Virtual power plant multi-unit coordinated optimization scheduling method and system based on spot market

Abstract

The invention relates to the technical field of energy scheduling, and provides a virtual power plant multi-unit coordinated optimization scheduling method and system based on a spot market, and the method comprises the steps: collecting the operation data of each aggregation resource, combining the electrical characteristics and adjustment capability of each aggregation resource, extracting a dynamic characteristic index, and building a node identifier; constructing a behavior model based on the resource type and uncertainty; recognizing a cooperative relationship and constructing a hypergraph by using a dynamic identifier and a model, and gathering resources with similar characteristics into a decision unit; a double-layer model is established, the upper layer optimizes market income, and the lower layer optimizes response precision and adjustment cost; and solving the model by combining an alternating direction multiplier method with rolling optimization to obtain an optimal scheduling scheme of each resource. According to the method, mechanisms such as multi-source data driving, behavior modeling and double-layer optimization are fused, and the scheduling response capability and market adaptability of the virtual power plant are improved.

Description

Technical Field

[0001] This invention relates to the field of energy dispatch optimization technology, specifically to a method and system for coordinated optimization dispatching of virtual power plants based on the spot market. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the increasing penetration of new energy sources and the ongoing reform of the power market, the demand for flexible regulation capabilities in the power system is growing. Virtual Power Plants (VPPs), as an advanced technology integrating distributed energy resources (such as photovoltaics, wind power, controllable loads, and energy storage devices), can aggregate, coordinate, and uniformly dispatch distributed resources through information communication and control technologies. They are a crucial solution supporting the coordinated operation of power generation, grid, load, and energy storage. VPPs not only possess the potential to enhance grid regulation capabilities and promote the consumption of new energy sources, but they are also increasingly becoming important players in the electricity spot market.

[0004] Under market mechanisms, although virtual power plants participate in transactions as independent market entities, their scheduling and market behavior exhibit significant two-way interaction characteristics because their coordination with actual aggregated resources is usually based on an agency relationship. This two-way decision-making not only needs to consider the overall revenue and market constraints of the virtual power plant, but also the individual behavioral preferences, adjustment capabilities, and game relationships of each aggregated resource. Therefore, constructing an efficient, controllable scheduling method with behavioral feedback mechanisms is a key technological challenge for achieving optimal resource allocation.

[0005] Existing virtual power plant dispatching methods suffer from several technical shortcomings: Firstly, dispatching models typically fail to fully consider the dynamic response characteristics of user-side resources, making it difficult to accurately reflect the adjustment capabilities of different types of loads and energy storage devices under various scenarios. Secondly, current methods generally neglect the impact of external environmental factors (such as weather, temperature, and seasonal changes) on resource availability and dispatching effectiveness, resulting in a lack of adaptability in dispatching strategies. Furthermore, dispatching mechanism designs often suffer from unclear aggregation hierarchies and weak internal and external coupling, making it difficult to achieve efficient coordination of resources within the virtual power plant and between the power grid. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a multi-unit coordinated optimization scheduling method and system for virtual power plants based on the spot market. This method integrates mechanisms such as multi-source data-driven approaches, behavioral modeling, and two-layer optimization to enhance the scheduling response capability and market adaptability of virtual power plants.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: One or more embodiments provide a method for coordinated and optimized scheduling of multiple units in a virtual power plant based on the spot market, including the following steps: Acquire real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. Based on the operational characteristics and uncertainty sources of different aggregated resource types, a behavioral model of aggregated resources is constructed using the response function or power change distribution used for scheduling calculations. Based on the dynamic identifiers and behavior models of the aggregated resource nodes, the collaborative relationships and adjustment capabilities among resources are determined. Distributed resources with similar response characteristics and spatiotemporal coupling relationships are grouped into the same hyperedge as a decision-making unit to construct a hypergraph structure. A two-level decision-making model is constructed, with the upper level making clearing decisions with the goal of maximizing market returns, and the lower level allocating resources with the goal of minimizing response accuracy and adjustment costs; By acquiring real-time operational data of each aggregated resource and the resource demand in the spot market, and based on the constructed hypergraph structure, the alternating direction multiplier method is used to solve the two-level decision model to obtain the scheduling scheme of each aggregated resource.

[0008] One or more embodiments provide a virtual power plant multi-unit coordinated optimization scheduling system based on the spot market, including: The dynamic identifier calculation module is configured to acquire real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. The behavior modeling module is configured to construct a behavior model of aggregated resources based on the operational characteristics and uncertainty sources of different aggregated resource types, using response functions or power change distributions for scheduling calculations. The decision unit generation and partitioning module is configured to determine the collaborative relationship and adjustment capability between resources based on the dynamic identifier and behavior model of the obtained aggregated resource nodes, and to classify distributed resources with similar response characteristics and spatiotemporal coupling relationship into the same hyperedge as a decision unit to construct a hypergraph structure. The decision model building module is configured to build a two-layer decision model, with the upper layer aiming to maximize market returns and make clearing decisions, and the lower layer aiming to minimize response accuracy and adjustment costs in resource allocation. The solution module is configured to acquire real-time operating data of each aggregated resource and the resource demand in the spot market. Based on the constructed hypergraph structure, it uses the alternating direction multiplier method to perform rolling optimization on the two-level decision model to obtain the scheduling scheme of each aggregated resource.

[0009] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the processor executes the computer instructions, the computer instructions perform the steps in the above-described virtual power plant multi-unit coordinated optimization scheduling method based on the spot market.

[0010] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the steps in the above-described virtual power plant multi-unit coordinated optimization scheduling method based on the spot market.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: This embodiment, by introducing dynamic feature indicators, can accurately characterize the real-time adjustment capabilities of each aggregated resource, improving the model's adaptability to resource heterogeneity. Based on behavioral modeling and hypergraph construction, it achieves the identification and quantification of complex collaborative relationships between resources, thereby constructing scheduling units more rationally and enhancing the system's scheduling flexibility and coordination. Through a two-layer decision-making architecture, it effectively balances the contradiction between market response and internal resource allocation, improving overall returns while taking into account adjustment efficiency and stability. In addition, the introduction of the alternating direction multiplier method enables the model to have high computational efficiency and convergence when dealing with large-scale resource coordination problems, adapting to the needs of rolling optimization and rapid response in practical applications.

[0012] The advantages of the present invention, as well as its additional advantages, will be described in detail in the following specific embodiments. Attached Figure Description

[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0014] Figure 1 This is a flowchart of the multi-unit coordinated optimization scheduling method of Embodiment 1 of the present invention; Figure 2 This is a block diagram of the multi-unit coordinated optimization scheduling system of Embodiment 1 of the present invention. Detailed Implementation

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0017] It should be noted that the terminology used herein is for describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features within those embodiments can be combined with each other. The embodiments will now be described in detail with reference to the accompanying drawings.

[0018] Example 1 A virtual power plant is an aggregate of various types of resources, including distributed power sources, energy storage systems, and controllable loads. Each aggregated resource is regarded as a scheduling unit with independent response capabilities in the virtual power plant and is called an aggregated resource node. The aggregated resource nodes achieve data sharing and collaborative control through information and communication technologies.

[0019] In one or more of the technical solutions disclosed in the embodiments, such as Figures 1 to 2 As shown, the virtual power plant multi-unit coordinated optimization scheduling method based on the spot market includes the following steps: Step 1: Obtain real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. Step 2: Based on the operational characteristics and uncertainty sources of different aggregated resource types, construct a behavioral model of the aggregated resource for each aggregated resource node using the response function or power change distribution used for scheduling calculation; Step 3: Based on the dynamic identifiers and behavior models of the aggregated resource nodes, determine the collaborative relationships and adjustment capabilities among resources, group distributed resources with similar response characteristics and spatiotemporal coupling relationships into the same hyperedge, construct a hypergraph structure, and generate decision-making units for the virtual power plant; divide the decision-making units into first-level decision-making units for direct participation in scheduling and second-level decision-making units for scheduling through resource allocation; Step 3: Construct a two-level decision-making model. The upper level makes clearing decisions with the goal of maximizing market returns, while the lower level allocates resources with the goal of minimizing response accuracy and adjustment costs. Step 4: Obtain real-time operating data of each aggregated resource and the resource demand in the spot market. Based on the constructed hypergraph structure, use the alternating direction multiplier method to solve the two-level decision model and obtain the scheduling scheme of each aggregated resource. In this embodiment, firstly, real-time operational data of multiple aggregated resources (such as photovoltaic, wind power, controllable loads, and energy storage systems) distributed in a virtual power plant are collected. Combined with their electrical parameters and adjustability characteristics, feature indicators representing their dynamic behavior are extracted to form a dynamic identifier for each resource node. These identifiers reflect the resource's responsiveness and stability under different operating states. Furthermore, by modeling the operational history and response mechanisms of various resources, behavioral models with scheduling significance are constructed. These models can use response functions or probability distributions to describe the characteristics of resource power changes.

[0020] By utilizing dynamic identifiers and behavioral models, potential collaborative relationships between resources are identified, such as the complementarity between energy storage and load, and the joint regulation capabilities of photovoltaic and wind power. Based on these correlations, resources with similar response characteristics and temporal or spatial coupling are aggregated into a single decision-making unit, modeled using a hypergraph approach, where each hyperedge represents a set of resources with coordinated relationships.

[0021] Based on this structure, a two-layer optimal scheduling model is proposed. The upper-layer model aims to participate in the electricity spot market transaction and calculates the optimal clearing scheme for overall resources to obtain maximum benefits. The lower-layer model, considering the specific resource regulation characteristics and behavioral feedback mechanisms, further optimizes the allocation of regulation tasks to minimize regulation costs and improve response accuracy. Finally, by introducing the Alternating Direction Multiplier Method (ADMM) and combining it with a rolling optimization mechanism, dynamically changing data is processed in real time, achieving efficient solution of the two-layer model and output of scheduling results.

[0022] First, by introducing dynamic feature indicators, this embodiment can accurately characterize the real-time adjustment capabilities of each aggregated resource, improving the model's adaptability to resource heterogeneity. Second, based on behavioral modeling and hypergraph construction, it enables the identification and quantification of complex collaborative relationships between resources, thereby constructing scheduling units more rationally and enhancing the system's scheduling flexibility and coordination. Third, through a two-layer decision-making architecture, it effectively balances the contradiction between market response and internal resource allocation, improving overall returns while considering adjustment efficiency and stability. Furthermore, the introduction of the alternating direction multiplier method gives the model high computational efficiency and convergence when dealing with large-scale resource coordination problems, adapting to the needs of rolling optimization and rapid response in practical applications.

[0023] In step 1, a dynamic identifier for the aggregated resources of the virtual power plant is constructed to dynamically characterize the dynamic features of each aggregated resource node of the virtual power plant, including regulation capacity and response speed. Traditional virtual power plant decision-making models, when characterizing the behavior or utility function of aggregated resources, either use a one-size-fits-all approach of "controllable / uncontrollable" to label aggregated resources, or only model them from the perspective of production and sales or electrical characteristics. These approaches are difficult to dynamically respond to changes in boundary conditions, and cannot accurately characterize the real adjustment capabilities and behavior patterns of aggregated resources, which can easily lead to failures or delays in virtual power plant coordination and scheduling.

[0024] To avoid the above-mentioned defects, this embodiment will adopt a method that combines dynamic labeling based on scheduling capabilities with aggregated resource modeling to dynamically characterize aggregated resources.

[0025] In step 1, real-time operating data of each aggregated resource is acquired. Based on the electrical characteristics and regulation capabilities of the aggregated resources, dynamic characteristic indicators are calculated as dynamic identifiers for each aggregated resource node in the virtual power plant. The generation process of the dynamic identifier for each aggregated resource node includes the following steps: Step 11: Obtain real-time operating data for each aggregated resource, including data such as voltage, power, and frequency; Optionally, voltage, power, and frequency data can be collected in real time from SCADA systems or smart meters; Step 12: Based on the acquired real-time operating data, calculate dynamic characteristic indicators as dynamic identifiers. Dynamic characteristic indicators include adjustment margin indicators, stability indicators, response latency indicators, and collaborative efficiency indicators. (1) Adjustment margin index The minimum value of the power regulation ratio and the voltage deviation ratio is used as the regulation margin of the aggregated resources to characterize the regulation capability. The specific expression is as follows: (1); In the formula, This represents the power that aggregated resource i can participate in adjusting; This represents the rated power of aggregated resource i; This indicates the maximum permissible voltage deviation. This represents the voltage of the currently aggregated resource i. Indicates the rated voltage of the current aggregated resource i; The regulation margin index considers both power capacity and voltage stability to avoid over-adjustment or exceeding limits. Therefore, this index calculates a multi-constraint-based regulation margin for dynamically optimizing dispatch decisions in virtual power plants.

[0026] (2) Stability Indicators The output fluctuation of aggregated resources over different time periods is used as a metric to characterize the output stability of aggregated resources, thereby determining whether they are suitable as adjustment units. The specific expression is as follows: (2); In the above formula, This represents calculating the standard deviation of a variable over T time periods throughout the day. This represents the output of the i-th aggregated resource during time period t.

[0027] The stability index in this embodiment reflects the degree of fluctuation in resource output. The smaller the value, the more stable the power output of the resource, making it suitable as a scheduling unit with high stability requirements. The larger the value, the more unpredictable the output changes may occur, resulting in higher adjustment risks.

[0028] (3) Response delay index The time difference between the scheduling instruction response and the response time is used as an indicator of the response sensitivity of the aggregated resource, characterizing the timeliness of the aggregated resource in responding to the scheduling instruction. The specific expression is: (3); In the above formula, Indicates the time when the scheduling instruction was issued; This indicates the moment when the aggregated resource completes its response.

[0029] The response latency metric reflects the speed at which aggregated resources execute scheduling commands. A smaller value indicates a more timely response and higher execution efficiency, suitable for critical scheduling units requiring rapid adjustments. A larger latency indicates a delayed response, which may affect system coordination and security. This metric can be used to help select resources with rapid response capabilities, improving the overall timeliness and reliability of the virtual power plant.

[0030] (4) Collaborative efficiency indicators The weighted combination of electrical coupling degree and power covariance among resource nodes is used as a coordination efficiency index to measure the adjustment and coordination capabilities among resources, and is used to characterize the coordination level of resources in the power grid structure and operation behavior. The specific expression is as follows: (4) In the above formula, and These represent the per-unit values ​​of the node voltages of the lines where aggregated resources i and j are located, respectively. Indicates the rated voltage of the node; This represents the impedance between i and j; The whole can be used to measure the electrical coupling between two points, therefore Weights representing the degree of electrical coupling; and This represents the standard deviation of power variation calculated based on the historical power / load curves of i and j, respectively. This represents the covariance given by the power sequences of i and j.

[0031] The collaborative efficiency index comprehensively considers the tightness of electrical connections and behavioral coordination between resources. A higher value indicates a better joint adjustment capability between the two resources, making them suitable for processing as joint units in scheduling and helping to improve the accuracy and stability of multi-resource collaborative adjustment. This index is an important basis for realizing resource hypergraph partitioning and scheduling path optimization. The above index values ​​will be used as the basic input data for generating scheduling instructions during the VPP resource aggregation process.

[0032] Step 2 is the process of constructing a behavioral model for aggregating resources in a virtual power plant; Aggregated resources can be categorized into retail users, industrial users, distributed energy storage, and traditional generating units. Since the output curves and cost functions of traditional generating units are stable and predictable within a given production plan, and the load changes of industrial users are rigid due to production reasons, there is no need for specific behavioral modeling to characterize their uncertainty. Distributed energy storage, in this embodiment, is illustrated using photovoltaic and wind power as examples. In step 2, based on the operational characteristics and sources of uncertainty of different aggregated resource types, a behavioral model of the aggregated resource is constructed for each aggregated resource node, including: Step 21, Retail User Behavior Modeling: For retail user load, an electricity price response model with asymmetric utility function and perceived probability correction is used as the behavior model to characterize the load adjustment behavior of retail users under different electricity price scenarios; Specifically, for retail users, their load changes are sensitive to price; therefore, price changes can be considered the core variable of their load changes, i.e., the source of uncertainty for retail users. The following utility model can be established: (5); in, This indicates the change in time-of-use electricity prices; and These represent the risk preference coefficients of users under different benefit domains; This represents the loss aversion coefficient.

[0033] The above equation (5) is modified by adding a probability weighting coefficient to adjust the user's perception of electricity price changes, specifically as follows: (6); in, This represents the probability of users revising their perception of changes in electricity prices. To learn the fitting parameters.

[0034] Combining the above two equations, we can obtain the probability distribution of user load adjustment as follows: (7); In the formula, A probability distribution function representing changes in user load; This indicates that the data obtained is from user satisfaction data. The maximum value.

[0035] The parameter fitting and distribution function generation for the above formulas can be achieved using the FL (Federated Learning) algorithm, and the specific process will not be elaborated here.

[0036] Step 22, Energy Storage Unit Behavior Modeling: For energy storage units, with the goal of maximizing charging revenue and minimizing charging and discharging costs and battery cycle life loss, an objective function is constructed to characterize the charging and discharging behavior and operation strategy of energy storage under different electricity prices and SOC states. Without considering other factors, the charging and discharging behavior of energy storage mainly depends on its cost per charge and discharge cycle. Generally, the charging and discharging plan of energy storage should minimize the cost over its cycle life.

[0037] According to the Wöhler model for energy storage cycle life decay, the number of charge-discharge cycles is: (8); In the above formula, This indicates the number of charge-discharge cycles of the energy storage battery. and These are the battery material parameters; It represents the change in the state of charge and is used instead of representing the depth of charge and discharge.

[0038] For simplicity of analysis, this embodiment assumes that the cost of a single charge-discharge cycle is the amortized cost of the total cycle life over the energy storage lifespan. Therefore, according to equation (8), the cost of a single cycle of energy storage is: (9); In the above formula, This represents the baseline cost of a single energy storage cycle; Indicates the reference depth for charging and discharging.

[0039] The following objective function is constructed to incorporate the energy storage charging and discharging schedule into the planning: (10); In the above formula, This represents the time-of-use electricity price for period t; This represents the charging and discharging power of energy storage during time period t. This indicates that energy storage is being discharged, at which point revenue is generated; when This indicates that energy storage and charging incur costs. This represents the cost of energy storage cycle losses during time period t; A coefficient representing the cost of energy storage cycle losses.

[0040] The constraints of the objective function of the above formula (10) include the state of charge constraints of energy storage, charge and discharge constraints, etc. By fitting historical data and solving the above equations, the objective function (10) can be obtained, and the charging and discharging power variation law (power variation distribution) of energy storage under different electricity prices and SOC states can be obtained. Finally, the executable charging and discharging power command can be generated and sent to the inverter.

[0041] Step 23: Distributed energy unit behavior modeling: For distributed energy units, a deep learning model is used to predict the short-term power output of distributed energy units. This model is used to construct a distributed energy unit behavior model to characterize the power fluctuation behavior of distributed resources under different time series and meteorological conditions. It can be directly used for subsequent power regulation and fluctuation suppression. The output of photovoltaic and wind power is uncertain. In order to incorporate them into aggregated resource modeling, we can train a deep learning model to predict their output and use the predicted sequence as the distribution of their output variation.

[0042] In terms of scene coverage, training stability, and the coherence of subsequent aggregation algorithms, this embodiment uses an improved adversarial network WGAN-GP to generate predicted sequences for distributed energy resources.

[0043] The improved adversarial network WGAN-GP, similar to general adversarial networks, also includes a discriminator and a generator. The improvement lies in using Wasserstein distance loss to construct the training model as follows: The discriminator's loss function is constructed as follows: (11); In the above formula, Indicate the process of finding the expected value; This represents the generator's predicted sequence; Represents a true historical sample sequence; This represents the probability distribution generated by the generator; Represents the distribution of the real samples; This represents the discriminator.

[0044] To improve training stability, the following gradient penalty is added during the training process of the improved adversarial network WGAN-GP; (12); In the above formula, Let represent the loss function with respect to the global gradient. This represents the penalty coefficient with respect to the gradient; Represents the interpolation sequence between the real sample and the generated sample; This represents calculating the gradient with respect to the interpolation sequence; Two-dimensional module.

[0045] By training a WGAN-GP based on equations (11) and (12), the output distribution sequence of distributed energy can be obtained. The results given in the above steps can be directly used for the active power control of distributed energy to smooth fluctuations.

[0046] Step 3, Aggregation and Generation of Decision Units: Based on the dynamic identifiers and behavioral models of the aggregated resource nodes, the collaborative relationships and adjustment capabilities among resources are determined. Distributed resources with similar response characteristics and spatiotemporal coupling relationships are grouped into the same hyperedge as a decision unit, a hypergraph structure is constructed, and the decision units are divided, including the following steps: Step 31: After completing the dynamic identification and behavior modeling of aggregated resources, calculate the behavioral correlation between different aggregated resource nodes, construct all aggregated resource nodes into a directed graph, and use each hyperedge of the directed graph as a decision unit to construct the hypergraph structure. Step 32: Calculate the behavioral correlation between resource nodes, and then obtain the conflict index and hyperedge weight of the directed graph node in the hyperedge, so as to quantify its performance ability in collaborative regulation. After completing the dynamic identification and behavior modeling of aggregated resources, the idea of ​​hypergraph aggregation can be adopted to transform all nodes of the virtual power plant aggregated resources into a directed graph. Through the hypergraph aggregation algorithm, different aggregated resources are assigned to different hyperedges. Each hyperedge, as a decision-making unit, contains a group of behavioral coordination nodes, such as photovoltaic, energy storage and users in the same power distribution area.

[0047] First, the behavioral correlation degree between two resource nodes is defined with reference to equation (4), specifically as follows: (13); In the above formula, This represents the sequence of behavioral changes for the i-th aggregated resource, determined by its category. For retail users, This represents load variation; for distributed energy resources, It is the output sequence; for energy storage, It is the charging and discharging power; the meanings of other parameters are consistent with equation (4).

[0048] Based on the behavioral correlation of resource nodes, define the node element k in the directed graph in the hyperedge l The conflict index is: (14); In the above formula, Indicates the superedge l The number of units in the array.

[0049] Furthermore, the node element k in the hyperedge can be defined by the following formula. l The weighting coefficients are: (15); Step 33, Classification and Authority Assignment of Decision Units: Combining the dynamic characteristic indicators and superedge weights of aggregated resources, decision units are divided into first-level decision units that directly participate in scheduling and second-level decision units that are allocated through resource optimization scheduling. In this embodiment, they are divided into three categories: bidding units, adjustment units, and standby units. The decision unit classification rules are shown in Table 1. Among them, dynamic characteristic indicators include adjustment margin indicators, stability indicators, response delay indicators, and collaborative efficiency indicators. Table 1. Rules for dividing decision-making units;

[0050] in, To set a threshold; Based on the above division rules, the corresponding decision-making units are granted decision-making authority in the following manner: Level 1 Decision-Making Unit: The decision-making unit is a bidding unit that meets the grid access conditions, and the virtual power plant directly acts as the bidding unit to participate in day-ahead bidding; Secondary decision-making units: Units other than bidding units conduct internal transactions through P2P transactions, using a points-based accounting system. Fund settlement is completed by the virtual power plant within the day, and then internal settlement is carried out according to the points record.

[0051] In step 4, a two-layer decision-making model for virtual power plants participating in the spot market is constructed. The upper layer makes clearing decisions with the goal of maximizing market returns, while the lower layer allocates resources with the goal of minimizing response accuracy and adjustment costs. When traditional virtual power plants participate in the spot market, due to differences in the regulatory characteristics of their internal aggregated resources and the mechanisms of profit sharing and risk sharing, their scheduling decisions should be an equilibrium of external conditions and internal game dynamics. Therefore, this embodiment constructs the following two-layer model to address the problem: Step 41: The lower-level model in the two-level decision-making model is constructed as an internal resource regulation model: For the second-level decision-making unit in the decision-making unit, with the goal of maximizing the resource correlation between resource providers and resource demanders and minimizing the conflict and distance, an objective function is constructed as the lower-level internal resource regulation model. When scheduling resources, since the primary decision-making unit can directly participate in the scheduling and energy supply, the purpose of this step in building the model is to optimize the allocation of the remaining resources after the primary decision-making unit has allocated the resources. For the secondary decision-making unit, a blockchain is deployed for the internal transactions of the virtual power plant. The allocation of internal resources is matched through smart contracts, and the activities of resource regulation are recorded through blocks.

[0052] To determine the optimal matching decision for resource allocation and control within the virtual power plant, the following objective function is constructed: (16); In the above formula, i refers to the resource demander and j refers to the resource provider. Therefore Represents the set of resource demanders; Represents a set of resource providers; This represents the Euclidean distance between two transaction units: the provider and the demander. A coefficient representing Euclidean distance; This is the penalty coefficient; The matching state variables representing the two parties in the transaction are subject to the following constraints: (17); in, This indicates the maximum flexible resource declaration amount for resource provider j; This represents the maximum adjustable power of resource provider j; This represents the maximum flexible resource demand of resource demander i.

[0053] Optionally, based on blockchain, the internal resource allocation of the secondary decision-making unit adopts a points-based ledger system. For every 1MW of flexible resources transacted, both parties to the transaction receive one point, and the exchange rate between points and cash is: (18); In the above formula, This indicates the amount of legal tender that can be exchanged for one unit of points; This represents the adjustment factor, which is generally taken as 0.8-1.2; This indicates the time-of-use electricity price corresponding to the time the transaction occurred; The credit rating of unit i can be estimated by the fulfillment rate of the unit's historical transactions.

[0054] The above model demonstrates that incentivizing resources within a VPP aggregate to participate in grid dispatch can significantly improve resource response performance.

[0055] The transaction volume and price involved in the above steps can be realized and recorded through smart contracts. Since the blockchain technology involved is only used as a tool in this embodiment, it will not be described in detail.

[0056] Step 42: The upper-level model in the two-layer decision-making model is constructed as a market and dispatch collaborative model: Based on the conditional value at risk (VaR) to measure market risk, the upper-level market and dispatch optimization model is constructed with the objectives of maximizing returns, controlling risks, and ensuring power grid security. The objective function is as follows: (19); In the above formula, This indicates the winning bid price for each time period within the day during the day-ahead phase of the virtual power plant; This indicates the amount of bids submitted by the virtual power plant for each time period within the day during the day-ahead phase; This indicates the time-of-use electricity price for different time periods within the day; This represents the bid deviation of the virtual power plant during the intraday period t. Indicates the risk aversion coefficient; Indicates confidence level; Indicates the power grid risk penalty coefficient; As a power grid risk indicator, it can be described by the frequency of risk events such as the number of voltage over-limit incidents; This represents the conditional value of risk in a risky scenario. The value-at-risk sequence under uncertain scenarios is expressed by the formula: (20); In the formula, n represents the total number of risk scenarios generated; This represents the winning bid price for the day-ahead period t under scenario i; This represents the time-of-use electricity price for period t within a day under scenario i; This represents the energy storage cycle loss cost in scenario i.

[0057] Optionally, the risk scenarios for the market and scheduling optimization model can be generated and optimized using WGAN-GP.

[0058] Optionally, the constraints of the market and scheduling optimization model include transaction constraints, power balance constraints, and power flow constraints.

[0059] The constraints of the market and scheduling optimization model at the transaction level are as follows: (twenty one); In the formula, FU represents the set of first-level trading units; This represents the output of the k-th primary trading unit during time period t; This indicates the power / flexibility resource demand gap that is being met by the secondary trading unit; This indicates the amount of bids placed in the previous period for time period t.

[0060] In step 5, the optimization solution of the two-level model is as follows: real-time operating data of each aggregated resource and the resource demand in the spot market are obtained. The alternating direction multiplier method (ADMM) is used to solve the two-level decision model to obtain the scheduling scheme of each aggregated resource, realizing the coordinated optimization of market decision-making and internal scheduling. This includes the following steps: Step 51: Convert the two-level decision model into a dual two-level problem. The upper level is a clearing decision subproblem with the goal of maximizing market returns and controlling risks, and the lower level is a resource allocation subproblem with the goal of minimizing response accuracy and adjustment costs. It can be seen that the problem to be optimized is a dual, bi-level problem. The problem of the upper-level model can be represented as: (twenty two); In the above formula, This represents the actual transaction volume within the virtual power plant during time period t; Alternating direction multipliers; Represents the dual variable; This represents a one-dimensional modulus.

[0061] The problem of the lower-level model can be represented as: (twenty three); In the above formula, This represents the time cost of matching i and j.

[0062] Meanwhile, the dual variable can be updated according to the following formula: (twenty four); This two-level model can be solved using ADMM. Therefore, under the two-level decision-making mechanism, the virtual power plant can perform rolling decision optimization in the following manner: Step 52, Initialization: Obtain real-time operating data of each aggregated resource and the resource demand in the spot market, and initialize the initial bid quantity to the total schedulable power of the first-level decision-making unit that can be directly scheduled in the decision-making unit; The initial bid quantity for the virtual power plant is set as the total dispatchable power of the first-level decision-making unit, such as... (25) In the above formula, This indicates the initial virtual power plant bidding decision quantity, and the superscript indicates the iteration step of the rolling decision. This represents the initial value of the virtual power plant's bid for time period t during the day-ahead phase. This represents the output of decision unit k in the first-level decision unit at time t; This represents the set of first-level decision-making units within a decision-making unit; Step 53: Calculate the demand gap: Based on the difference between the total schedulable power of the primary decision-making unit and the target bid amount, obtain the amount of resources to be scheduled in the current round. As shown in the following formula: (26); In the above formula, m represents the m-th step of the rolling decision; This indicates the amount of adjustable and flexible resources required in the current rolling process; Step 54: Solve the lower-level decision model to obtain the resource scheduling results of the second-level decision unit, and return the actual resource quantity at the lower level. ; Step 55: Based on the resource scheduling results of the secondary decision-making unit, solve the upper-level model of the two-level decision-making model and update the bidding quantity of the virtual power plant. + (27) Step 56: Obtain the predicted electricity price and the state of the decision unit updated in Step 3, return to Step 52, traverse all time periods until the algorithm converges, and finally output the optimal decision scheduling amount.

[0063] This embodiment effectively improves the accuracy of identifying resource regulation capabilities and response behaviors by constructing a dynamic indicator system and differentiated behavior modeling methods for aggregated resources, significantly reducing scheduling failure rates and execution lag rates. It employs a hypergraph aggregation method to form collaborative scheduling units with behavioral consistency and electrical coupling characteristics, enhancing resource synergy and system controllability. By introducing CVaR (Value at Risk) and grid penalty factors at the optimization layer, it achieves a synergistic balance between maximizing returns and controlling risks, adapting to the highly uncertain characteristics of spot market prices and operating states. Combined with a two-layer optimization architecture and rolling update mechanism, it enhances the real-time performance and robustness of system scheduling, ensuring long-term implementability and dynamic self-adaptation. Simultaneously, through point-based P2P transactions and a blockchain ledger, it achieves traceability and compliance of internal resource trading processes, solving the problem of poor implementation of traditional virtual power plant internal coordination mechanisms. Overall, it significantly improves the resource allocation efficiency and operational capabilities of virtual power plants in a spot market environment.

[0064] Example 2 Based on Embodiment 1, this embodiment provides 8. A virtual power plant multi-unit coordinated optimization scheduling system based on the spot market, including: The dynamic identifier calculation module is configured to acquire real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. The behavior modeling module is configured to construct a behavior model of aggregated resources based on the operational characteristics and uncertainty sources of different aggregated resource types, using response functions or power change distributions for scheduling calculations. The decision unit generation and partitioning module is configured to determine the collaborative relationship and adjustment capability between resources based on the dynamic identifier and behavior model of the obtained aggregated resource nodes, and to classify distributed resources with similar response characteristics and spatiotemporal coupling relationship into the same hyperedge as a decision unit to construct a hypergraph structure. The decision model building module is configured to build a two-layer decision model, with the upper layer aiming to maximize market returns and make clearing decisions, and the lower layer aiming to minimize response accuracy and adjustment costs in resource allocation. The solution module is configured to acquire real-time operating data of each aggregated resource and the resource demand in the spot market. Based on the constructed hypergraph structure, it uses the alternating direction multiplier method to perform rolling optimization on the two-level decision model to obtain the scheduling scheme of each aggregated resource.

[0065] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment 1, and their specific implementation process is the same, so it will not be repeated here.

[0066] Example 3 Based on Embodiment 1, this embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps in the virtual power plant multi-unit coordinated optimization scheduling method based on the spot market described in Embodiment 1.

[0067] Example 4 Based on Embodiment 1, this embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they complete the steps in the virtual power plant multi-unit coordinated optimization scheduling method based on the spot market described in Embodiment 1.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0069] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A multi-unit coordinated optimization scheduling method for virtual power plants based on the spot market, characterized in that, Includes the following steps: Acquire real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. Based on the operational characteristics and uncertainty sources of different aggregated resource types, a behavioral model of aggregated resources is constructed using the response function or power change distribution used for scheduling calculations. Based on the dynamic identifiers and behavior models of the aggregated resource nodes, the collaborative relationships and adjustment capabilities among resources are determined. Distributed resources with similar response characteristics and spatiotemporal coupling relationships are grouped into the same hyperedge as a decision-making unit to construct a hypergraph structure. A two-level decision-making model is constructed, with the upper level making clearing decisions with the goal of maximizing market returns, and the lower level allocating resources with the goal of minimizing response accuracy and adjustment costs; By acquiring real-time operational data of each aggregated resource and the resource demand in the spot market, and based on the constructed hypergraph structure, the alternating direction multiplier method is used to solve the two-level decision model to obtain the scheduling scheme of each aggregated resource.

2. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 1, characterized in that: The process of generating dynamic identifiers for each aggregated resource node includes the following steps: Obtain real-time operating data for each aggregated resource, including data such as voltage, power, and frequency; Based on the acquired real-time operational data, dynamic characteristic indicators are calculated as dynamic identifiers. These dynamic characteristic indicators include adjustment margin indicators, stability indicators, response latency indicators, and collaborative efficiency indicators.

3. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 2, characterized in that: The minimum value of the power regulation ratio and the voltage deviation ratio is used as the regulation margin of the aggregated resources; The degree of output fluctuation of aggregated resources in different time periods is used as a stability indicator; The time difference in response to scheduling instructions is used as an indicator of the response sensitivity of aggregated resources. The weighted combination of electrical coupling degree and power covariance between resource nodes is used as a collaborative efficiency index to measure the adjustment and coordination capabilities between resources.

4. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 1, characterized in that: Based on the operational characteristics and sources of uncertainty of different aggregated resource types, a behavioral model of the aggregated resource is constructed for each aggregated resource node, including: For retail user load, an electricity price response model with asymmetric utility function and perceived probability correction is used as a behavioral model to characterize the load adjustment behavior of retail user load under different electricity price scenarios. For energy storage units, an objective function is constructed with the goal of maximizing charging revenue, minimizing charging and discharging costs and battery cycle life loss. This function is used to characterize the charging and discharging behavior and operation strategy of energy storage under different electricity prices and SOC states. For distributed energy units, a deep learning model is used to predict the short-term power output of distributed energy units, which serves as the constructed behavior model for distributed energy units.

5. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 1, characterized in that: Based on the obtained dynamic identifiers and behavioral models of aggregated resource nodes, the collaborative relationships and adjustment capabilities among resources are determined. Distributed resources with similar response characteristics and spatiotemporal coupling relationships are grouped into the same hyperedge as a decision unit, a hypergraph structure is constructed, and the decision units are divided, including the following steps: Calculate the behavioral correlation between different aggregated resource nodes, construct a directed graph of all aggregated resource nodes, and use each hyperedge of the directed graph as a decision unit to construct the hypergraph structure; Calculate the behavioral correlation between resource nodes, and then obtain the conflict index and hyperedge weight of the directed graph node in the hyperedge; By combining the dynamic characteristic indicators and super-edge weights of aggregated resources, the decision-making units are divided into first-level decision-making units that directly participate in scheduling and second-level decision-making units that allocate resources through optimized scheduling.

6. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 1, characterized in that: The constructed two-level decision-making model includes: The lower-level model is constructed as an internal resource regulation model. For the second-level decision-making unit in the decision-making unit, the objective function is constructed with the goal of maximizing the resource correlation between resource providers and resource demanders and minimizing the conflict and distance. This objective function serves as the lower-level internal resource regulation model. The upper-level model is constructed as a market and scheduling collaborative model: based on conditional risk value to measure market risk, the upper-level market and scheduling optimization model is constructed with the goals of maximizing returns, controlling risks, and ensuring power grid security.

7. The virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in claim 1, characterized in that: The alternating direction multiplier method is used for rolling optimization to solve the two-level decision model, and the scheduling schemes for each aggregated resource are obtained, including the following steps: Step 51: Convert the two-level decision model into a dual two-level problem. The upper level is a clearing decision subproblem with the goal of maximizing market returns and controlling risks, and the lower level is a resource allocation subproblem with the goal of minimizing response accuracy and adjustment costs. Step 52: Obtain real-time operating data of each aggregated resource and the resource demand in the spot market, and initialize the initial bid quantity to the total schedulable power of the first-level decision-making unit that can be directly scheduled in the decision-making unit; Step 53: Based on the difference between the total schedulable power of the primary decision-making unit and the target bid amount, obtain the amount of resources to be scheduled in the current round; Step 54: Solve the lower-level decision model to obtain the resource scheduling results of the second-level decision unit, and return the actual resource quantity of the lower level; Step 55: Based on the resource scheduling results of the secondary decision-making unit, solve the upper-level model of the two-level decision-making model and update the bidding quantity of the virtual power plant. Step 56: Obtain the predicted electricity price and the state of the decision unit updated in Step 3, return to Step 52, traverse all time periods until the algorithm converges, and finally output the optimal decision scheduling amount.

8. A virtual power plant multi-unit coordinated optimization scheduling system based on the spot market, characterized in that, include: The dynamic identifier calculation module is configured to acquire real-time operating data of each aggregated resource, and calculate dynamic characteristic indicators based on the electrical characteristics and regulation capabilities of the aggregated resources, which serve as dynamic identifiers for each aggregated resource node in the virtual power plant. The behavior modeling module is configured to construct a behavior model of aggregated resources based on the operational characteristics and uncertainty sources of different aggregated resource types, using response functions or power change distributions for scheduling calculations. The decision unit generation and partitioning module is configured to determine the collaborative relationship and adjustment capability between resources based on the dynamic identifier and behavior model of the obtained aggregated resource nodes, and to classify distributed resources with similar response characteristics and spatiotemporal coupling relationship into the same hyperedge as a decision unit to construct a hypergraph structure. The decision model building module is configured to build a two-layer decision model, with the upper layer aiming to maximize market returns and make clearing decisions, and the lower layer aiming to minimize response accuracy and adjustment costs in resource allocation. The solution module is configured to acquire real-time operating data of each aggregated resource and the resource demand in the spot market. Based on the constructed hypergraph structure, it uses the alternating direction multiplier method to perform rolling optimization on the two-level decision model to obtain the scheduling scheme of each aggregated resource.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the processor executes the computer instructions, it performs the steps in the virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the steps in the virtual power plant multi-unit coordinated optimization scheduling method based on the spot market as described in any one of claims 1-7.

Images