Electric vehicle and power grid combined dispatching method, device and equipment and storage medium

Abstract

The invention provides an electric vehicle and power grid combined dispatching method, device and equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps that an electric vehicle and power grid combined dispatching problem is coded into a bigraph structure, the electric vehicle and power grid combined dispatching problem is modeled into an MISOCP, the MISOCP is limited by target constraints, and the target constraints comprise a power distribution network constraint and a vehicle-network cooperation constraint; according to the bipartite graph structure, scheduling prediction is carried out, and a scheduling prediction probability is obtained; constructing a confidence domain according to the scheduling prediction probability; and solving the sub-problem determined by the confidence domain in the confidence domain by taking minimization of power generation cost and minimization of electric vehicle energy interaction cost as targets, and obtaining scheduling parameters in the scheduling period. According to the scheme, the scheduling parameters are solved in the confidence domain, so that the calculation complexity of electric vehicle and power grid joint scheduling can be reduced, and the calculation overhead is reduced.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a method, apparatus, equipment, and storage medium for joint scheduling of electric vehicles and power grids. Background Technology

[0002] With the widespread adoption of electric vehicles (EVs) and the continued advancement of clean energy policies, the demand for unified scheduling and control of EV charging and discharging behavior by the power grid is increasing. Traditional EV scheduling typically handles optimal power flow (OPF) and vehicle-to-grid (V2G) behavior separately, employing a mixed-integer programming model to jointly model and optimize EV charging and discharging strategies and grid operating states. This approach focuses on power system power flow balance, considering physical constraints such as node voltage, power limits, and state of charge (SOC) boundaries, and solves the system cost minimization problem through global optimization. In this process, the EV state is modeled using integer variables, forming a typical mixed-integer second-order cone programming (MISOCP) problem. However, due to the large scale, numerous variables, and strong nonlinearity of the OPF-V2G problem, traditional solvers (such as Solving Constraint Integer Programs (SCIP)) need to traverse a large state space during the solution process, resulting in extremely high computational overhead and making it difficult to meet real-time requirements. Summary of the Invention

[0003] The purpose of this application is to provide a method, device, equipment and storage medium for joint dispatching of electric vehicles and power grids, which solves the problems of high computational complexity and large computational overhead in joint dispatching of electric vehicles and power grids.

[0004] To achieve the above objectives, embodiments of this application provide a method for joint dispatching of electric vehicles and the power grid, comprising:

[0005] The joint scheduling problem of electric vehicles and power grid is encoded as a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include distribution network constraints and vehicle-grid coordination constraints.

[0006] Based on the bipartite graph structure, a scheduling prediction is performed to obtain the scheduling prediction probability;

[0007] Construct a confidence region based on the scheduling prediction probability;

[0008] Within the confidence region, the sub-problems determined by the confidence region are solved with the objectives of minimizing power generation costs and minimizing electric vehicle energy interaction costs to obtain scheduling parameters within the scheduling cycle;

[0009] The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

[0010] Optionally, the scheduling parameters include at least one of the following: generator power, generator voltage, electric vehicle charging capacity, electric vehicle discharging capacity, electric vehicle remaining capacity, electric vehicle charging power, electric vehicle discharging power, and electric vehicle charging / discharging mode.

[0011] Optionally, encoding the joint scheduling problem of electric vehicles and the power grid into a bipartite graph structure includes:

[0012] Construct a set of variable nodes and a set of constraint nodes based on power grid physical nodes and electric vehicle aggregation units;

[0013] The continuous variables in optimal power flow calculation and the integer variables in electric vehicle scheduling are mapped to the node attributes of a bipartite graph, and the target constraints are represented as edge connections across sets to obtain a bipartite graph structure with nodes and edges.

[0014] The integer variable is used to represent the charging and discharging status of the electric vehicle in each scheduling cycle, and the continuous variable is used to describe at least one of the following: power flow, voltage amplitude, and remaining power level.

[0015] Optionally, the step of performing scheduling prediction based on the bipartite graph structure and obtaining the scheduling prediction probability includes:

[0016] A graph neural network (GCNN) is used to perform embedding representation learning on the bipartite graph structure to extract spatial structure information between variables and constraints.

[0017] Based on spatial structure information, the Long Short-Term Memory (LSTM) network is used to process the dynamic scheduling behavior in the time dimension and obtain the scheduling prediction probability for each time step.

[0018] Optionally, the step of using a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph structure and extracting spatial structure information between variables and constraints includes:

[0019] In the bipartite graph structure, the information of the first constraint node is aggregated with the information of its adjacent variable nodes and edges, and the embedded representation of the first constraint node is updated.

[0020] In the bipartite graph structure, the information of the first variable node is aggregated with the information of its adjacent constraint nodes and edges, and the embedded representation of the first variable node is updated.

[0021] Based on the embedded representation of the first constraint node and the embedded representation of the first variable node, the spatial structure information between variables and constraints is determined;

[0022] Wherein, the first constraint node is any node in the set of constraint nodes corresponding to the bipartite graph structure, and the first variable node is any node in the set of variable nodes corresponding to the bipartite graph structure.

[0023] Optionally, the step of using a Long Short-Term Memory (LSTM) network to process the dynamic scheduling behavior in the time dimension based on spatial structure information and obtaining the scheduling prediction probability for each time step includes:

[0024] The control variables of each electric vehicle in each time period and the spatial structure information of the charging station node corresponding to it are input into the LSTM model in ascending order of time to capture the cross-time state evolution and constraint transmission of electric vehicles and obtain the scheduling prediction probability at each time step.

[0025] Optionally, constructing the confidence region based on the scheduling prediction probability includes:

[0026] Get the k positions with the highest predicted probability among all the scheduling prediction probabilities at each time step, where k is an integer greater than or equal to 1;

[0027] An initial solution is constructed based on the patterns corresponding to the k positions, where each position corresponds to one of the charging mode, discharging mode, or idle mode.

[0028] The confidence radius is determined based on the target norm according to the initial solution, and the confidence region is obtained.

[0029] Optionally, the distribution network constraints include at least one of the following:

[0030] Power flow constraints, power balance constraints, voltage amplitude constraints, current amplitude constraints, and equipment capacity and operation limitations constraints;

[0031] and / or

[0032] The vehicle-to-everything (V2X) coordination constraints include at least one of the following:

[0033] Charge / discharge state constraints, charge / discharge power constraints, and battery capacity constraints.

[0034] This application also provides an electric vehicle and power grid joint dispatching device, including:

[0035] The encoding module is used to encode the joint scheduling problem of electric vehicles and power grid into a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include distribution network constraints and vehicle-grid coordination constraints.

[0036] The first acquisition module is used to perform scheduling prediction based on the bipartite graph structure and acquire the scheduling prediction probability.

[0037] The construction module is used to construct a confidence region based on the scheduling prediction probability;

[0038] The second acquisition module is used to solve the sub-problems determined by the confidence domain within the confidence domain with the objectives of minimizing power generation cost and minimizing electric vehicle energy interaction cost, and to obtain the scheduling parameters within the scheduling period;

[0039] The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

[0040] This application embodiment also provides an electric vehicle and power grid joint dispatching device, including: a transceiver, a processor, a memory, and a program or instructions stored in the memory and executable on the processor; when the processor executes the program or instructions, it implements the above-described electric vehicle and power grid joint dispatching method.

[0041] This application also provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps in the above-described electric vehicle and power grid joint scheduling method.

[0042] The beneficial effects of the above technical solution in this application are as follows:

[0043] The method in this application embodiment encodes the joint scheduling problem of electric vehicles and the power grid into a bipartite graph structure. Based on the bipartite graph structure, scheduling prediction is performed to obtain scheduling prediction probabilities. Based on the scheduling prediction probabilities, a confidence region is constructed. Within the confidence region, the sub-problems determined by the confidence region are solved with the objectives of minimizing generation costs and minimizing the energy interaction costs of electric vehicles. The scheduling parameters within the scheduling period are obtained. By solving the scheduling parameters in the confidence region, the computational complexity and computational overhead of joint scheduling of electric vehicles and the power grid can be reduced. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the electric vehicle and power grid joint dispatching method according to an embodiment of this application;

[0045] Figure 2 This is a schematic diagram illustrating the overall framework of the implementation principle of the embodiments of this application;

[0046] Figure 3 A schematic diagram of the overall structure for obtaining the bipartite graph structure;

[0047] Figure 4This is a schematic diagram of the modules of the electric vehicle and power grid joint dispatching device according to an embodiment of this application;

[0048] Figure 5 This is a schematic diagram of the structure of the electric vehicle and power grid joint dispatching equipment according to an embodiment of this application. Detailed Implementation

[0049] To make the technical problems, technical solutions and advantages of this application clearer, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments.

[0050] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0051] In the various embodiments of this application, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0052] In addition, the terms "system" and "network" are often used interchangeably in this article.

[0053] In the embodiments provided in this application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean that B is determined solely based on A, but can also be determined based on A and / or other information.

[0054] like Figure 1 As shown in the figure, an embodiment of this application provides a method for joint dispatching of electric vehicles and the power grid, comprising:

[0055] Step S101: Encode the joint scheduling problem of electric vehicles and power grid into a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include: distribution network constraints and vehicle-grid coordination constraints.

[0056] Step S102: Based on the bipartite graph structure, perform scheduling prediction and obtain the scheduling prediction probability;

[0057] Step S103: Construct a confidence region based on the scheduling prediction probability;

[0058] Step S104: Solve the sub-problems determined by the confidence region within the confidence region with the objectives of minimizing power generation cost and minimizing electric vehicle energy interaction cost, and obtain the scheduling parameters within the scheduling cycle;

[0059] The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

[0060] It should be noted that, in this embodiment of the application, the joint scheduling problem of electric vehicles and the power grid is encoded into a bipartite graph structure. Based on the bipartite graph structure, scheduling prediction is performed to obtain the scheduling prediction probability. Based on the scheduling prediction probability, a confidence region is constructed. Within the confidence region, the sub-problems determined by the confidence region are solved with the objectives of minimizing the generation cost and minimizing the energy interaction cost of electric vehicles, and the scheduling parameters within the scheduling period are obtained. In this way, by solving the scheduling parameters in the confidence region, the computational complexity of joint scheduling of electric vehicles and the power grid can be reduced, and the computational overhead can be reduced.

[0061] Optionally, the scheduling parameters include at least one of the following: generator power, generator voltage, electric vehicle charging capacity, electric vehicle discharging capacity, electric vehicle remaining capacity, electric vehicle charging power, electric vehicle discharging power, and electric vehicle charging / discharging mode.

[0062] Optionally, the distribution network constraints include at least one of the following:

[0063] Power flow constraints, power balance constraints, voltage amplitude constraints, current amplitude constraints, and equipment capacity and operation limitations constraints.

[0064] Optionally, the vehicle-to-everything (V2X) coordination constraints include at least one of the following:

[0065] Charge / discharge state constraints, charge / discharge power constraints, and battery capacity constraints.

[0066] It should be noted that this application proposes a hybrid learning and confidence region optimization (LeTRO) method for joint scheduling of electric vehicles and the power grid, aiming to solve the problems of high computational complexity and low feasibility in the joint modeling of optimal power flow (OPF) and vehicle-to-grid (V2G) scheduling. This method first uses a graph neural network (GCNN) to model the physical topology of the power grid and combines it with a long short-term memory network (LSTM) to extract time-series features of electric vehicle charging and discharging behavior, achieving accurate prediction of scheduling patterns. Then, the prediction results are transformed into a prior distribution of integer control variables, and a confidence region is constructed based on this to dynamically limit the search space of the MISOCP model. Finally, an iterative shrinking confidence region optimization strategy is used to quickly solve the problem within the feasible region, thereby significantly improving computational speed and the quality of scheduling solutions while ensuring physical feasibility such as power flow constraints, voltage safety, and SOC limitations.

[0067] The embodiments of this application are structurally divided into two phases: the learning phase and the optimization phase, and their overall framework is as follows: Figure 2 As shown, the learning phase extracts spatial and temporal features to predict the initial scheduling solution; the optimization phase solves for feasible solutions within the confidence region.

[0068] in, Figure 2 The meanings of the relevant parameters in the table are as follows: X is the set of continuous control variables (e.g., generator power). ,Voltage EV charging power EV discharge power (etc.). Z is the set of binary integer control variables (i.e., the charging mode of the electric vehicle). Discharge mode t represents a time period within the optimization cycle. T represents the total set of time periods within the optimization cycle. g represents a generator. G represents the set of nodes connected to distributed generators. j represents a node in the power grid. This is a function representing the power generation cost of the generator. Let be the active power generated by generator g located at node j at time t. Let e ​​represent the amount of electricity at time t. Let e ​​refer to a single electric vehicle. Let V be the set of all EVs connected to the charging station. The charging power of electric vehicle e at time t. Let be the discharge power of electric vehicle e at time t.

[0069] It should be noted that the joint scheduling problem is first modeled as MISOCP before constructing the bipartite graph. In existing technologies, OPF and V2G scheduling are usually modeled separately or global convexity processing is used, making it difficult to simultaneously solve the integer charge-discharge control and power flow second-order cone constraints. The embodiments of this application introduce distribution network constraints and vehicle-network coordination constraints simultaneously under the same MISOCP framework, realizing coupled modeling of physical constraints and scheduling decisions, and providing a unified mathematical foundation for subsequent optimization.

[0070] It should be noted that by establishing a joint optimization scheduling model for electric vehicles and the power grid, with the goals of reducing the total system operating cost, mitigating peak load impacts, and ensuring stable grid operation, the optimization objectives and physical constraints are modeled within a unified mathematical framework. Based on MISOCP, this model uses integer variables to represent the charging and discharging states of electric vehicles in each scheduling cycle, and continuous variables to describe physical quantities such as power flow, voltage amplitude, and remaining charge (SOC) level.

[0071] In model construction, constraints directly determine the shape and extent of the feasible solution space, including:

[0072] Power flow constraints: Second-order cone constraints are established based on the power flow equations of the distribution network to characterize the nonlinear relationship between node voltage and branch power. Through coneification and linearization, the computation is made feasible and convergent during the optimization process.

[0073] Electric vehicle SOC constraint (i.e., battery capacity constraint): Based on the vehicle's initial SOC, charging and discharging power and efficiency parameters, the range of SOC variation is constrained to avoid overcharging or over-discharging.

[0074] Charge / discharge state constraints: Use 0-1 integer variables to represent the three states of charging, discharging, and standby, and add mutual exclusion conditions to ensure that it can only be in one state at the same time.

[0075] Equipment capacity and operating limits: Limit the maximum power capacity and safe operating range of equipment such as transformers and feeders to prevent overload operation.

[0076] Power balance constraint: Ensure the balance between power injection at each node and load, satisfying the basic physical laws of distribution network operation.

[0077] The relationship between these constraints and the MISOCP model is that they collectively constitute the feasible region of the optimization problem, determining the boundary of the search space; while the optimization objective function seeks the solution with the minimum system operating cost within this feasible region. By uniformly embedding the above constraints into the MISOCP framework, the economic efficiency and physical feasibility of the solution can be guaranteed simultaneously, providing a stable mathematical foundation for subsequent prediction and confidence region optimization steps.

[0078] During the initialization phase of the scheduling system, the optimization objective is to minimize the generation cost and the energy interaction cost of electric vehicles throughout the entire scheduling cycle. The model achieves refined collaborative control of the EV group by combining electricity prices, generation unit output characteristics, and EV user behavior across different time periods. The objective function of the OPF and V2G collaborative scheduling problem is modeled as follows: .

[0079] Among them, the cost of generators The cost of charging and discharging the electric vehicle is the sum of the active power generated. It is the difference between the hourly electricity price and the charge / discharge rate.

[0080] The constraints in the optimization problem include distribution network constraints (grid power flow constraints, power balance constraints, voltage amplitude constraints, current amplitude constraints, equipment capacity and operation limit constraints) and vehicle-to-grid (V2G) constraints (charge and discharge power constraints, battery capacity constraints, charge and discharge state constraints).

[0081] The power balance equation is used to describe the active and reactive power balance state of each node in any time slot. It takes into account the power flow of upstream / downstream branches, the charging and discharging power of electric vehicles, and local load demand to ensure the conservation of node power.

[0082] The active power balance equation is:

[0083] .

[0084] in, Let be the active power load of node j at time t. , These are the active power flows from node j to k and from node i to j, respectively. Let be the resistance on branch (i,j). It is the square of the current in branch (i,j). Let be the active power generated by the generator at node j at time t. Let be the set of EVs that access charging station j at time t. η is the charging / discharging efficiency factor.

[0085] The reactive power balance equation is:

[0086] .

[0087] in, Let be the reactive power load of node j at time t. , These are the reactive power flows from node j to k and from node i to j, respectively. Let be the reactance on branch (i,j). Let be the reactive power generated by the generator at node j at time t.

[0088] Voltage amplitude constraints at nodes are one of the core constraints for ensuring voltage quality and safe system operation. The voltage at each distribution network node must be maintained within the allowable physical range to avoid equipment failure or power instability caused by voltage deviations from standard values.

[0089] .

[0090] in, Let the square of the voltage at node j at time t be ; Let be the square of the voltage at node i at time t; L represents all branches.

[0091] Building upon this, a second-order cone constraint is introduced to achieve convex relaxation modeling of the relationship between branch power and voltage. This constraint ensures that node voltage variations remain within physically permissible limits while also taking into account the nonlinear coupling relationships in the power flow equations.

[0092] .

[0093] Furthermore, the scheduling control variables for electric vehicles are constrained by SOC state, upper and lower limits of charging and discharging power, and mutual exclusion constraints. These constraints ensure that each EV charges and discharges as needed within the permitted time window, while simultaneously meeting the requirements for battery capacity and stable grid operation.

[0094] .

[0095] .

[0096] .

[0097] in, This represents the charge / discharge state at the previous time. The current charging / discharging state. The length of a single time period. , These represent the lower and upper limits of the battery's State of Charge (SOC). =1 indicates charging mode. =1 indicates the discharge mode. = =0 indicates that the space is empty.

[0098] To ensure the safety of power grid operation and the rationality of equipment operating conditions, it is also necessary to introduce physical boundary constraints on variables such as node voltage, current, power generation output, and electric vehicle charging and discharging power.

[0099] .

[0100] .

[0101] .

[0102] .

[0103] .

[0104] .

[0105] in, Let be the lower and upper limits of the squared voltage at node i. This is the upper limit of the square of the current in branch (i,j). These represent the lower and upper limits of the generator's active power. , These are the lower and upper limits of the active power during charging and discharging.

[0106] Optionally, in one implementation, the specific implementation of encoding the joint scheduling problem of electric vehicles and the power grid into a bipartite graph structure includes:

[0107] Step 11: Construct a set of variable nodes and a set of constraint nodes based on the power grid physical nodes and electric vehicle aggregation units;

[0108] Step 12: Map the continuous variables in the optimal power flow calculation and the integer variables in the electric vehicle scheduling to the node attributes of the bipartite graph, and represent the target constraint as an edge connection across the set to obtain the bipartite graph structure with nodes and edges;

[0109] The integer variable is used to represent the charging and discharging status of the electric vehicle in each scheduling cycle, and the continuous variable is used to describe at least one of the following: power flow, voltage amplitude, and remaining power level.

[0110] It should be noted that, to efficiently characterize the multi-layered coupling relationship between variables and constraints in the joint optimization scheduling problem, this application proposes an innovative bipartite graph structure modeling mechanism. This structure uses power grid physical nodes and electric vehicle aggregation units as two types of vertex sets, uniformly mapping continuous variables in optimal power flow calculation and integer variables in electric vehicle scheduling to the node attributes of the bipartite graph. Furthermore, it explicitly represents physical relationships such as distribution network constraints and / or vehicle-grid coordination constraints as cross-set edge connections. Unlike traditional vectorized input, this structured modeling not only preserves the topological information and constraint dependencies of the physical system but also enables the subsequent graph neural network to simultaneously capture spatial adjacency relationships and constraint propagation characteristics during propagation and aggregation, thus embedding the prior structure of the optimization problem at the input layer. Its overall structure is as follows: Figure 3 As shown.

[0111] By optimizing the joint dispatch of electric vehicles and the power grid (i.e., the primal optimization problem) Mapped to a weighted bipartite graph ,in:

[0112] Variable node set Each node This corresponds to a decision variable (such as power, voltage, SOC, charge / discharge rate, etc.). Node characteristics include variable type (continuous / integer), upper and lower limits, objective function coefficients, relaxation solution values, and the activity of the variable in the constraints.

[0113] Constraint node set C Each node This represents the physical or mathematical constraints in an optimization model, such as power flow balancing, cone constraints, voltage calculation, and SOC updates. Constraint node characteristics include constraint type, mathematical form (linear / conical), whether it includes EV behavior, dual solution values, and time information.

[0114] Edge set E Indicative variables Under constraints The participation relationship in the constraint. The weight of the edge is the coefficient of the variable in the constraint (linear or second-order term). Edge features also include standardized coefficients, gradient information between the variable and the constraint, cone constraint labels, etc.

[0115] Through the above mapping, the embodiments of this application can standardize each optimized scheduling instance into a well-structured and attribute-rich graph, providing a structural foundation for neural network learning.

[0116] It should be noted that after completing the bipartite graph modeling, LeTRO uses a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph to extract spatial structure information between variables and constraints. Optionally, in one implementation, the specific method of performing scheduling prediction based on the bipartite graph structure to obtain the scheduling prediction probability includes steps 21 and 22:

[0117] Step 21: Use a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph structure to extract spatial structure information between variables and constraints;

[0118] Optionally, in one implementation, the specific implementation of using a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph structure and extract spatial structure information between variables and constraints includes steps 211-213:

[0119] Step 211: Aggregate the information of the first constraint node in the bipartite graph structure with the information of its adjacent variable nodes and edges, and update the embedded representation of the first constraint node.

[0120] Optionally, this step can be understood as the first stage of extracting spatial structure information (constraint direction propagation), that is, each constraint node... Aggregate its adjacent variable nodes and edge Information to update its own embedded representation (Node representation of variable nodes):

[0121] = .

[0122] Step 212: Aggregate the information of the first variable node in the bipartite graph structure with the information of its adjacent constraint nodes and edges, and update the embedded representation of the first variable node.

[0123] Optionally, this step can be understood as the second stage of extracting spatial structure information (variable directional propagation), that is, each variable node... Aggregate its associated constraint nodes and edge Information to update variable embedding representation (Node representation of constraint nodes):

[0124] .

[0125] in, , , , All are multilayer perceptrons (MLPs) using the ReLU activation function.

[0126] Step 213: Based on the embedded representation of the first constraint node and the embedded representation of the first variable node, determine the spatial structure information between variables and constraints;

[0127] Wherein, the first constraint node is any node in the set of constraint nodes corresponding to the bipartite graph structure, and the first variable node is any node in the set of variable nodes corresponding to the bipartite graph structure.

[0128] It should be noted that, through multiple rounds of propagation, GCNN can capture global topological characteristics, such as strong correlations between nodes, redundant dependencies, and structural tension, providing a spatial perspective for subsequent EV scheduling prediction.

[0129] Step 22: Based on the spatial structure information, use LSTM to process the dynamic scheduling behavior in the time dimension and obtain the scheduling prediction probability for each time step.

[0130] Optionally, the specific implementation of processing the dynamic scheduling behavior in the time dimension using a Long Short-Term Memory (LSTM) network based on spatial structure information to obtain the scheduling prediction probability for each time step includes:

[0131] Step 221: Input the control variables of each electric vehicle in each time period and the spatial structure information of the charging station node corresponding to it into the LSTM model in ascending order of time to capture the cross-time state evolution and constraint transmission of electric vehicles and obtain the scheduling prediction probability of each time step.

[0132] It should be noted that EV scheduling tasks have significant time dependencies, such as SOC state evolution, user connection duration, and electricity price curves. Therefore, this application's embodiments introduce LSTM (Long Short-Term Memory) networks to handle the dynamic scheduling behavior in the time dimension. The core process is as follows:

[0133] Control variables for each EV in each time period (e.g.) , ) and its charging station node The embedded features are used as time series input:

[0134] .

[0135] in, The input vector of a specific electric vehicle e is input to the LSTM at time t; The hidden graph features corresponding to the charging power variable of electric vehicle e at time t; Let be the discharge power variable of electric vehicle e at time t; The hidden graph features are those corresponding to the charging station node where electric vehicle e is located.

[0136] Input sequence The data is fed into the LSTM model in ascending order of time to capture the cross-time state evolution and constraint propagation of the EV.

[0137] The LSTM outputs the scheduling prediction probability vector for each time step:

[0138] .

[0139] in, , , and represent the probabilities that the EV is in a charging, discharging, or idle state at time t, respectively, and are used to construct the initial scheduling solution. Let represent the hidden state of the LSTM at time t. is the weight matrix of the output layer. b is the bias vector of the output layer.

[0140] It should be noted that after completing the spatial topological features and time series modeling, LeTRO enters the prediction phase. The goal is to use a trained neural network model to perform soft prediction of the electric vehicle charging and discharging patterns (binary optimization variables), and thereby construct an initial solution region (confidence region) for optimization. Optionally, in one implementation, the specific implementation of constructing the confidence region based on the scheduling prediction probability includes:

[0141] Step 31: Obtain the k positions with the highest prediction probability among all scheduling prediction probabilities at each time step, where k is an integer greater than or equal to 1;

[0142] Step 32: Construct an initial solution based on the patterns corresponding to the k positions. The pattern corresponding to each position is one of the following: charging mode (also known as charging state), discharging mode (also known as discharging state), and idle mode (also known as idle state).

[0143] Step 33: Determine the confidence radius based on the target norm according to the initial solution, and obtain the confidence region;

[0144] For example, the target norm is Norm.

[0145] It should be noted that for each electric vehicle e in any time period t, its possible scheduling patterns include three categories:

[0146] =1 indicates that the EV is in charging mode;

[0147] =1 indicates that the EV is in discharge mode;

[0148] When both are 0, it indicates an idle state.

[0149] The final output of the neural network is a set of three-class probability vectors:

[0150] = .

[0151] in, To represent the probability that EV is in one of the three states at time t, the hidden state output by LSTM is obtained by passing through a fully connected layer and a Softmax layer. Let represent the output feature vector of LSTM at time t.

[0152] To accelerate the subsequent optimization process, LeTRO uses the predicted probability results to construct a set of initial scheduling decisions. And define the optimization region around it:

[0153] 1. Sort selection: Select the k positions (e, t) with the highest predicted probabilities among all predictions, considering these predictions to be relatively reliable;

[0154] 2. Constructing a partially fixed solution: Construct an initial solution based on the mode corresponding to the highest probability (e.g., charging / discharging / idling). ;

[0155] 3. Generate confidence region: around the initial point... The norm (corresponding to the target norm above) defines a confidence radius Δ, and the constraints are constructed as follows: .

[0156] This constraint limits the search range within the integer variable space, meaning the optimizer can only search in the solution domain close to the initial solution, thereby improving feasibility and solution efficiency.

[0157] After completing the initial solution prediction and confidence region construction, LeTRO enters the final optimization stage. In this stage, by restricting the main problem (MISOCP) to the space of credible solutions, a simpler and smaller subproblem is generated. The subproblem inherits all the physical constraints and nonlinear structures of the original MISOCP problem, only shrinking the feasible region for integer variables. This forms a collaborative solution structure of "prediction guidance + optimization verification".

[0158] Subproblems are handled by calling a mixed-integer optimization solver (such as SCIP). Due to the confidence region constraint, a large number of irrelevant solutions are eliminated, which significantly reduces the solver branch traversal space, thereby achieving fast convergence while maintaining feasibility.

[0159] Actual testing shows that, compared to solving the entire space, this application reduces the solution time by an average of 56.38% in the Case 141 test system, and the optimality gap is controlled within 0.48%.

[0160] Based on all the above steps, the pseudocode of the LeTRO algorithm flow in this embodiment is as follows:

[0161] Parameters: Confidence region radius Δ and an integer k;

[0162] Input: Original OPF-V2G problem instance P;

[0163] Output: Solution X, Y, which are the scheduling parameters obtained within the scheduling cycle with the objectives of minimizing generation cost and minimizing energy interaction cost of electric vehicles;

[0164] Step 1. Encode P as a bipartite graph G;

[0165] Step 2. Use a neural network to predict. ;

[0166] Step 3. For Sort the k largest values ​​in the array to obtain and ;

[0167] Step 4. According to Restore integer control variables ;

[0168] Step 5. Construct new confidence region constraints ;

[0169] Step 6. Construct new subproblems ;

[0170] Step 7. X, Y Call the MIP solver to solve the subproblem ;

[0171] Step 8. Return X, Y.

[0172] Where P is an instance of the original OPF-V2G problem, namely the joint scheduling problem of electric vehicles and the power grid. Let m be the probability of the predicted three states, and m be the probability of the EV being in a charging, discharging, or idle state as output by the neural network. This is the set of indices corresponding to the k maximum probability values. According to Constructed partial fixed solutions; Let be the initial solution, and y be the initial point of the confidence region. Let be the confidence radius. This is a subproblem, namely, a problem consisting of all objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and the power grid. X represents the scheduling parameters to be solved within the scheduling period with the objective of minimizing generation costs, and Y represents the scheduling parameters to be solved within the scheduling period with the objective of minimizing the energy interaction costs of electric vehicles. This represents the probability prediction results of the neural network for the EV pattern.

[0173] It should be noted that this application proposes a joint scheduling method for electric vehicles and the power grid that integrates spatiotemporal feature modeling using graph neural networks and confidence region optimization. By unifying the optimal power flow (OPF) and vehicle-to-grid (V2G) scheduling into a single model called MISOCP, and mapping continuous variables, integer variables, and their second-order cone / inequality constraints in the optimization problem to an attributed bipartite graph structure, the physical topology features of the power grid are extracted using a graph convolutional neural network (GCNN), and the time dependence of electric vehicle charging and discharging behavior is captured using a long short-term memory network (LSTM), achieving spatiotemporal joint prediction of the scheduling state. Unlike existing technologies that only use tiling vectors or local statistical features as input, this application retains the physical adjacency relationships and constraint propagation structures between variables, allowing the prior structure of the problem to be embedded in the prediction stage. This solves the problems of low prediction accuracy and unstable optimization solutions caused by the lack of structural information in traditional methods, significantly improving the accuracy and physical feasibility of state prediction in large-scale scheduling scenarios.

[0174] This application proposes an application of a trust region optimization method based on prediction confidence in the joint dispatching of electric vehicles and the power grid: it utilizes the integer variable pattern probabilities predicted by a neural network to construct... Using the confidence region subproblem under norm constraints as the initial optimization point, the solution range of MISOCP is limited to the high-probability solution space. An iterative shrinking confidence region strategy improves the convergence speed and the proportion of feasible solutions. Unlike existing technologies that directly solve within the global feasible region, this application dynamically reduces the search space based on prediction results, significantly reducing the computational overhead caused by integer combination explosion. It solves the problems of long optimization computation time and low feasible solution proportion when large-scale EV access is implemented. While ensuring power flow constraints, voltage stability, and SOC safety, it achieves real-time scheduling solution and high-quality output.

[0175] In summary, compared with existing OPF-V2G joint scheduling schemes based on traditional mixed-integer optimization models, the LeTRO (Learning and Trust Region Optimization) method proposed in this application has significant technical advantages in several aspects, including computational efficiency, scheduling accuracy, and model feasibility. Specifically, these advantages are manifested in the following points:

[0176] 1. Computational efficiency is significantly improved.

[0177] Traditional mixed-integer solvers (such as SCIP) are slow in solving large-scale EV access scenarios, making it difficult to meet the real-time requirements of actual scheduling. This application's embodiment uses a deep learning model to predict the optimal distribution of some binary integer variables, and then constructs a confidence region to limit the optimization search range, significantly accelerating the overall solution process while maintaining optimization accuracy. Experimental results show that LeTRO reduces the solution time by more than 56% in typical distribution network test cases.

[0178] 2. Optimized for better quality and stronger stability.

[0179] This application utilizes neural networks to accurately predict integer variables and combines this with optimization algorithms to further solve within a reliable region, achieving rapid acquisition of near-optimal solutions. Test results show that the average optimality gap of the LeTRO output solution is as low as 0.48%, significantly better than traditional heuristic or fixed strategies (3.33%).

[0180] 3. Strong ability to ensure the feasibility of scheduling results.

[0181] Traditional prediction strategies (such as fixed patterns or soft decision schemes) often lead to feasibility collapse when dealing with constrained scenarios, with some strategies having a feasibility rate of only 6%. The embodiments of this application introduce a confidence region mechanism to limit the optimization range based on the prediction, ensuring that the solution is feasible within the original constraint space. In practical applications, 100% of physically feasible scheduling solutions are obtained.

[0182] 4. The model structure is highly adaptable and has good versatility.

[0183] The bipartite graph modeling and GCNN-LSTM fusion prediction strategy of this application has good scalability and can be flexibly applied to distribution network structures of different scales and different types of EV load distribution scenarios. It eliminates the need to reconstruct the mathematical model for different instances or manually adjust parameters, greatly enhancing the model's generalization ability and engineering practicality.

[0184] 5. Learning and optimization are integrated to avoid the blind traversal of the solution space by traditional methods.

[0185] Compared with existing B&B-based MIP solution strategies, the LeTRO implementation in this application guides the optimization of the initial point by learning prior information, which significantly reduces the search space and the number of iterations, making the scheduling system more forward-looking and intelligent.

[0186] The embodiments of this application address the need for efficient optimization of joint dispatching of electric vehicles and power grids, and have significant application value.

[0187] Firstly, in terms of application areas, the method of this application embodiment can be widely applied to various scenarios such as intelligent dispatching of electric vehicle fleets in smart grids, flexible resource management of distribution networks, V2G commercialization platforms, and optimization of power market ancillary services. It is particularly suitable for joint optimal power flow and load response dispatching tasks under large-scale electric vehicle access. This method can effectively improve grid operating efficiency, alleviate peak load pressure, promote the consumption of new energy, and bring economic benefits to electric vehicle users. It is in line with the national "dual-carbon" strategy and the development trend of "source-load-storage-vehicle" collaborative optimization, and has long-term industrialization development potential. Secondly, in terms of implementation feasibility and cost, as a pure algorithmic software solution, the embodiments of this application can be directly integrated into existing energy management systems (EMS), dispatch automation systems (DMS), or V2G operation platforms without the need for additional hardware modifications, resulting in high system compatibility. Model training and prediction can be deployed on edge servers or dispatch centers, with low computational resource requirements and no increase in equipment burden. By reducing the optimization scale through prediction, the cost of dispatching computation and communication can be significantly reduced, which is conducive to rapid promotion in various power grid and vehicle-grid collaborative platforms. Furthermore, from a market perspective, this technology aligns with the trend of increasing the proportion of new energy sources and electric vehicles gradually becoming a flexible resource for the power grid, possessing a broad market potential within the industrial landscape that integrates smart energy and intelligent transportation.

[0188] like Figure 4 As shown in the figure, an electric vehicle and power grid joint dispatching device according to an embodiment of this application includes:

[0189] Encoding module 401 is used to encode the joint scheduling problem of electric vehicles and power grid into a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include distribution network constraints and vehicle-grid coordination constraints.

[0190] The first acquisition module 402 is used to perform scheduling prediction based on the bipartite graph structure and acquire the scheduling prediction probability.

[0191] Construction module 403 is used to construct a confidence region based on the scheduling prediction probability;

[0192] The second acquisition module 404 is used to solve the sub-problems determined by the confidence domain within the confidence domain with the objectives of minimizing power generation cost and minimizing electric vehicle energy interaction cost, and to obtain the scheduling parameters within the scheduling period;

[0193] The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

[0194] Optionally, the scheduling parameters include at least one of the following: generator power, generator voltage, electric vehicle charging capacity, electric vehicle discharging capacity, electric vehicle remaining capacity, electric vehicle charging power, electric vehicle discharging power, and electric vehicle charging / discharging mode.

[0195] Optionally, the encoding module 401 is used for:

[0196] Construct a set of variable nodes and a set of constraint nodes based on power grid physical nodes and electric vehicle aggregation units;

[0197] The continuous variables in optimal power flow calculation and the integer variables in electric vehicle scheduling are mapped to the node attributes of a bipartite graph, and the target constraints are represented as edge connections across sets to obtain a bipartite graph structure with nodes and edges.

[0198] The integer variable is used to represent the charging and discharging status of the electric vehicle in each scheduling cycle, and the continuous variable is used to describe at least one of the following: power flow, voltage amplitude, and remaining power level.

[0199] Optionally, the first acquisition module 402 is configured to:

[0200] A graph neural network (GCNN) is used to perform embedding representation learning on the bipartite graph structure to extract spatial structure information between variables and constraints.

[0201] Based on spatial structure information, the Long Short-Term Memory (LSTM) network is used to process the dynamic scheduling behavior in the time dimension and obtain the scheduling prediction probability for each time step.

[0202] Optionally, the method of using a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph structure and extracting spatial structure information between variables and constraints includes:

[0203] In the bipartite graph structure, the information of the first constraint node is aggregated with the information of its adjacent variable nodes and edges, and the embedded representation of the first constraint node is updated.

[0204] In the bipartite graph structure, the information of the first variable node is aggregated with the information of its adjacent constraint nodes and edges, and the embedded representation of the first variable node is updated.

[0205] Based on the embedded representation of the first constraint node and the embedded representation of the first variable node, the spatial structure information between variables and constraints is determined;

[0206] Wherein, the first constraint node is any node in the set of constraint nodes corresponding to the bipartite graph structure, and the first variable node is any node in the set of variable nodes corresponding to the bipartite graph structure.

[0207] Optionally, the method for obtaining the scheduling prediction probability for each time step by using a Long Short-Term Memory (LSTM) network to process the dynamic scheduling behavior in the time dimension based on spatial structure information includes:

[0208] The control variables of each electric vehicle in each time period and the spatial structure information of the charging station node corresponding to it are input into the LSTM model in ascending order of time to capture the cross-time state evolution and constraint transmission of electric vehicles and obtain the scheduling prediction probability at each time step.

[0209] Optionally, the construction module 403 is configured to:

[0210] Get the k positions with the highest predicted probability among all the scheduling prediction probabilities at each time step, where k is an integer greater than or equal to 1;

[0211] An initial solution is constructed based on the patterns corresponding to the k positions, where each position corresponds to one of the charging mode, discharging mode, or idle mode.

[0212] The confidence radius is determined based on the target norm according to the initial solution, and the confidence region is obtained.

[0213] Optionally, the distribution network constraints include at least one of the following:

[0214] Power flow constraints, power balance constraints, voltage amplitude constraints, current amplitude constraints, and equipment capacity and operation limitations constraints;

[0215] and / or

[0216] The vehicle-to-everything (V2X) coordination constraints include at least one of the following:

[0217] Charge / discharge state constraints, charge / discharge power constraints, and battery capacity constraints.

[0218] It should be noted that this device embodiment corresponds one-to-one with the above method embodiments. All implementation methods in the above method embodiments are applicable to this device embodiment and can achieve the same technical effect.

[0219] Another embodiment of this application provides a joint dispatching device for electric vehicles and the power grid, such as... Figure 5As shown, it includes a transceiver 510, a processor 500, a memory 520, and a program or instructions stored in the memory 520 and executable on the processor 500; when the processor 500 executes the program or instructions, it implements the above-mentioned method for joint scheduling of electric vehicles and power grids.

[0220] The transceiver 510 is used to receive and send data under the control of the processor 500.

[0221] Among them, Figure 5 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 500) and memory (memory 520). The bus architecture can also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 510 can be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over a transmission medium.

[0222] The processor 500 is responsible for managing the bus architecture and general processing, while the memory 520 can store the data used by the processor 500 when performing operations.

[0223] This application provides a readable storage medium storing a program or instructions. When executed by a processor, the program or instructions implement the steps in the electric vehicle and power grid joint scheduling method described above, achieving the same technical effect. To avoid repetition, further details are omitted here. The computer-readable storage medium may include read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0224] It should be further noted that the terminals described in this specification include, but are not limited to, smartphones, tablets, etc., and many of the functional components described are referred to as modules in order to emphasize the independence of their implementation.

[0225] In this embodiment, the module can be implemented in software so that it can be executed by various types of processors. For example, an identified executable code module may include one or more physical or logical blocks of computer instructions, which may be constructed as objects, procedures, or functions. Nevertheless, the executable code of the identified module does not need to be physically located together, but may include different instructions stored in different bits, which, when logically combined, constitute the module and achieve the module's intended purpose.

[0226] In practice, an executable code module can be a single instruction or many instructions, and can even be distributed across multiple different code segments, different programs, and across multiple memory devices. Similarly, operational data can be identified within the module and can be implemented in any suitable form and organized within any suitable data structure. This operational data can be collected as a single dataset or distributed across different locations (including different storage devices), and can exist, at least in part, solely as electronic signals within the system or network.

[0227] When a module can be implemented using software, considering the current level of hardware technology, modules that can be implemented in software can be implemented using hardware circuits by those skilled in the art to achieve the corresponding functions, without considering cost. These hardware circuits include conventional very-large-scale integrated circuits (VLSI) or gate arrays, as well as existing semiconductors such as logic chips and transistors, or other discrete components. Modules can also be implemented using programmable hardware devices, such as field-programmable gate arrays, programmable array logic, and programmable logic devices.

[0228] The exemplary embodiments described above are with reference to the accompanying drawings. Many different forms and embodiments are feasible without departing from the spirit and teachings of this application. Therefore, this application should not be construed as limiting the exemplary embodiments set forth herein. Rather, these exemplary embodiments are provided to make this application complete and convey the scope of this application to those skilled in the art. In these drawings, component dimensions and relative dimensions may be exaggerated for clarity. The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. As used herein, unless clearly indicated otherwise, the singular forms “a,” “an,” and “the” are intended to include all such forms. It will be further understood that the terms “comprising” and / or “including”, when used in this specification, indicate the presence of the stated features, integers, steps, operations, components, and / or elements, but do not exclude the presence or addition of one or more other features, integers, steps, operations, components, and / or groups thereof. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of the range and any subranges in between.

[0229] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for joint dispatching of electric vehicles and power grid, characterized in that, include: The joint scheduling problem of electric vehicles and power grid is encoded as a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include distribution network constraints and vehicle-grid coordination constraints. Based on the bipartite graph structure, a scheduling prediction is performed to obtain the scheduling prediction probability; Construct a confidence region based on the scheduling prediction probability; Within the confidence region, the sub-problems determined by the confidence region are solved with the objectives of minimizing power generation costs and minimizing electric vehicle energy interaction costs to obtain scheduling parameters within the scheduling cycle; The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

2. The method according to claim 1, characterized in that, The scheduling parameters include at least one of the following: generator power, generator voltage, electric vehicle charging power, electric vehicle discharging power, electric vehicle remaining power, electric vehicle charging power, electric vehicle discharging power, and electric vehicle charging / discharging mode.

3. The method according to claim 1, characterized in that, The method of encoding the joint scheduling problem of electric vehicles and the power grid into a bipartite graph structure includes: Construct a set of variable nodes and a set of constraint nodes based on power grid physical nodes and electric vehicle aggregation units; The continuous variables in optimal power flow calculation and the integer variables in electric vehicle scheduling are mapped to the node attributes of a bipartite graph, and the target constraints are represented as edge connections across sets to obtain a bipartite graph structure with nodes and edges. The integer variable is used to represent the charging and discharging status of the electric vehicle in each scheduling cycle, and the continuous variable is used to describe at least one of the following: power flow, voltage amplitude, and remaining power level.

4. The method according to claim 1, characterized in that, The step of performing scheduling prediction and obtaining scheduling prediction probabilities based on the bipartite graph structure includes: A graph neural network (GCNN) is used to perform embedding representation learning on the bipartite graph structure to extract spatial structure information between variables and constraints. Based on spatial structure information, the Long Short-Term Memory (LSTM) network is used to process the dynamic scheduling behavior in the time dimension and obtain the scheduling prediction probability for each time step.

5. The method according to claim 4, characterized in that, The step of using a graph neural network (GCNN) to perform embedding representation learning on the bipartite graph structure and extracting spatial structure information between variables and constraints includes: In the bipartite graph structure, the information of the first constraint node is aggregated with the information of its adjacent variable nodes and edges, and the embedded representation of the first constraint node is updated. In the bipartite graph structure, the information of the first variable node is aggregated with the information of its adjacent constraint nodes and edges, and the embedded representation of the first variable node is updated. Based on the embedded representation of the first constraint node and the embedded representation of the first variable node, the spatial structure information between variables and constraints is determined; Wherein, the first constraint node is any node in the set of constraint nodes corresponding to the bipartite graph structure, and the first variable node is any node in the set of variable nodes corresponding to the bipartite graph structure.

6. The method according to claim 4, characterized in that, The step of processing the dynamic scheduling behavior in the time dimension using a Long Short-Term Memory (LSTM) network based on spatial structure information to obtain the scheduling prediction probability for each time step includes: The control variables of each electric vehicle in each time period and the spatial structure information of the charging station node corresponding to it are input into the LSTM model in ascending order of time to capture the cross-time state evolution and constraint transmission of electric vehicles and obtain the scheduling prediction probability at each time step.

7. The method according to claim 1, characterized in that, The step of constructing a confidence region based on the scheduling prediction probability includes: Get the k positions with the highest predicted probability among all the scheduling prediction probabilities at each time step, where k is an integer greater than or equal to 1; An initial solution is constructed based on the patterns corresponding to the k positions, where each position corresponds to one of the charging mode, discharging mode, or idle mode. The confidence radius is determined based on the target norm according to the initial solution, and the confidence region is obtained.

8. The method according to claim 1, characterized in that, The power distribution network constraints include at least one of the following: Power flow constraints, power balance constraints, voltage amplitude constraints, current amplitude constraints, and equipment capacity and operation limitations constraints; and / or The vehicle-to-everything (V2X) coordination constraints include at least one of the following: Charge / discharge state constraints, charge / discharge power constraints, and battery capacity constraints.

9. A device for joint dispatching of electric vehicles and power grid, characterized in that, include: The encoding module is used to encode the joint scheduling problem of electric vehicles and power grid into a bipartite graph structure. The joint scheduling problem of electric vehicles and power grid is modeled as a mixed integer second-order cone programming (MISOCP). The MISOCP is constrained by objective constraints, which include distribution network constraints and vehicle-grid coordination constraints. The first acquisition module is used to perform scheduling prediction based on the bipartite graph structure and acquire the scheduling prediction probability. The construction module is used to construct a confidence region based on the scheduling prediction probability; The second acquisition module is used to solve the sub-problems determined by the confidence domain within the confidence domain with the objectives of minimizing power generation cost and minimizing electric vehicle energy interaction cost, and to obtain the scheduling parameters within the scheduling period; The sub-problem consists of all the objective constraints and confidence region constraints of the joint scheduling problem of electric vehicles and power grid.

10. A device for joint dispatching of electric vehicles and power grid, comprising: A transceiver, a processor, a memory, and a program or instructions stored in the memory and executable on the processor; characterized in that, when the processor executes the program or instructions, it implements the electric vehicle and power grid joint scheduling method as described in any one of claims 1-8.

11. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps in the electric vehicle and power grid joint scheduling method as described in any one of claims 1-8.

Images