A Physical Information Neural Network-Based Optimal Scheduling Method for V2G-Integrated Virtual Power Plants

Abstract

This invention discloses an optimal scheduling method for virtual power plants containing V2G based on a physical information neural network, belonging to the field of virtual power plant optimal scheduling. The method includes: constructing a virtual power plant aggregation system architecture comprising renewable energy generation equipment and V2G electric vehicle charging stations; determining the optimal scheduling problem for virtual power plants containing V2G electric vehicles; constructing an objective function based on the optimal scheduling problem, with the goal of maximizing user benefits; determining the physical constraints of the optimal scheduling problem, including power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; constructing a physical information neural network, embedding the objective function and the physical constraints into a loss function, training the physical information neural network, and solving for the optimal scheduling strategy for virtual power plants containing V2G. This method achieves optimal scheduling of virtual power plants containing electric vehicles, taking into account multiple needs such as user benefits, renewable energy consumption, and grid security.

Description

Technical Field

[0001] This invention belongs to the field of virtual power plant optimization scheduling, specifically involving a V2G virtual power plant optimization scheduling method based on physical information neural network. Background Technology

[0002] As distributed energy resources are gradually integrated into the power grid, the dispatching mode and stability of traditional power systems face challenges. Virtual power plants, leveraging new information and intelligent control technologies, aggregate dispersed distributed resources, promote their joint operation and optimize regulation, and serve as a fundamental platform for the system and an important carrier for high-level management of demand-side resources.

[0003] Currently, vehicle-to-grid (V2G) technology for electric vehicles (EVs) is showing a trend of large-scale application. When a large number of distributed EV resources are integrated into the power grid, the grid load may fluctuate due to the disordered charging and discharging of these vehicles, threatening the stable operation of the power grid. For EV clusters integrating V2G functions in virtual power plants, it is necessary to build a multi-objective collaborative optimization scheduling system. On the one hand, this involves the spatiotemporal reconfiguration of charging and discharging resources to create quantifiable economic gains for the user side; on the other hand, it involves improving the absorption rate of new energy sources and enhancing the basic performance of virtual power plants and the performance indicators of grid interaction.

[0004] Traditional mathematical models struggle to accurately depict the charging and discharging load of EV charging stations, stemming from the high complexity resulting from the aggregation of numerous random behaviors of EV users. Current technologies for optimizing the scheduling of V2G virtual power plants often employ conventional optimization algorithms, but these methods fall short in handling complex constraints and uncertainties, failing to meet the demands for efficient and precise scheduling. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide an optimized scheduling method for virtual power plants with V2G based on physical information neural networks, thereby solving the problems in existing technologies.

[0006] The objective of this invention can be achieved through the following technical solutions: A method for optimal scheduling of virtual power plants with V2G based on physical information neural networks includes: Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and determine the optimization scheduling problem of virtual power plants with V2G electric vehicles. Based on the virtual power plant optimization scheduling problem, an objective function is constructed with the goal of maximizing user benefits; Determine the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; A physical information neural network is constructed, and the objective function and the physical constraints are embedded in the loss function. The physical information neural network is trained and the optimal scheduling strategy containing V2G virtual power plants is obtained by solving the problem.

[0007] Furthermore, the virtual power plant aggregates new energy power generation equipment and electric vehicle charging stations with vehicle-to-grid interaction capabilities, and performs intraday optimized scheduling within a period T; the new energy power generation equipment includes photovoltaic equipment and wind power equipment.

[0008] Furthermore, the objective function is: in, This represents the objective function value representing the overall benefit for all electric vehicle users in a distributed energy system. For the discharge revenue of electric vehicles, The cost of charging electric vehicles, The reward coefficient for the amount of waste disposed of. This refers to the total amount of new energy consumed. It refers to the set of individual electric vehicles participating in the scheduling and control of a distributed energy system. The penalty coefficient is... The state of charge required for the user's next trip. For the first When electric vehicles are actually disconnected from the grid value; and They are respectively Time of the first The charging and discharging power of an electric vehicle yes Real-time electricity price for grid connection It is an electric car Discharge subsidy at any time, The scheduling period is [number].

[0009] Furthermore, the total amount of new energy consumed The formula for calculation is: in, This refers to the total power generated by new energy sources in a distributed energy system. and These are the power generation capacities of photovoltaic and wind power, respectively. For the total load demand of distributed energy systems, To fix the foundation load, This refers to the amount of new energy consumed.

[0010] Furthermore, the power balance constraint is: in, For the first The net power output of an electric vehicle to the power grid. Current time Download demand; The equipment capacity constraint is: in, , This represents the maximum power generation capacity of new energy photovoltaic and wind power. , They are the first The maximum charging and discharging power of a vehicle; The constraints on the electric vehicle equipment are: in, , These are the upper and lower limits of SOC, respectively. For the time before departure, For state variables, , These represent the time periods for connecting to and disconnecting from the power grid.

[0011] Furthermore, the loss function of the physical information neural network is: in, Loss to the target For power balance loss, For SOC range loss, , , The weights of each loss term.

[0012] Furthermore, the physical information neural network adopts a multi-layer fully connected neural network, including an input layer, a hidden layer, and an output layer. The input data is normalized and then sent to the input layer for weight mapping and linear combination of features to obtain an initial feature vector. Then, the hidden layer performs nonlinear activation function mapping and deep physical feature extraction to obtain high-dimensional abstract features containing the physical constraints of the power system. Finally, the output layer outputs the optimized scheduling strategy.

[0013] The V2G-based virtual power plant optimization scheduling device based on physical information neural network, executing the above method, is characterized by including: Problem identification module: Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and identify the virtual power plant optimization scheduling problem involving V2G electric vehicles; Objective function construction module: Based on the virtual power plant optimization scheduling problem, an objective function is constructed with the goal of maximizing user benefits; Constraint Construction Module: Determines the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; Strategy solving module: Construct a physical information neural network, embed the objective function and the physical constraints into the loss function, train the physical information neural network, and solve for the optimal scheduling strategy containing V2G virtual power plants.

[0014] A computer storage medium storing a readable program that, when executed, instructs a computing device to perform the above-described method for optimizing the scheduling of a virtual power plant with V2G based on a physical information neural network.

[0015] An electronic device includes: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform operations corresponding to the above-described optimized scheduling method for V2G virtual power plants based on physical information neural networks.

[0016] The beneficial effects of this invention are: This invention focuses on the optimization of Physical Information Neural Network (PINN). By deeply integrating physical constraints with data-driven approaches, it solves the problems of "complex and difficult-to-solve models" and "poor real-time performance" in traditional virtual power plant optimization scheduling. The PINN network can directly embed physical rules such as power balance and SOC safety without the need to build additional complex constraint solvers, significantly improving scheduling decision efficiency. Attached Figure Description

[0017] 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, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the optimized scheduling method of the present invention; Figure 2 This data represents the electric vehicle access time of charging stations in a certain location. Figure 3 This is a graph showing the trend of total loss during the training process of the PINN algorithm. Figure 4 To optimize the comparison of EV user revenue before and after scheduling; Figure 5 A comparative analysis of load curves before and after optimization for the PINN network. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 like Figure 1 As shown, the optimized scheduling method for V2G-enabled virtual power plants based on physical information neural networks includes the following steps: S1. Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and determine the virtual power plant optimization scheduling problem that includes V2G electric vehicles. In the virtual power plant distributed energy aggregation system described in this invention, multiple distributed energy sources are aggregated, including new energy power generation equipment such as photovoltaic and wind power, as well as electric vehicle charging stations with vehicle-to-grid (V2G) interaction capabilities. To simplify the optimization scheduling problem, the charging and discharging of energy storage devices are not considered in this part of the study. Intraday optimization scheduling is performed within a scheduling cycle T (set as one day, T=24h) to optimize EV charging and discharging strategies, maximize the total user benefits (charging cost savings, discharging benefits) while maximizing the absorption of new energy sources, and ensuring grid security.

[0021] S2, based on the virtual power plant optimization scheduling problem determined in S1, constructs an objective function with the goal of maximizing user benefits; In S2, with the goal of maximizing user benefits and considering factors such as renewable energy consumption, and combining multiple objectives such as economic benefits and the needs of electric vehicle users, and taking into account the characteristics of distributed energy, a comprehensive optimization objective is established to maximize the total benefits of all individual EV users in the distributed energy system; the specific process includes: S21, Design the overall objective optimization function. In the virtual power plant distributed energy aggregation system described in this invention, multiple distributed energy sources are aggregated, including new energy power generation equipment such as photovoltaic and wind power, as well as electric vehicle charging stations with vehicle-to-grid interaction capabilities. To simplify the optimization scheduling problem, the charging and discharging of energy storage devices are not considered in this part of the study. Intraday optimization scheduling is performed within a scheduling period T to optimize the EV charging and discharging strategy, maximize the total user benefits (charging cost savings, discharging benefits) while maximizing the absorption of new energy sources, and ensuring grid security.

[0022] With the goal of maximizing user benefits and considering factors such as renewable energy consumption, this paper combines multiple objectives including economic efficiency and the needs of electric vehicle users, and takes into account the characteristics of distributed energy. It establishes a comprehensive optimization objective that maximizes the total benefit for all individual EV users in the distributed energy system. The objective function is... The objective function representing the overall benefit to all electric vehicle users in a distributed energy system is written as: (1) in, For the discharge revenue of electric vehicles, The cost of charging electric vehicles, The reward coefficient for the amount of waste disposed of. This refers to the total amount of new energy consumed. It refers to the set of individual electric vehicles participating in the scheduling and control of a distributed energy system. The penalty coefficient is... The state of charge required for the user's next trip. For the first When electric vehicles are actually disconnected from the grid value; S22 defines the detailed calculation formulas for each part of the objective function; (1) Costs and benefits of charging and discharging electric vehicles Electric vehicle charging costs and discharge benefits The calculation formulas are as follows: (2) (3) in, and They are respectively Time of the first The charging and discharging power of an electric vehicle yes Real-time electricity price for grid connection It is an electric car Discharge subsidy at any time, The scheduling period is [number].

[0023] (2) Allocation of new energy consumption Total power generation of new energy sources in distributed energy systems for: (4) in, and These are the power generation capacities of photovoltaic and wind power, respectively. Assuming the total load demand of the distributed energy system for: (5) in, For fixed foundation load.

[0024] Regarding the consumption and distribution of new energy sources, if If all EV charging is powered by new energy sources, then the amount of new energy consumed will be... for: (6) Otherwise, new energy sources will be allocated proportionally, with the EV consumption capacity as follows: (7) Right now: (8) The total renewable energy consumption is the cumulative value within the dispatch cycle: (9) (3) Penalties for users' charging needs In the scheduling model, the SOC penalty term directly reflects the actual usage needs of electric vehicle users, ensuring the experience needs of EV users and serving as a key physical constraint for the feasibility of the scheduling strategy.

[0025] The SOC penalty term in the objective function is defined as follows: (10) in, The state of charge required for the user's next trip; For the first The State of Charge (SOC) value of an electric vehicle when it is actually disconnected from the grid; The penalty coefficient represents the economic loss caused by each unit of SOC failing to meet the standard.

[0026] S3, determine the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints and electric vehicle equipment constraints; The main constraints considered in the optimization scheduling problem are determined. In the optimization scheduling problem studied in this invention, physical constraints are an important basis for constructing the core physical constraint layer of the PINN network. The main constraints considered in this invention include: power balance constraints, device capacity constraints, and EV device constraints. The specific details are as follows: (1) Power balance constraint At every moment Under this system, the distributed energy generation capacity within the distributed energy aggregation system should be balanced with load demand and the charging and discharging power of EVs. That is: (11) in, and These are the power generation capacities of photovoltaic and wind power, respectively. For the first The net power output of an electric vehicle to the power grid. Current time Download demand.

[0027] (2) Equipment capacity constraints The power generation capacity of new energy photovoltaic and wind power cannot exceed their maximum power generation capacity, which can be expressed mathematically as: (12) in, , This represents the maximum power generation capacity of new energy photovoltaic and wind power.

[0028] Each electric vehicle has a permissible range for charging and discharging power, namely: (13) in, , They are the first The maximum charging and discharging power of an electric vehicle.

[0029] (3) EV equipment constraints In this invention, the state of charge (SBC) of each electric vehicle's battery must be within a reasonable range and must meet the user's needs when the vehicle leaves. Furthermore, each electric vehicle can only perform charging and discharging operations within its connection and departure time windows. That is: (14) in, , These are the upper and lower limits of SOC, respectively. For the time before departure, For state variables, , These represent the time periods for connecting to and disconnecting from the power grid.

[0030] By embedding the aforementioned physical constraints into the network's loss function, both physical constraints and data-driven optimization can be achieved.

[0031] S4. Construct a physical information neural network, and embed the objective function of S2 and the physical constraints of S3 into the loss function. Train the physical information neural network and solve for the optimal scheduling strategy containing the V2G virtual power plant.

[0032] This step constructs a multi-layered objective function with maximizing user revenue as the primary objective, maximizing renewable energy consumption as the secondary objective, and user charging demand penalty as the constraint, and transforms it into minimizing the loss term; the Adam optimizer and a dynamic learning rate decay strategy are used to train the network, and the weight parameters are updated through backpropagation until the loss function converges, outputting the optimal charging and discharging strategy for V2G electric vehicles; thus realizing the intraday optimized scheduling of the virtual power plant aggregation system.

[0033] The specific methods for optimizing scheduling strategies based on PINN neural networks include: Based on the architecture of the physical information neural network, S41 normalizes the input data features and ultimately constructs the input feature vector for each EV. .

[0034] (15) This includes: the current moment Load data Photovoltaic power generation Wind power generation capacity Initial battery level of each electric vehicle Access time for each electric vehicle and departure time Electricity price and discharge subsidy The number of neurons in the input layer depends on the dimensionality of the input information, as shown in Equation (15). Each EV corresponds to 9 input features: S42, Define the PINN network structure. PINN uses a multi-layer fully connected neural network, including an input layer, hidden layers, and an output layer. The mapping relationship between each layer is as follows: The input layer accepts a normalized feature vector matrix. The matrix has 9 nodes. ( (Number of EVs), number of hidden layers Set as 3 floors, the first layer( The input of ) is the output of the previous layer. Output The calculation formula is: (16) in, For the first Layer weight matrix (dimension: , For the first Number of neurons in a layer, in this invention ), For bias vectors, ReLU activation function ( ); The output layer outputs the charging and discharging power of each EV, with a node count of 2. The calculation formula is: (17) in This is the output layer weight matrix. This is the output layer bias vector.

[0035] S43, based on physical constraints, is transformed into a physical constraint layer embedded PINN structure.

[0036] The constructed physical constraint layer further includes: (18) (19) in, For power balance loss, the power balance constraint is transformed into part of the loss function, which is minimized during training. This ensures that the network output meets power balance constraints; The SOC security loss is a loss function composed of two parts: the SOC range constraint and the EV off-grid SOC demand constraint.

[0037] S44, based on the objective function defined in S2, constructs the network composite loss function, further including: (20) (twenty one) (twenty two) (twenty three) (twenty four) in For discharge benefits, For charging costs, This refers to the total amount of new energy consumed. and These are discharge subsidies and time-of-use electricity prices, respectively. To reduce the incentive coefficient, This is the penalty coefficient; , and The loss weight is used to balance the priority between objective optimization and physical constraint satisfaction.

[0038] Based on a similar inventive concept, embodiments of the present invention also provide a computer storage medium storing a readable program that, when run by a processor, can execute the above-described optimized scheduling method for V2G-based virtual power plants using a physical information neural network.

[0039] Based on a similar inventive concept, this invention provides an electronic device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described optimized scheduling method for V2G virtual power plants based on physical information neural networks.

[0040] Based on a similar inventive concept, embodiments of the present invention also provide a computer program product, including computer instructions, which instruct a computing device to perform the operations corresponding to the above-described optimized scheduling method for V2G virtual power plants based on a physical information neural network.

[0041] Example 2 In this embodiment, the superiority of the optimized scheduling method proposed in Embodiment 1 is verified in a specific instance; Based on the electric vehicle access time data of charging piles in a certain area ( Figure 2 The PINN network uses Monte Carlo sampling to obtain EV access times and then schedules and controls the charging and discharging behavior of electric vehicles within the virtual power plant. The PINN network parameters are set as follows: 3 hidden layers with 500 neurons each, an initial learning rate of 0.001, dynamically updated with network iterations; and 5000 training iterations. Figure 3The curve showing the change in the total loss function value as a function of the number of training iterations reveals that in the early stages of training (0 to 1000 iterations), the total loss function value drops rapidly from its initial high value, with a very steep curve. This indicates that the model possesses efficient convergence performance, enabling the PINN network to quickly capture the physical constraints and data characteristics of electric vehicle charging and discharging behavior. As the number of iterations increases to the mid-to-late stages (2000 to 5000 iterations), the rate of decrease in the loss value slows down, but still maintains an overall downward trend without significant oscillations or gradient vanishing. This dynamically updated learning rate ensures refined training, validating the effectiveness of the patented strategy of "dynamically updating the learning rate with network iterations." This successfully avoids the model getting trapped in local optima and ensures that even with a complex architecture of 3 hidden layers and 500 neurons per layer, the network can still be fine-tuned. When the number of iterations approaches 5000, the total loss function value drops to a lower order of magnitude and tends to stabilize, with a significantly reduced curve fluctuation, indicating that the model achieves extremely high computational accuracy and robustness. This conclusion further confirms that the EV access time data obtained through Monte Carlo sampling has been fully trained, and the PINN model used has achieved a deep integration of the randomness of electric vehicle access and the scheduling constraints of virtual power plants, which can provide highly reliable decision support for the subsequent regulation of charging and discharging behavior.

[0042] To verify the effectiveness of the optimized scheduling, this embodiment constructs a random EV access scenario based on charging pile EV access data. It compares and analyzes the income differences and load curves of EV user groups after random charging / discharging and after optimized scheduling. The results are as follows: Figure 4 as well as Figure 5 As shown in the figure, in the random charging and discharging scenario, most electric vehicles only consider charging demand, and the number of charging users further increases after 18:00. After optimized scheduling, the overall income of the electric vehicle group significantly improves, but income decreases after 18:00. After optimized scheduling, both the standard deviation and mean load of the optimized load decrease to a certain extent. The standard deviation of the load decreases from the initial 92.8259 to 77.2773, an overall decrease of approximately 16.75%, indicating that the PINN network effectively reduces the dispersion of load data, and the load fluctuation is more stable. At the same time, the mean load also decreases from 235.66kW to 171.41kW, indicating that the optimization strategy also plays a certain role in regulating the load overall.

[0043] Example 3 In this embodiment, a V2G-based virtual power plant optimization scheduling device based on physical information neural networks is proposed, specifically including: Problem identification module: Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and identify the virtual power plant optimization scheduling problem involving V2G electric vehicles; Objective function construction module: Based on the virtual power plant optimization scheduling problem, an objective function is constructed with the goal of maximizing user benefits; Constraint Construction Module: Determines the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; Strategy solving module: Construct a physical information neural network, embed the objective function and the physical constraints into the loss function, train the physical information neural network, and solve for the optimal scheduling strategy containing V2G virtual power plants.

[0044] The methods of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be processed by software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the methods described herein. Furthermore, when a general-purpose computer accesses the code used to implement the methods shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for performing the methods shown herein.

[0045] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for optimal scheduling of virtual power plants with V2G based on physical information neural networks, characterized in that, include: Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and determine the optimization scheduling problem of virtual power plants with V2G electric vehicles. Based on the virtual power plant optimization scheduling problem, an objective function is constructed with the goal of maximizing user benefits; Determine the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; A physical information neural network is constructed, and the objective function and the physical constraints are embedded in the loss function. The physical information neural network is trained and the optimal scheduling strategy containing V2G virtual power plants is obtained by solving the problem.

2. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 1, characterized in that, The virtual power plant aggregates new energy power generation equipment and electric vehicle charging stations with vehicle-to-grid interaction capabilities, and performs intraday optimized scheduling within a period T; the new energy power generation equipment includes photovoltaic equipment and wind power equipment.

3. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 1, characterized in that, The objective function is: in, This represents the objective function value representing the overall benefit for all electric vehicle users in a distributed energy system. For the discharge revenue of electric vehicles, The cost of charging electric vehicles, The reward coefficient for the amount of waste disposed of. This refers to the total amount of new energy consumed. It refers to the set of individual electric vehicles participating in the scheduling and control of a distributed energy system. The penalty coefficient is... The state of charge required for the user's next trip. For the first When electric vehicles are actually disconnected from the grid value; and They are respectively Time of the first The charging and discharging power of electric vehicles yes Real-time electricity price for grid connection It is an electric car Discharge subsidy at any time, The scheduling period is [number].

4. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 3, characterized in that, The total amount of new energy consumption The formula for calculation is: in, This refers to the total power generated by new energy sources in a distributed energy system. and These are the power generation capacities of photovoltaic and wind power, respectively. For the total load demand of distributed energy systems, To fix the foundation load, This refers to the amount of new energy consumed.

5. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 4, characterized in that, The power balance constraint is: in, For the first The net power output of an electric vehicle to the power grid. Current time Download demand; The equipment capacity constraint is: in, , This represents the maximum power generation capacity of new energy photovoltaic and wind power. , They are the first The maximum charging and discharging power of a vehicle; The constraints on the electric vehicle equipment are: in, , These are the upper and lower limits of SOC, respectively. For the time before departure, For state variables, , These represent the time periods for connecting to and disconnecting from the power grid.

6. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 5, characterized in that, The loss function of the physical information neural network is: in, Loss to the target For power balance loss, For SOC range loss, , , The weights of each loss term.

7. The optimized scheduling method for V2G-based virtual power plants based on physical information neural networks according to claim 1, characterized in that, The physical information neural network adopts a multi-layer fully connected neural network, which includes an input layer, a hidden layer and an output layer. The input data is normalized and then sent to the input layer for weight mapping and linear combination of features to obtain an initial feature vector. After that, the hidden layer performs non-linear activation function mapping and deep physical feature extraction to obtain high-dimensional abstract features containing the physical constraints of the power system. Finally, the optimized scheduling strategy is output through the output layer.

8. A V2G-based virtual power plant optimization scheduling device based on a physical information neural network, performing the method described in any one of claims 1-7, characterized in that, include: Problem identification module: Construct a virtual power plant aggregation system architecture that includes new energy power generation equipment and V2G electric vehicle charging stations, and identify the virtual power plant optimization scheduling problem involving V2G electric vehicles; Objective function construction module: Based on the virtual power plant optimization scheduling problem, an objective function is constructed with the goal of maximizing user benefits; Constraint Construction Module: Determines the physical constraints of the optimization scheduling problem, including: power balance constraints, equipment capacity constraints, and electric vehicle equipment constraints; Strategy solving module: Construct a physical information neural network, embed the objective function and the physical constraints into the loss function, train the physical information neural network, and solve for the optimal scheduling strategy containing V2G virtual power plants.

9. A computer storage medium storing a readable program, characterized in that, When the program runs, it can instruct the computing device to execute the V2G virtual power plant optimization scheduling method based on physical information neural network as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the V2G virtual power plant optimization scheduling method based on physical information neural network as described in any one of claims 1-7.

Images