Load type virtual power plant resource pool optimization method based on industrial risk modeling

Abstract

The invention discloses a load type virtual power plant resource pool optimization method based on industrial risk modeling, and belongs to the technical field of power systems and energy internet. The method comprises the steps of firstly constructing a resource association network based on an industry input-output relationship; simulating a dynamic propagation process of industrial risks in a resource network by adopting an SIRS model, and depicting an evolution law of the long-period adjustable capacity of the resources; a short-period disturbance scene is constructed by further combining a power grid topology fault and scheduling demand fluctuation; generating a multi-time scale scene tree by combining long and short period uncertainty, and extracting a typical scene set by clustering; and finally, a double-layer random optimization model, an upper layer optimization resource pool structure selection and a lower layer optimization scheduling strategy are constructed, a resource long-period stability coefficient is introduced into the lower layer model through capacity boundary constraint, and collaborative optimization of the structure and scheduling is achieved. According to the method, the income redemption capability and the anti-risk robustness of the load-type virtual power plant in a complex and uncertain environment are effectively improved.

Description

Technical Field

[0001] This application relates to the field of power system operation and control technology, and in particular to a load-based virtual power plant resource pool optimization method based on industry risk modeling. Background Technology

[0002] Current methods for constructing resource pools for load-based virtual power plants are mostly based on the screening and aggregation of individual resources. They typically assume that the size and adjustment capacity of the resource pool remain constant during the contract period and directly extrapolate intraday scheduling models for long-term revenue evaluation. However, load-based virtual power plants use industrial enterprises as the aggregation entities, and their adjustability is significantly affected by industry cycles, order fluctuations, policy adjustments, and other industry factors, exhibiting strong volatility and uncertainty.

[0003] Currently, resource response uncertainty has become a key factor restricting the stable operation and commercialization of load-based virtual power plants. Constrained by inconsistencies in communication and control standards and imperfect market trading mechanisms, aggregators need to lock in adjustable resources in advance through long-term contracts to build a responsive resource pool. Therefore, the performance capability of a virtual power plant is highly dependent on the operational stability of the initial resource pool. Since load-based virtual power plants are aggregated by users, their adjustability is directly affected by industry cycles and order fluctuations, exhibiting significant volatility and uncertainty. This mismatch between resource providers prioritizing their main production activities and aggregators pursuing controllability and stable revenue further exacerbates resource response uncertainty, impacting the performance capability and economic returns of virtual power plants. Existing research mainly focuses on three dimensions—resource characteristic modeling, agent behavior modeling, and system uncertainty modeling—to improve the adjustment performance and profitability of resource pools under uncertain environments. In the resource dimension, the research covers modeling the steady-state capacity characteristics and dynamic response characteristics of resources, capturing the response behavior of resources at different time scales. In the subject dimension, the research introduces risk preference functions and prospect theory to characterize the response intentions of resource providers, and combines game theory to describe multi-objective conflict scenarios. In the system dimension, the research further considers external uncertainties such as demand fluctuations, grid topology changes, and communication system stability to improve the model's adaptability to real-world environments. However, existing research often focuses on technical optimization, typically using the implicit assumption of "constant resource pool size" to directly extrapolate intraday scheduling models to long-term benefit assessments, neglecting the real-world problem of resource pool capacity fluctuating over time during long-term operation. In summary, existing research provides insights into understanding and handling the uncertainty of virtual power plant regulation capabilities in different dimensions, but still has two limitations: 1) Lack of modeling for inter-resource industrial linkages and risk propagation mechanisms. Existing studies mostly use individual resources as modeling units, estimating the resource cluster regulation capacity through linear superposition and mapping transformation, which fails to reflect the risk transmission and linkage effects caused by industrial coupling between resources. Although complex network research has explained the risk propagation mechanism from the perspective of linkage structure, related methods have not yet been effectively embedded into the modeling system of virtual power plant resource pools. Existing resource pool optimization research lacks a systematic characterization of industrial risk transmission paths and their impact on resource regulation capacity.

[0004] 2) Lack of generalized modeling methods applicable to large-scale screening of heterogeneous resources. The industrial load resources aggregated by virtual power plants are highly heterogeneous in terms of energy consumption characteristics, response behavior, and control costs. Traditional individual-based refined models are difficult to effectively screen and combine when faced with massive resources. There is an urgent need to build an optimization method system that integrates structural cognition and generalization capabilities.

[0005] To address the aforementioned issues, this application takes load-type virtual power plants that aggregate industrial loads as the research object, focuses on resource volatility analysis driven by industry risks, supplements and improves risk characterization with short-cycle disturbances, and constructs a multi-timescale evaluation and optimization methodology for resource pool adjustment capabilities based on the dynamic evolution of resource pool adjustability within the contract period. This aims to enhance the operational robustness and revenue protection capabilities of virtual power plants in complex and uncertain environments. Summary of the Invention

[0006] The purpose of this application is to provide a load-based virtual power plant resource pool optimization method based on industry risk modeling. By constructing an industry risk-driven resource association network, the method simulates the coupled impact of long-cycle risk propagation and short-cycle physical disturbances, thereby achieving coordinated optimization of resource pool structure and scheduling strategy.

[0007] To achieve the above objectives, this application provides the following solution: This application provides a method for optimizing the resource pool of a load-based virtual power plant based on industry risk modeling, characterized by comprising: Construct a resource association network that reflects the characteristics of industry risk propagation; Multiple long-period scenarios and their corresponding probabilities are generated based on the resource association network. Generate multiple short-cycle scenarios and their corresponding probabilities; For each long-cycle scenario, a corresponding short-cycle scenario set is generated nested, and a two-level nested joint scenario tree is constructed. Determine the resource adjustable capacity stability coefficient under any joint scenario; A comprehensive evaluation index is constructed based on the mean and standard deviation of the resource adjustable capacity stability coefficient under the aforementioned joint scenario; Based on the evaluation indicators, the joint scenarios are divided into three risk level scenarios: normal, balanced, and extreme. The K-means clustering method is used to screen the typical scenarios for each scenario set to obtain a typical scenario subset. A two-layer stochastic optimization model is constructed using the aforementioned subset of typical scenarios as input; the two-layer stochastic optimization model includes an upper-layer model and a lower-layer model. The two-layer model is transformed into a single-layer linear optimization problem for solution, thereby obtaining the optimal resource pool structure and corresponding scheduling strategy.

[0008] Optionally, constructing a resource association network that reflects the characteristics of industry risk propagation specifically includes the following steps: Based on the industry input-output table, calculate the propagation weight of nodes within the industry and the propagation weight of enterprise nodes between industries to form a basic interconnection network. By combining regional enterprise market capitalization ranking and electricity consumption ranking data, an industry heterogeneity indicator is constructed from three dimensions: ranking overlap rate, ranking consistency, and ranking standard deviation. The edge weights of the basic association network are modified based on the industry heterogeneity index to obtain a resource association network that reflects the characteristics of industry risk propagation.

[0009] Optionally, generating multiple long-period scenarios and their corresponding probabilities based on the resource association network specifically includes: Using the resource association network as a propagation carrier, the SIRS model is used to model the state of resource nodes, including normal, infected, and recovered states. Construct node susceptibility parameters based on enterprise micro-attributes; Calculate the node state transition probability based on the node susceptibility parameters; Based on the node state transition probability, the long-term evolution process of resource adjustable capacity under different industry risk types is simulated, a long-term resource capacity stability coefficient is defined, and multiple long-term scenarios and their corresponding probabilities are generated.

[0010] Optionally, generating multiple short-period scenarios and their corresponding probabilities specifically includes the following steps: Based on the power grid topology, a short-term physical availability matrix for the resource pool is generated by simulating line fault events. Generate a scheduling demand activation matrix based on historical scheduling data; By combining the resource physical availability matrix and the scheduling demand activation matrix, a short-term adjustable capacity coefficient for resources is defined, and multiple short-cycle scenarios and their corresponding probabilities are generated through multiple rounds of sampling.

[0011] Optionally, the upper-level model uses resource selection variables as decision variables, with the objective of maximizing the expected value of scheduling revenue in all scenarios, and constraints including fixed resource costs; the lower-level model, given the resource selection results, aims to meet the scheduling needs of each time period, and constraints include response latency, response rate, and capacity boundary constraints, wherein the instantaneous adjustable capacity in the capacity boundary constraints is determined by the product of the maximum adjustable capacity of the resource and its capacity stability coefficient in the corresponding joint scenario.

[0012] Optionally, the expression for the node propagation weight within the industry is as follows: ; in, This refers to the funds invested by a particular industry for its own development. Indicates industry I For the industry K The invested funds Indicates the propagation weight of nodes within the industry. This represents the total number of industries. Indicates industry Resources , Indicates industry Resources .

[0013] The expression for the inter-industry enterprise node propagation weight is as follows: ; in, This indicates investment funds between industries. This indicates the propagation weight of enterprise nodes across industries.

[0014] Optionally, the node susceptibility parameter constructed based on enterprise micro-attributes specifically adopts the following formula: ; in, Indicates the node susceptibility parameter. These represent the weights of each indicator. Indicates the company's profit margin. Indicates the company's leverage ratio. Indicates company size. Indicates the growth rate of enterprise size. Indicates the degree of corporate information disclosure. This indicates the company's risk appetite.

[0015] Optionally, the edge weights of the basic association network are modified based on the industry heterogeneity index using the following formula: ; in, This indicates the modified edge weight. Indicates inter-industry before the revision With the industry Node propagation weight, Represents a node Industry-related indicators of electricity consumption Represents a node Industry-related indicators of electricity consumption.

[0016] Optionally, the expression of the upper-level model is as follows: ; ; in, x This indicates the resources to be filtered. N This represents the total amount of resources to be filtered. Indicates a specific scenario, Indicates the total number of scenes. This indicates the probability of a scenario occurring. T Indicates the total response time of the resource pool. t Indicates the specific response time. i Indicates the resource sequence number. Choose variables for resources; Representing resources The dispatch price, Represents the scenario at time t. The response quantity after selecting a set of resources. This indicates the degree of importance the optimization objective places on the resource pool's response performance. Indicates the first Moment Scene scheduling requirements Indicates fixed resource costs, This indicates the upper limit of the number of resources that can be aggregated. Enterprises are adjustable resource providers; therefore, enterprises and resources are considered the same concept.

[0017] Optionally, the expression of the lower-level model is as follows: ; ; in, for t The scheduling requirement for a time period is the response requirement for that time period. With demand time series The product; The constraints include: Response delay constraint: ; in, Representing resources The time delay from receiving an instruction to the actual response; Response rate constraint: ; in, For resources The maximum regulation rate; Capacity boundary constraints: ; in, Representing resources Maximum adjustable capacity, For resources in the scene The capacity stability coefficient is as follows.

[0018] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a load-based virtual power plant resource pool optimization method based on industry risk modeling, which has the following significant advantages: Enhanced return realization capability: Although the expected return of the constructed long-term stable resource pool is slightly lower than that of traditional methods, the return is consistently higher in actual operation, especially in extreme scenarios, indicating that it has a stronger return realization capability. Enhance resilience against risks: By identifying and avoiding high-volatility industry resources (such as oil and chemicals), and prioritizing the aggregation of highly stable industries (such as general equipment manufacturing and R&D services), the risk contagion path can be effectively suppressed and the volatility of returns can be reduced. Achieve multi-timescale collaborative modeling: Integrate industrial economic risks (long cycle) and power grid physical disturbances (short cycle) into the scenario generation framework to more realistically reflect the dynamic evolution of resource availability; Supports efficient screening of large-scale resources: By constructing generalized evaluation indicators and clustering of typical scenarios, it balances model accuracy and computational efficiency, and is suitable for the aggregation optimization of massive heterogeneous industrial loads. Provides a sustainable operation path: It provides a resource pool construction paradigm that balances economy and reliability for load-based virtual power plants to move from the "pilot demonstration" stage to the "commercial promotion" stage. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a load-based virtual power plant resource pool optimization method based on industry risk modeling in one embodiment of this application; Figure 2 This application provides a mapping relationship between enterprise nodes and power grid nodes in one embodiment. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] See Figure 1 , Figure 1 This diagram illustrates the virtual power plant resource pool optimization framework under multi-timescale scenarios in this application. In the resource risk simulation module, a resource association network is constructed based on inter-industry production relationships. The SIRS infection model is used to simulate the propagation and evolution of resource states within the network, depicting the trend changes in the long-term adjustable capacity of the resource pool and forming a long-term industry risk scenario. In the multi-timescale scenario construction module, the short-term adjustability of resources is assessed by combining the power grid topology and demand disturbance factors, constructing a short-term disturbance scenario. By combining long and short-term scenarios, a multi-stage scenario tree is constructed to support the optimization analysis of the resource pool. In the resource pool optimization decision module, based on the constructed multi-timescale scenario tree, a stochastic optimization model is designed. Starting from two levels—resource selection optimization and operation scheduling optimization—it improves the resource pool's ability to realize benefits and its risk control level under complex environments.

[0024] In one exemplary embodiment, such as Figure 1 As shown, a method for optimizing the resource pool of a load-based virtual power plant based on industry risk modeling is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, it includes the following steps: Step 1: Construct a resource association network that reflects the characteristics of industry risk propagation.

[0025] Specifically, it includes: Step 1.1: Based on the industry input-output table, calculate the propagation weight of nodes within the industry and the propagation weight of enterprise nodes between industries to form a basic association network.

[0026] Step 1.2: Combining regional enterprise market capitalization ranking and electricity consumption ranking data, construct industry heterogeneity indicators from three dimensions: ranking overlap rate, ranking consistency, and ranking standard deviation.

[0027] Step 1.3: Based on the industry heterogeneity index, modify the edge weights of the basic association network to obtain a resource association network that reflects the characteristics of industry risk propagation.

[0028] Specifically, to describe the dynamics of industry risk propagation in resources, this application uses graph theory to represent resources and their relationships in the form of nodes and edges: edges reflect economic connections between industries, and nodes embody the heterogeneity of a firm's adjustable resources after disturbances. Using the SIRS (Susceptible-Infected-Recovery-Susceptible) model, node states are classified into three categories: normal (S), infected (I), and recovered (R), characterizing the resource state transition process and providing a dynamic evolutionary basis for subsequent stability analysis of resource pool adjustment capabilities.

[0029] This embodiment constructs a resource edge weight model from two aspects: propagation path and industry structure. First, the economic risk propagation path is characterized based on the industry input-output relationship, forming a basic network of connections between resources. Second, considering the structural differences in economic scale and energy consumption levels among enterprises within the industry, an evaluation index for enterprise heterogeneity within the industry is constructed to correct the edge weights of the risk propagation path and enhance the model's ability to characterize industry electricity consumption fluctuations under risk.

[0030] 1) Risk transmission path modeling Considering data availability, this application uses the industry-specific capital flow definition in the input-output table to determine the propagation weights: (1) Industry-specific enterprise node propagation weight (1) in, This represents the funds invested in the development of a particular industry. The ratio of internal capital investment to total investment is used as the industry's propagation weight, reflecting the degree of interconnectedness among resources within the industry and its sensitivity to external shocks. This represents the funds invested by industry I in industry K.

[0031] (2) Inter-industry enterprise node propagation weight (2) in, This represents inter-industry investment. The ratio of investment from one industry to other industries to total investment is used as the inter-industry propagation weight, measuring the strength of cross-industry dependence and risk transmission among enterprise nodes.

[0032] Due to differences in production structure, electrification level, and value creation methods across industries, their risk propagation characteristics in interconnected networks exhibit heterogeneity. To accurately quantify network edge weights, this application proposes an edge weight adjustment method that integrates multi-dimensional industry characteristics based on ranking similarity theory. By integrating regional enterprise market capitalization ranking and electricity consumption ranking data, indicators reflecting industry heterogeneity are constructed from three dimensions: enterprise distribution, correlation, and volatility. A fuzzy membership function is used to map differences in industry characteristics, dynamically adjusting network edge weights to reflect the impact of the coupling characteristics of industry electricity consumption behavior and value creation on risk propagation paths.

[0033] (1) Ranking overlap rate: This measures the proportion of companies in a certain industry that simultaneously possess both high market value and high electricity consumption characteristics in the total number of companies in that industry. The higher this indicator is, the stronger the consistency and homogeneity of companies within that industry in terms of economic scale and energy consumption level.

[0034] (3) in, Indicates industry The number of companies appearing on both rankings This indicates the industry in the region. How many companies does the CCP include?

[0035] (2) Ranking consistency: This measures the consistency between the market capitalization of enterprises and their electricity consumption ranking. The smaller the value of this indicator, the more significant the correlation between the market capitalization and electricity consumption of enterprises within the industry.

[0036] (4) (5) (6) (7) in, Indicates industry The relative positions of the listed companies on the market capitalization table. Indicates enterprise The serial number on the list, This represents the total number of companies on the market capitalization table. This indicates the company's relative position on the power consumption meter. Indicates enterprise The serial number on the power consumption meter This indicates the total number of companies on the power consumption meter.

[0037] (3) Ranking standard deviation: indicates the degree of dispersion of the ranking difference of enterprises in the two dimensions of market value and electricity consumption, reflecting the heterogeneity of the correlation between market value and electricity consumption within the industry. The larger the value, the higher the degree of difference and structural differentiation among enterprises.

[0038] (8) (9) in, This represents the maximum difference in rankings among all listed companies.

[0039] Combining the above indicators, a comprehensive indicator reflecting the correlation between industry market capitalization and electricity consumption is constructed: (10) (11) in, , indicating industry The correlation between market capitalization and electricity consumption.

[0040] Risk does not spread uniformly within a network, but is dominated by the evolution of "critical bottleneck" industries. To enhance the network model's ability to represent potential systemic risks and more accurately characterize the criticality and dominance of industries in the risk propagation process, a risk bottleneck effect is introduced. By comparing the market capitalization-electricity consumption correlation between industries at both ends of the risk propagation path, the industry with the smaller index value and weaker risk tolerance is identified as the bottleneck industry in this propagation path. Based on this, the inter-industry electricity consumption correlation index is adjusted for risk sensitivity, forming a complete edge weighting for inter-resource electricity consumption correlation: (12).

[0041] Step 2: Generate multiple long-period scenarios and their corresponding probabilities based on the resource association network.

[0042] Specifically, it includes: Step 2.1: Using the resource association network as a propagation carrier, the SIRS model is used to model the state of resource nodes, including normal, infected, and recovered states.

[0043] Step 2.2: Construct node susceptibility parameters based on enterprise micro-attributes.

[0044] Based on the network structure, considering the heterogeneous response capabilities of enterprise nodes to external shocks, node susceptibility parameters are constructed. By embedding enterprise micro-attributes into the propagation network, the differential response behavior of nodes in the risk contagion process is characterized, and an industrial risk propagation network model with both structural and behavioral characteristics is constructed.

[0045] (13) in, This indicates the weight of each indicator. Indicates the company's profit margin. Indicates the company's leverage ratio. Indicates company size. Indicates the growth rate of enterprise size. Indicates the degree of corporate information disclosure. This indicates the company's risk appetite.

[0046] Step 2.3: Calculate the node state transition probability based on the node susceptibility parameters.

[0047] To systematically characterize the impact of industry risks on the operational stability of the resource pool, industry risk scenarios are divided into three typical types according to their propagation characteristics and initial impact features. The parameter settings are shown in Table 1.

[0048] Table 1 SIRS parameter settings

[0049] Under the influence of industry risks, enterprise nodes will exhibit the following three states: 1) S (Susceptible): The enterprise's adjustable resources are in a normal and available state; 2) I (Infected): The firm's adjustable resources decrease due to market, policy, or other external shocks; 3) R (Recovery): Enterprise's adjustable resources are restored to an available state after external conditions improve or strategies are adjusted.

[0050] The probability of resource state transitions at different time periods is jointly determined by risk propagation parameters, edge weights between resources, and node susceptibility, as defined below: (14) in, Represents a node Susceptibility parameters. This represents the impact propagation parameter, and its specific value is related to the risk type. Indicates the resource-related network edge weights, Represents nodes The sum of all weights of adjacent connected components.

[0051] Step 2.4: Based on the node state transition probability, simulate the long-term evolution process of resource adjustable capacity under different industry risk types, define the long-term capacity stability coefficient of resources, and generate multiple long-term scenarios and their corresponding probabilities.

[0052] Based on the above analysis, to quantify the impact of different resource adjustability, a long-term capacity stability coefficient for resources is defined and dynamically adjusted according to the infection status. (15) in, This indicates the impact of firm heterogeneity on adjustable capacity. This represents economic volatility. Among them, Indicates economic volatility. i.e., resources The correlation between the market capitalization of an industry and electricity consumption is a key indicator. The product of these two indicators represents the impact of economic fluctuations on the adjustable capacity of resources.

[0053] Step 3: Generate multiple short-cycle scenarios and their corresponding probabilities.

[0054] To improve the real-world adaptability of the resource pool optimization results, this application further constructs a short-cycle operation scenario. The short-term disturbance screening follows the following two principles: (1) the factor must have a direct impact on whether the resource can participate in the adjustment; (2) the factor must have a reliable data foundation and be quantifiable. Based on this, this application selects two types of key uncertainty factors: line faults and changes in scheduling requirements.

[0055] 1) Line fault factors reflect the vulnerability of physical access paths. When the upstream line of a resource access node fails, the resource loses its regulation capability due to power outage. By simulating line failure states at different times, a short-term physical availability matrix of the resource pool is formed.

[0056] 2) Changes in dispatch demand reflect the dynamic fluctuations in the power system's demand for regulatory resources. During a specific time period, if there is no response demand, resources will not be activated or allocated. A dispatch activation matrix is ​​generated through random sampling on the demand side, reflecting the probability of allocation at different times.

[0057] Treating line failures and demand response as independent random events, for any resource, at any given time, the resource only has actual callability when the physical path is available and there is a scheduling demand. Based on this, the short-term adjustable capacity coefficient of the resource is defined as follows: (16) in, The value is 1 when the resource is available and 0 when it is invalid. The value is 1 if the scheduling requirement exists, and 0 if it does not exist.

[0058] By conducting multiple rounds of independent sampling, a scenario set containing various combinations of short-term uncertainties is generated to assess the short-term adjustment capability of the resource pool.

[0059] Step 4: For each long-cycle scenario, generate a corresponding short-cycle scenario set nested within it, and construct a two-level nested joint scenario tree.

[0060] Step 5: Determine the resource adjustable capacity stability coefficient under any joint scenario.

[0061] Step 6: Construct a comprehensive evaluation index based on the mean and standard deviation of the resource adjustable capacity stability coefficient under the joint scenario.

[0062] Step 7: Based on the evaluation indicators, the joint scenarios are divided into three risk level scenarios: normal, balanced and extreme. The K-means clustering method is used to screen the typical scenarios for each scenario set to obtain a typical scenario subset.

[0063] Combining steps 4-7, in order to incorporate the aforementioned multi-timescale uncertainties into the resource pool optimization model, an application is made to generate a two-layer nested scenario tree structure based on the aforementioned long-cycle risks (industrial economic disturbances) and short-cycle risks (physical failures and scheduling demand fluctuations) to construct a set of typical representative scenarios.

[0064] 1) In long-term scenarios, the SIRS propagation model is used to simulate the diffusion process of risk in resource-related networks, obtaining the characteristics of resource adjustable capacity changes at different time periods (quarterly or annual scales). Each scene corresponds to a probability .

[0065] 2) Short-cycle scenarios are tailored to each long-cycle scenario. Generate a set of short-cycle scenes Simulate instantaneous disturbances during scheduling and assign probabilities to each short-cycle scenario. .

[0066] This leads to the construction of a complete joint operation scenario tree. : (17) In any joint scenario, a single enterprise node at time t The adjustable capacity stability coefficient is denoted as: (18) in, For a long period of time, For short cycles, .

[0067] The corresponding joint probability is: (19) To comprehensively characterize the evolution of the adjustable capacity of the resource pool under different disturbance scenarios and improve the robustness and adaptability of the optimization model to multiple risks, this application constructs the following comprehensive evaluation index based on the mean and standard deviation of the stability coefficient of the adjustable capacity of the resources generated by simulation: (20) in, The mean of the resource adjustable capacity stability coefficient. The standard deviation of the resource adjustability capacity stability coefficient. These are the weighting coefficients.

[0068] Based on the evaluation indicators, all joint scenarios are divided into three risk levels: "mild", "moderate" and "severe". Further, three sets of industry risk-physical risk coupling scenarios are constructed, as shown in Table 2.

[0069] Table 2. Methodology for constructing typical scenarios of multi-timescale industrial risk-physical risk

[0070] Considering the sparsity of scene sets in high-dimensional resource spaces, and the fact that structural relationships between resources can easily lead to highly similar scene response patterns, directly using the full sample would result in an excessive computational burden on the optimization model. This application uses the K-means clustering method to perform typical selection on different scene sets, and extracts a representative subset of typical scenes through compressed sampling.

[0071] While preserving the resource structure response characteristics and uncertainty distribution characteristics, the computational complexity of the optimization model is effectively reduced, and the model solution efficiency is improved. The final set of typical scenarios will serve as the input basis for the resource pool structure optimization and scheduling model.

[0072] Step 8: Using the subset of typical scenarios as input, construct a two-layer stochastic optimization model; the two-layer stochastic optimization model includes an upper-layer model and a lower-layer model.

[0073] Step 9: Transform the two-layer model into a single-layer linear optimization problem and solve it to obtain the optimal resource pool structure and corresponding scheduling strategy.

[0074] To balance the expected performance and response stability of resource structure under various shocks, this application constructs a two-layer stochastic optimization model based on a multi-timescale scenario tree to collaboratively optimize the resource pool structure and scheduling strategy. The model aims to maximize expected revenue; the upper layer optimizes the resource pool structure, while the lower layer solves the scheduling response strategy under a given resource set, achieving collaborative optimization under structural disturbances and scheduling uncertainties.

[0075] Upper-level model: Resource pool optimization considering cost and long-term adjustable capacity stability In long-term industry risk scenarios, the upper layer optimizes the structure and configuration of the resource pool. The goal is to maximize the expected return of the resource pool under all possible disturbance scenarios.

[0076] (twenty one) (twenty two) in, The selection of variables for resources is decided by higher levels. Indicating a disturbance scenario Next, the Given a set of resources at any given time Then, the response is calculated by the lower-level model; Representing a scene Download resources Dispatch price; This indicates the fixed cost of resources.

[0077] Lower-level model: Scheduling optimization considering response reliability Lower-level models in each scene The following is a detailed optimization scheme for resource scheduling, aiming to meet scheduling requirements while considering the physical constraints and adjustability of resource response. The scheduling optimization model is as follows: (twenty three) (twenty four) in, The scheduling demand for time period t is the response demand for that time period. With demand time series The product of.

[0078] The constraints include: 1) Response delay constraint: The output of resources is zero during the response delay time.

[0079] (25) in, Representing resources The time delay from receiving an instruction to the actual response.

[0080] 2) Response rate constraint: The change in the response of a resource within any time interval shall not exceed its maximum response rate. (26) in, For resources The maximum adjustment rate.

[0081] 3) Capacity boundary constraint: The response capability of a resource at any given time shall not exceed its instantaneous adjustable capacity, reflecting the availability boundary of the resource under the influence of industry risks. (27) in, Representing resources Maximum adjustable capacity, For resources in the scene The capacity stability coefficient is as follows.

[0082] To improve the computational efficiency of the model, this application further transforms the two-level stochastic optimization problem into a single-level optimization model. Since the lower-level objective is a convex function and satisfies linear constraints, auxiliary variables and linear inequality constraints are introduced based on the strong duality principle, transforming the problem into a single-level linear optimization problem, thus achieving a unified solution for resource selection and scheduling strategies.

[0083] To verify the effectiveness of the proposed method, this application conducts a case study based on the typical industrial structure of a city in Inner Mongolia. The study covers 257 enterprises across 14 typical industries. The industry classification and numbering are shown in Table 3.

[0084] Table 3 Industry Classification and Number of Various Resources

[0085] In terms of resource association modeling, to balance data integrity and availability, this application refers to the data provided in the 2018 China Time Series Input-Output Table to calculate the association weights between industries and between enterprise nodes within industries, constructing a resource association network. Based on the constructed industry risk propagation network, it simulates the long-term adjustable capacity evolution characteristics of the resource pool under three typical industry risk scenarios. Based on the 33-node distribution network model, it constructs the mapping relationship between enterprise nodes and power grid nodes, such as... Figure 2 As shown in Table 4, the distribution of enterprise resources within the power grid is as follows.

[0086] Table 4 Spatial Layout of Adjustable Resources in Distribution Network

[0087] 1) Combine historical operation and maintenance data of the region to estimate the failure probability of the line during the contract period, construct the line failure probability distribution, and generate a line failure dataset by simulating power supply path interruption through independent sampling.

[0088] 2) Based on the resource response history, establish a scheduling demand probability model and generate a set of resource response demands.

[0089] 3) By jointly sampling the two types of uncertainties, the short-term adjustable capacity stability coefficient matrix of each resource at different time stages is obtained, and combined with the long-term resource adjustable capacity stability coefficient matrix to form a complete multi-timescale scenario tree. Typical representative scenarios are extracted to construct sets of normal, equilibrium, and extreme scenarios.

[0090] A small virtual power plant was selected as a case study. The upper limit of resource aggregation was set to 10 enterprises. Resource pools were constructed using the method proposed in this application (long-term stable resource pool) and the control method (traditional resource pool that only considers resource adjustment characteristics and ignores the long-term reliability of the resource pool). Resource scheduling and benefit evaluation were carried out in three typical scenarios.

[0091] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for optimizing the resource pool of a load-based virtual power plant based on industry risk modeling, characterized in that, The load-based virtual power plant resource pool optimization method based on industry risk modeling includes: Construct a resource association network that reflects the characteristics of industry risk propagation; Multiple long-period scenarios and their corresponding probabilities are generated based on the resource association network. Generate multiple short-cycle scenarios and their corresponding probabilities; For each long-cycle scenario, a corresponding short-cycle scenario set is generated nested, and a two-level nested joint scenario tree is constructed. Determine the resource adjustable capacity stability coefficient under any joint scenario; A comprehensive evaluation index is constructed based on the mean and standard deviation of the resource adjustable capacity stability coefficient under the aforementioned joint scenario; Based on the evaluation indicators, the joint scenarios are divided into three risk level scenarios: normal, balanced, and extreme. The K-means clustering method is used to screen the typical scenarios for each scenario set to obtain a typical scenario subset. A two-layer stochastic optimization model is constructed using the aforementioned subset of typical scenarios as input; the two-layer stochastic optimization model includes an upper-layer model and a lower-layer model. The two-layer model is transformed into a single-layer linear optimization problem for solution, thereby obtaining the optimal resource pool structure and corresponding scheduling strategy.

2. The resource pool optimization method for load-type virtual power plants based on industry risk modeling according to claim 1, characterized in that, The construction of a resource association network that reflects the characteristics of industry risk propagation specifically includes the following steps: Based on the industry input-output table, calculate the propagation weight of nodes within the industry and the propagation weight of enterprise nodes between industries to form a basic interconnection network. By combining regional enterprise market capitalization ranking and electricity consumption ranking data, an industry heterogeneity indicator is constructed from three dimensions: ranking overlap rate, ranking consistency, and ranking standard deviation. The edge weights of the basic association network are modified based on the industry heterogeneity index to obtain a resource association network that reflects the characteristics of industry risk propagation.

3. The resource pool optimization method for load-type virtual power plants based on industry risk modeling according to claim 1, characterized in that, The generation of multiple long-period scenarios and their corresponding probabilities based on the resource association network specifically includes: Using the resource association network as a propagation carrier, the SIRS model is used to model the state of resource nodes, including normal, infected, and recovered states. Construct node susceptibility parameters based on enterprise micro-attributes; Calculate the node state transition probability based on the node susceptibility parameters; Based on the node state transition probability, the long-term evolution process of resource adjustable capacity under different industry risk types is simulated, a long-term resource capacity stability coefficient is defined, and multiple long-term scenarios and their corresponding probabilities are generated.

4. The load-based virtual power plant resource pool optimization method based on industry risk modeling according to claim 1, characterized in that, The generation of multiple short-period scenarios and their corresponding probabilities specifically includes the following steps: Based on the power grid topology, a short-term physical availability matrix for the resource pool is generated by simulating line fault events. Generate a scheduling demand activation matrix based on historical scheduling data; By combining the resource physical availability matrix and the scheduling demand activation matrix, a short-term adjustable capacity coefficient for resources is defined, and multiple short-cycle scenarios and their corresponding probabilities are generated through multiple rounds of sampling.

5. The resource pool optimization method for load-type virtual power plants based on industry risk modeling according to claim 1, characterized in that, The upper-level model uses resource selection variables as decision variables, with the goal of maximizing the expected value of scheduling revenue in all scenarios, and constraints include fixed resource costs. The lower-level model, given the resource selection results, aims to meet the scheduling needs of each time period, and the constraints include response latency, response rate, and capacity boundary constraints. The instantaneous adjustable capacity in the capacity boundary constraints is determined by the product of the maximum adjustable capacity of the resource and its capacity stability coefficient in the corresponding joint scenario.

6. The method for optimizing the resource pool of a load-type virtual power plant based on industry risk modeling according to claim 1, characterized in that, The expression for the node propagation weight within the industry is as follows: ； in, This refers to the funds invested by a particular industry for its own development. Indicates industry I For the industry K The invested funds Indicates the propagation weight of nodes within the industry. This represents the total number of industries. Indicates industry Resources , Indicates industry Resources . The expression for the inter-industry enterprise node propagation weight is as follows: ； in, This indicates investment funds between industries. This indicates the propagation weight of enterprise nodes across industries.

7. The resource pool optimization method for load-type virtual power plants based on industry risk modeling according to claim 1, characterized in that, The specific formula used to construct the node susceptibility parameter based on enterprise micro-attributes is as follows: ； in, Indicates the node susceptibility parameter. These represent the weights of each indicator. Indicates the company's profit margin. Indicates the company's leverage ratio. Indicates company size. Indicates the growth rate of enterprise size. Indicates the degree of corporate information disclosure. This indicates the company's risk appetite.

8. The method for optimizing the resource pool of a load-type virtual power plant based on industry risk modeling according to claim 2, characterized in that, The following formula is used to modify the edge weights of the basic association network based on the industry heterogeneity index: ； in, This indicates the modified edge weight. Indicates inter-industry before the revision With the industry Node propagation weight, Represents a node Industry-related indicators of electricity consumption Represents a node Industry-related indicators of electricity consumption.

9. The method for optimizing the resource pool of a load-type virtual power plant based on industry risk modeling according to claim 1, characterized in that, The expression for the upper-level model is as follows: ； ； in, x This indicates the resources to be filtered. N This represents the total amount of resources to be filtered. Indicates a specific scenario, Indicates the total number of scenes. This indicates the probability of a scenario occurring. T Indicates the total response time of the resource pool. t Indicates the specific response time. i Indicates the resource sequence number. Choose variables for resources; Representing resources The dispatch price, Represents the scenario at time t. The response quantity after selecting a set of resources. This indicates the degree of importance the optimization objective places on the resource pool's response performance. Indicates the first Moment Scene scheduling requirements Indicates fixed resource costs, This indicates the upper limit of the number of resources that can be aggregated.

10. The resource pool optimization method for load-type virtual power plants based on industry risk modeling according to claim 9, characterized in that, The expression for the lower-level model is as follows: ； ； in, for t The scheduling requirement for a time period is the response requirement for that time period. With demand time series The product; The constraints include: Response delay constraint: ; in, Representing resources The time delay from receiving an instruction to the actual response; Response rate constraint: ; in, For resources The maximum regulation rate; Capacity boundary constraints: ; in, Representing resources Maximum adjustable capacity, For resources in the scene The capacity stability coefficient is as follows.

Images