battery state data. Align the timestamps of the preprocessed video stream data and battery status data.

[0007] Preferably, the process of extracting physical connection stability features and establishing an early warning mechanism includes: The preprocessed video stream data is divided into video segments containing continuous image frames according to a preset time step, and used as input to a three-dimensional convolutional neural network model. The three-dimensional convolutional neural network model performs convolution operations on video segments using three-dimensional convolutional kernels to extract spatiotemporal features that include both time and spatial dimensions. The extracted spatiotemporal features are input into the fully connected regression layer of the 3D convolutional neural network model, and the output is a scalar value, which is defined as the physical connection stability feature. Set a safety threshold for physical connection stability, compare the value of the physical connection stability feature with the safety threshold in real time, and generate an early warning signal when the value of the physical connection stability feature is lower than the safety threshold, and remove the corresponding electric vehicle from the schedulable resource pool.

[0008] Preferably, the process of extracting thermal stress features and energy state features from the preprocessed battery state data includes: Calculate the difference between the preprocessed battery surface temperature data at the current moment and the preprocessed battery surface temperature data at the previous sampling moment, and divide the difference by the time interval between the two sampling moments to obtain the thermal stress characteristics; Energy state features are extracted from the preprocessed battery state data, which are standardized battery state of charge data.

[0009] Preferably, the process of outputting the battery's health status parameters includes: The thermal stress characteristics and energy state characteristics are combined into a two-dimensional input feature vector at each sampling time. The two-dimensional input feature vectors within continuous time steps are constructed into an input sequence with a time order relationship. The input sequence is fed into the gated recurrent unit network model. The gated recurrent unit network model processes the input sequence through its internal update and reset gates and updates the hidden state of the gated recurrent unit network model to capture the temporal dependencies in the input sequence. The hidden state output by the gated recurrent unit network model after processing the input sequence is input into the fully connected output layer; The fully connected output layer outputs a two-dimensional vector, which is then defined as a battery health state parameter.

[0010] Preferably, the process of dividing the connected electric vehicles into different resource clusters includes: The physical connection stability characteristics and health status parameters of each connected electric vehicle are combined into three-dimensional data points, and the three-dimensional data points from each connected electric vehicle are used to form a feature space. To apply a density-based spatial clustering algorithm to noise, we set the neighborhood radius parameter and the minimum number of neighborhood points for the core object. Calculate the Euclidean distance between two 3D data points in the feature space; For a 3D data point in the feature space, the number of 3D data points contained in the neighborhood of the 3D data point within the range of the neighborhood radius parameter is determined by comparing the Euclidean distance with the neighborhood radius parameter. If the number of 3D data points contained in the neighborhood of a 3D data point is not less than the minimum number of neighborhood points of the core object, then the 3D data point is marked as the core object. Starting with an unvisited core object, create a new cluster, find 3D data points that are density-reachable from the core object, and assign the found 3D data points to the new cluster; Repeat the process of creating new clusters from unvisited core objects, finding density-reachable 3D data points, and assigning density-reachable 3D data points to new clusters until all core objects have been accessed. The set of electric vehicles corresponding to the three-dimensional data points within a cluster is defined as a resource cluster.

[0011] Preferably, the process of performing distributed collaborative solving to obtain the total power adjustment command for each resource cluster includes: Resource clusters are defined as independent intelligent agents, and these agents constitute a multi-agent system. Construct a digital twin collaborative optimization model based on a multi-agent deep reinforcement learning algorithm; Each agent is given an independent actor network, and a central critic network is set up for the multi-agent system, forming an actor-critic architecture; Define the state space, action space, and reward function of a multi-agent system; The state space includes aggregated parameters for each resource cluster and global information about the power grid. Aggregated parameters include the average physical connectivity stability characteristics of electric vehicles within the cluster, average health state parameters, total charging power, and total available capacity. The action space of each agent is the continuous power adjustment range of the resource cluster; The reward function is used to evaluate the immediate reward obtained by each agent after performing an action during training. The immediate reward is calculated based on the grid frequency deviation and the expected battery loss of electric vehicles in the cluster. In the digital twin collaborative optimization model, training is performed through continuous interaction between the multi-agent system and the digital twin power grid environment, and the parameters of the actor network and the critic network are updated until the digital twin collaborative optimization model converges. The current state space is input into the actor network after training convergence. Each agent's actor network selects and outputs an action from the action space. The action is the total power adjustment instruction for the resource cluster.

[0012] Preferably, the process of performing secondary power allocation and generating charge / discharge control commands includes: Within the resource cluster that receives the total power adjustment command, a normalized allocation weight coefficient is calculated based on the health status parameters and physical connection stability characteristics of each electric vehicle. The total power regulation command is allocated to each electric vehicle in the resource cluster according to the normalized allocation weight coefficient, and a charging and discharging control command is generated for each electric vehicle. The charging and discharging control commands are sent to the corresponding charging pile execution terminal for execution; The data acquisition unit acquires the updated battery status data after executing the charge and discharge control command in real time, and uses the updated battery status data to update the health status parameters and resource clusters to form the state space for the next moment, thus forming closed-loop control.

[0013] Preferably, the process of visualizing physical connection stability characteristics, health status parameters, resource clusters, and charge / discharge control commands includes: Construct a digital twin 3D topology view, which is used to show the geographical distribution and connectivity of charging infrastructure and connected electric vehicles; In the digital twin 3D topology view, create an interactive 3D model for each connected electric vehicle; The physical connection stability characteristics, health status parameters, and charge / discharge control commands are data-bound to the corresponding 3D models; By integrating time-series graphs into a digital twin 3D topology view, the changing trend of physical connection stability characteristics over time can be demonstrated. By integrating a data dashboard into a digital twin 3D topology view, real-time values ​​of health status parameters and charge / discharge control commands are displayed. In the digital twin 3D topology view, a preset color mapping rule is used to color the 3D model in real time according to the resource cluster to which each electric vehicle belongs, so as to visualize the results of dynamic clustering.

[0014] A digital twin-based virtual power plant vehicle-to-grid resource optimization control system, used to implement the above methods, includes: The twin status acquisition module is used to acquire and preprocess video stream data from the charging port area and aggregate battery status data of electric vehicles to obtain preprocessed video stream data and battery status data. The physical health feature extraction module is used to input preprocessed video stream data into a three-dimensional convolutional neural network model, extract physical connection stability features, and establish an early warning mechanism. It also extracts thermal stress features and energy state features from preprocessed battery state data. The battery health assessment module is used to input thermal stress characteristics and energy state characteristics into the gated cyclic cell network model and output the battery's health state parameters. The dynamic resource clustering module is used to divide the connected electric vehicles into different resource clusters based on the physical connection stability characteristics and health status parameters using a density-based noise spatial clustering algorithm. The collaborative optimization decision module is used to take the aggregation parameters of resource clusters as input to the digital twin collaborative optimization model based on multi-agent deep reinforcement learning algorithm, perform distributed collaborative solution, obtain the total power adjustment command for each resource cluster, and perform secondary power allocation to generate charging and discharging control commands. The status monitoring visualization module is used to visualize the stability characteristics of physical connections, health status parameters, resource clusters, and charge / discharge control commands.

[0015] The beneficial effects of this invention are as follows: This invention achieves real-time quantitative evaluation of the physical connection status of the charging interface by introducing analysis of video stream data in the charging port area, and establishes an early warning mechanism to avoid issuing scheduling instructions to electric vehicles with unreliable physical connections, thereby improving the success rate of scheduling instruction execution and the operational safety of the system.

[0016] This invention achieves differentiated and refined control of batteries in different health states by accurately assessing the battery health status of each electric vehicle and dynamically clustering resources based on the assessment results. This avoids damage to the battery caused by indiscriminate charge and discharge control and effectively slows down the process of battery capacity decay.

[0017] This invention constructs a digital twin collaborative optimization model that integrates physical connection reliability with battery health status parameters, and uses a multi-agent deep reinforcement learning algorithm for distributed solution, thereby achieving global optimization scheduling of large-scale heterogeneous vehicle-to-grid resources and improving the overall control accuracy and response reliability of the virtual power plant. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0020] Example 1: As Figure 1 As shown, the virtual power plant vehicle-to-grid resource optimization control method based on digital twins includes the following steps: S1. Collect and preprocess the video stream data of the charging port area and aggregate the battery status data of the electric vehicle to obtain the preprocessed video stream data and battery status data.

[0021] Video capture devices are deployed on each charging station, with the devices positioned directly opposite the charging port area of ​​the connected electric vehicle, to acquire continuous video stream data of the charging port area.

[0022] A data acquisition unit is integrated into the control system of the charging pile. The data acquisition unit establishes a communication connection with the battery management system of the connected electric vehicle through the charging communication protocol to acquire battery status data of the connected electric vehicle, including battery state of charge data and battery surface temperature data.

[0023] Preprocessing of video stream data: Frame sampling is performed on the video stream data to obtain a video sequence; Perform image transformation on each frame of the video sequence to convert multi-channel color images into single-channel images; The pixel values ​​of a single-channel image are normalized to scale them to a standardized range, resulting in preprocessed video stream data.

[0024] Preprocess the battery status data: The collected battery status data is cleaned to remove missing and outlier values. A filtering algorithm based on a state estimation model is used to denoise the cleaned battery state data; The denoised battery state data is standardized to eliminate the influence of different physical quantities, resulting in preprocessed battery state data.

[0025] Align the timestamps of the preprocessed video stream data and battery status data.

[0026] S2. Input the preprocessed video stream data into a three-dimensional convolutional neural network model, extract physical connection stability features, and establish an early warning mechanism to extract thermal stress features and energy state features from the preprocessed battery state data.

[0027] The preprocessed video stream data is divided into video segments containing continuous image frames according to a preset time step, and used as input to a three-dimensional convolutional neural network model. The three-dimensional convolutional neural network model performs convolution operations on video segments using three-dimensional convolutional kernels to extract spatiotemporal features that include both time and spatial dimensions. The extracted spatiotemporal features are input into the fully connected regression layer of the 3D convolutional neural network model, and the output is a scalar value, which is defined as the physical connection stability feature. Set a safety threshold for physical connection stability, compare the value of the physical connection stability feature with the safety threshold in real time, and generate an early warning signal when the value of the physical connection stability feature is lower than the safety threshold, and remove the corresponding electric vehicle from the schedulable resource pool. Calculate the difference between the preprocessed battery surface temperature data at the current moment and the preprocessed battery surface temperature data at the previous sampling moment, and divide the difference by the time interval between the two sampling moments to obtain the thermal stress characteristics; Energy state features are extracted from the preprocessed battery state data, which are standardized battery state of charge data.

[0028] S3. Input the thermal stress characteristics and energy state characteristics into the gated cyclic unit network model, and output the battery health state parameters.

[0029] The thermal stress characteristics and energy state characteristics are combined into a two-dimensional input feature vector at each sampling time. The two-dimensional input feature vectors within continuous time steps are constructed into an input sequence with a time order relationship. The input sequence is fed into the gated recurrent unit network model. The gated recurrent unit network model processes the input sequence through its internal update and reset gates and updates the hidden state of the gated recurrent unit network model to capture the temporal dependencies in the input sequence. The reset gate in the gated recurrent unit network model determines the extent to which hidden state information from the previous time step is retained in the current time step; In the gated recurrent unit network model, the update gate determines whether the hidden state information from the previous time step and the new input information from the current time step are included in the hidden state at the current time step. The hidden state output by the gated recurrent unit network model after processing the input sequence is input into the fully connected output layer; The fully connected output layer outputs a two-dimensional vector. The first dimension of the two-dimensional vector represents the current capacity decay of the battery, and the second dimension represents the current increase in the internal resistance of the battery. The two-dimensional vector is defined as the battery's health state parameter.

[0030] S4. Based on the physical connection stability characteristics and health status parameters, a density-based noise application spatial clustering algorithm is used to divide the connected electric vehicles into different resource clusters.

[0031] The physical connection stability characteristics and health status parameters of each connected electric vehicle are combined into three-dimensional data points, and the three-dimensional data points from each connected electric vehicle are used to form a feature space. To apply a density-based spatial clustering algorithm to noise, we set the neighborhood radius parameter and the minimum number of neighborhood points for the core object. Calculate the Euclidean distance between two 3D data points in the feature space; For a 3D data point in the feature space, the number of 3D data points contained in the neighborhood of the 3D data point within the range of the neighborhood radius parameter is determined by comparing the Euclidean distance with the neighborhood radius parameter. If the number of 3D data points contained in the neighborhood of a 3D data point is not less than the minimum number of neighborhood points of the core object, then the 3D data point is marked as the core object. Starting with an unvisited core object, create a new cluster, find 3D data points that are density-reachable from the core object, and assign the found 3D data points to the new cluster; Repeat the process of creating new clusters from unvisited core objects, finding density-reachable 3D data points, and assigning density-reachable 3D data points to new clusters until all core objects have been accessed. The set of electric vehicles corresponding to the three-dimensional data points within a cluster is defined as a resource cluster.

[0032] S5. Using the aggregation parameters of resource clusters as input to the digital twin collaborative optimization model based on multi-agent deep reinforcement learning algorithm, perform distributed collaborative solution to obtain the total power adjustment command for each resource cluster, and perform secondary power allocation to generate charging and discharging control commands.

[0033] Resource clusters are defined as independent intelligent agents, and these agents constitute a multi-agent system. Construct a digital twin collaborative optimization model based on a multi-agent deep reinforcement learning algorithm; Each agent is given an independent actor network, and a central critic network is set up for the multi-agent system, forming an actor-critic architecture; Define the state space, action space, and reward function of a multi-agent system; The state space includes aggregated parameters for each resource cluster and global information about the power grid. Aggregated parameters include the average physical connectivity stability characteristics of electric vehicles within the cluster, average health state parameters, total charging power, and total available capacity. The action space of each agent is the continuous power adjustment range of the resource cluster; The reward function is used to evaluate the immediate reward obtained by each agent after performing an action during training. The immediate reward is calculated based on the grid frequency deviation and the expected battery loss of electric vehicles in the cluster. In the digital twin collaborative optimization model, training is performed through continuous interaction between the multi-agent system and the digital twin power grid environment, and the parameters of the actor network and the critic network are updated until the digital twin collaborative optimization model converges. The current state space is input into the actor network after training convergence. Each agent's actor network selects and outputs an action from the action space. The action is the total power adjustment instruction for the resource cluster. Within the resource cluster that receives the total power adjustment command, a normalized allocation weight coefficient is calculated based on the health status parameters and physical connection stability characteristics of each electric vehicle. The total power regulation command is allocated to each electric vehicle in the resource cluster according to the normalized allocation weight coefficient, and a charging and discharging control command is generated for each electric vehicle. The charging and discharging control commands are sent to the corresponding charging pile execution terminal for execution; The data acquisition unit acquires the updated battery status data after executing the charge and discharge control command in real time, and uses the updated battery status data to update the health status parameters and resource clusters to form the state space for the next moment, thus forming closed-loop control.

[0034] S6. Visualize physical connection stability characteristics, health status parameters, resource clusters, and charge / discharge control commands.

[0035] Construct a digital twin 3D topology view, which is used to show the geographical distribution and connectivity of charging infrastructure and connected electric vehicles; In the digital twin 3D topology view, create an interactive 3D model for each connected electric vehicle; The physical connection stability characteristics, health status parameters, and charge / discharge control commands are data-bound to the corresponding 3D models; By integrating time-series graphs into a digital twin 3D topology view, the changing trend of physical connection stability characteristics over time can be demonstrated. By integrating a data dashboard into a digital twin 3D topology view, real-time values ​​of health status parameters and charge / discharge control commands are displayed. In the digital twin 3D topology view, a preset color mapping rule is used to color the 3D model in real time according to the resource cluster to which each electric vehicle belongs, so as to visualize the results of dynamic clustering.

[0036] Example 2: Figure 2 As shown, the virtual power plant vehicle-to-grid resource optimization control system based on digital twins is used to implement the method in Example 1, including: The twin status acquisition module is used to acquire and preprocess video stream data from the charging port area and aggregate battery status data of electric vehicles to obtain preprocessed video stream data and battery status data.

[0037] The physical health feature extraction module is used to input preprocessed video stream data into a three-dimensional convolutional neural network model, extract physical connectivity stability features, establish an early warning mechanism, and extract thermal stress features and energy state features from preprocessed battery state data.

[0038] The battery health assessment module is used to input thermal stress characteristics and energy state characteristics into the gated cyclic cell network model and output the battery's health state parameters.

[0039] The dynamic resource clustering module is used to divide the connected electric vehicles into different resource clusters based on the physical connection stability characteristics and health status parameters, using a density-based noise-based spatial clustering algorithm.

[0040] The collaborative optimization decision module is used to take the aggregate parameters of resource clusters as input to the digital twin collaborative optimization model based on multi-agent deep reinforcement learning algorithm, perform distributed collaborative solution, obtain the total power adjustment command for each resource cluster, perform secondary power allocation, and generate charging and discharging control commands.

[0041] The status monitoring visualization module is used to visualize the stability characteristics of physical connections, health status parameters, resource clusters, and charge / discharge control commands.

Claims

1. A virtual power plant vehicle-to-grid resource optimization control method based on digital twins, characterized in that, Includes the following steps: The video stream data of the charging port area and the battery status data of the electric vehicle are collected and preprocessed to obtain the preprocessed video stream data and battery status data. The preprocessed video stream data is input into a three-dimensional convolutional neural network model to extract physical connection stability features and establish an early warning mechanism. Thermal stress features and energy state features are extracted from the preprocessed battery state data. The thermal stress characteristics and energy state characteristics are input into the gated cyclic unit network model, and the battery health state parameters are output. Based on the physical connection stability characteristics and health status parameters, a density-based noise-based spatial clustering algorithm is used to divide the connected electric vehicles into different resource clusters. Using the aggregation parameters of resource clusters as input to a digital twin collaborative optimization model based on a multi-agent deep reinforcement learning algorithm, a distributed collaborative solution is performed to obtain the total power adjustment command for each resource cluster, and a secondary power allocation is executed to generate charging and discharging control commands. Visualize physical connection stability characteristics, health status parameters, resource clusters, and charge / discharge control commands.

2. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of obtaining the preprocessed video stream data and battery status data includes: Video acquisition devices are deployed on each charging pile, with the devices positioned directly opposite the charging port area of ​​the connected electric vehicle, to acquire continuous video stream data of the charging port area. A data acquisition unit is integrated into the control system of the charging pile. The data acquisition unit establishes a communication connection with the battery management system of the connected electric vehicle through the charging communication protocol to acquire battery status data of the connected electric vehicle, including battery state of charge data and battery surface temperature data. Frame sampling is performed on the video stream data to obtain a video sequence; Perform image transformation on each frame of the video sequence to convert multi-channel color images into single-channel images; The pixel values ​​of a single-channel image are normalized to scale them to a standardized range, resulting in preprocessed video stream data. The collected battery status data is cleaned to remove missing and outlier values. A filtering algorithm based on a state estimation model is used to denoise the cleaned battery state data; The denoised battery state data is standardized to eliminate the influence of different physical quantities, resulting in preprocessed battery state data. Align the timestamps of the preprocessed video stream data and battery status data.

3. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of extracting physical connection stability features and establishing an early warning mechanism includes: The preprocessed video stream data is divided into video segments containing continuous image frames according to a preset time step, and used as input to a three-dimensional convolutional neural network model. The three-dimensional convolutional neural network model performs convolution operations on video segments using three-dimensional convolutional kernels to extract spatiotemporal features that include both time and spatial dimensions. The extracted spatiotemporal features are input into the fully connected regression layer of the 3D convolutional neural network model, and the output is a scalar value, which is defined as the physical connection stability feature. Set a safety threshold for physical connection stability, compare the value of the physical connection stability feature with the safety threshold in real time, and generate an early warning signal when the value of the physical connection stability feature is lower than the safety threshold, and remove the corresponding electric vehicle from the schedulable resource pool.

4. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 2, characterized in that, The process of extracting thermal stress features and energy state features from the preprocessed battery state data includes: Calculate the difference between the preprocessed battery surface temperature data at the current moment and the preprocessed battery surface temperature data at the previous sampling moment, and divide the difference by the time interval between the two sampling moments to obtain the thermal stress characteristics; Energy state features are extracted from the preprocessed battery state data, which are standardized battery state of charge data.

5. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of outputting the battery's health status parameters includes: The thermal stress characteristics and energy state characteristics are combined into a two-dimensional input feature vector at each sampling time. The two-dimensional input feature vectors within continuous time steps are constructed into an input sequence with a time order relationship. The input sequence is fed into the gated recurrent unit network model. The gated recurrent unit network model processes the input sequence through its internal update and reset gates and updates the hidden state of the gated recurrent unit network model to capture the temporal dependencies in the input sequence. The hidden state output by the gated recurrent unit network model after processing the input sequence is input into the fully connected output layer; The fully connected output layer outputs a two-dimensional vector, which is then defined as a battery health state parameter.

6. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of dividing the connected electric vehicles into different resource clusters includes: The physical connection stability characteristics and health status parameters of each connected electric vehicle are combined into three-dimensional data points, and the three-dimensional data points from each connected electric vehicle are used to form a feature space. To apply a density-based spatial clustering algorithm to noise, we set the neighborhood radius parameter and the minimum number of neighborhood points for the core object parameter. Calculate the Euclidean distance between two 3D data points in the feature space; For a 3D data point in the feature space, the number of 3D data points contained in the neighborhood of the 3D data point within the range of the neighborhood radius parameter is determined by comparing the Euclidean distance with the neighborhood radius parameter. If the number of 3D data points contained in the neighborhood of a 3D data point is not less than the minimum number of neighborhood points of the core object, then the 3D data point is marked as the core object. Starting with an unvisited core object, create a new cluster, find 3D data points that are density-reachable from the core object, and assign the found 3D data points to the new cluster; Repeat the process of creating new clusters from unvisited core objects, finding density-reachable 3D data points, and assigning density-reachable 3D data points to new clusters until all core objects have been accessed. The set of electric vehicles corresponding to the three-dimensional data points within a cluster is defined as a resource cluster.

7. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of performing distributed collaborative solving to obtain the total power adjustment command for each resource cluster includes: Resource clusters are defined as independent intelligent agents, and these agents constitute a multi-agent system. Construct a digital twin collaborative optimization model based on a multi-agent deep reinforcement learning algorithm; Each agent is given an independent actor network, and a central critic network is set up for the multi-agent system, forming an actor-critic architecture; Define the state space, action space, and reward function of a multi-agent system; The state space includes aggregated parameters for each resource cluster and global information about the power grid. Aggregated parameters include the average physical connectivity stability characteristics of electric vehicles within the cluster, average health state parameters, total charging power, and total available capacity. The action space of each agent is the continuous power adjustment range of the resource cluster; The reward function is used to evaluate the immediate reward obtained by each agent after performing an action during training. The immediate reward is calculated based on the grid frequency deviation and the expected battery loss of electric vehicles in the cluster. In the digital twin collaborative optimization model, training is performed through continuous interaction between the multi-agent system and the digital twin power grid environment, and the parameters of the actor network and the critic network are updated until the digital twin collaborative optimization model converges. The current state space is input into the actor network after training convergence. Each agent's actor network selects and outputs an action from the action space. The action is the total power adjustment instruction for the resource cluster.

8. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of performing secondary power allocation and generating charge / discharge control commands includes: Within the resource cluster that receives the total power adjustment command, a normalized allocation weight coefficient is calculated based on the health status parameters and physical connection stability characteristics of each electric vehicle. The total power regulation command is allocated to each electric vehicle in the resource cluster according to the normalized allocation weight coefficient, and a charging and discharging control command is generated for each electric vehicle. The charging and discharging control commands are sent to the corresponding charging pile execution terminal for execution; The data acquisition unit acquires the updated battery status data after executing the charge and discharge control command in real time, and uses the updated battery status data to update the health status parameters and resource clusters to form the state space for the next moment, thus forming closed-loop control.

9. The method for optimizing and controlling virtual power plant vehicle-to-grid resources based on digital twins according to claim 1, characterized in that, The process of visualizing physical connection stability characteristics, health status parameters, resource clusters, and charge / discharge control commands includes: Construct a digital twin 3D topology view, which is used to show the geographical distribution and connectivity of charging infrastructure and connected electric vehicles; In the digital twin 3D topology view, create an interactive 3D model for each connected electric vehicle; The physical connection stability characteristics, health status parameters, and charge / discharge control commands are data-bound to the corresponding 3D models; By integrating time-series graphs into a digital twin 3D topology view, the changing trend of physical connection stability characteristics over time can be demonstrated. By integrating a data dashboard into a digital twin 3D topology view, real-time values ​​of health status parameters and charge / discharge control commands are displayed. In the digital twin 3D topology view, a preset color mapping rule is used to color the 3D model in real time according to the resource cluster to which each electric vehicle belongs, so as to visualize the results of dynamic clustering.

10. A virtual power plant vehicle-to-grid resource optimization control system based on digital twins, used to implement the method described in any one of claims 1-9, characterized in that, include: The twin status acquisition module is used to acquire and preprocess video stream data from the charging port area and aggregate battery status data of electric vehicles to obtain preprocessed video stream data and battery status data. The physical health feature extraction module is used to input preprocessed video stream data into a three-dimensional convolutional neural network model, extract physical connection stability features, and establish an early warning mechanism. It also extracts thermal stress features and energy state features from preprocessed battery state data. The battery health assessment module is used to input thermal stress characteristics and energy state characteristics into the gated cyclic cell network model and output the battery's health state parameters. The dynamic resource clustering module is used to divide the connected electric vehicles into different resource clusters based on the physical connection stability characteristics and health status parameters using a density-based noise spatial clustering algorithm. The collaborative optimization decision module is used to take the aggregation parameters of resource clusters as input to the digital twin collaborative optimization model based on multi-agent deep reinforcement learning algorithm, perform distributed collaborative solution, obtain the total power adjustment command for each resource cluster, and perform secondary power allocation to generate charging and discharging control commands. The status monitoring visualization module is used to visualize the stability characteristics of physical connections, health status parameters, resource clusters, and charge / discharge control commands.

Images