A virtual power plant reserve ancillary service bidding method considering physical settlement risk and economic tail risk

Abstract

A virtual power plant reserve auxiliary service declaration method considering physical realization risk and economic tail risk, comprising: constructing an asymmetric physical disturbance interval with a robust radius expansion according to source load prediction data to form a physical uncertainty set; generating a typical penalty rate scenario set reserving a high penalty rate tail scenario according to a deviation evaluation parameter; determining a day-ahead declaration power according to a commitment coefficient, and establishing a deviation evaluation cost model between the day-ahead declaration power and real-time public connection point exchange power; constructing a representative physical scenario representing key adverse realization in the physical uncertainty set; constructing an upper boundary identification model and a lower risk scheduling model to form a commitment boundary; performing declaration control according to the commitment boundary to output a day-ahead declaration power or an aggregated resource scheduling strategy. The method can output a quantitative declaration upper limit and a risk margin at the same time, avoid the simple combination of physical realization risk and economic tail risk, and significantly improve the robustness and implementability of the declaration strategy.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant optimized scheduling technology, specifically to a method for submitting standby ancillary services for virtual power plants that takes into account physical fulfillment risk and economic tail risk. Background Technology

[0002] With the continuous integration of new energy generation and distributed flexible resources into the power system, Virtual Power Plants (VPPs) utilize communication, aggregation control, and optimized dispatch technologies to organize dispersed resources such as wind power, photovoltaics, energy storage, controllable generation resources, and aggregated loads into integrated units capable of participating in market transactions and grid regulation. For standby ancillary services, VPPs must declare their available service capabilities to the ancillary services market during the day-ahead phase and undergo deviation assessments and fee settlements during the real-time operation phase based on dispatch instructions and actual execution. Therefore, developing a reasonable day-ahead declaration strategy under conditions of uncertainty on both the source and load sides is a core technical issue for VPPs participating in the standby ancillary services market. Existing literature has proposed technical solutions from different perspectives regarding the declaration and dispatch issues of VPPs participating in the ancillary services market.

[0003] Reference [1] (CN122267720A) discloses a method for regulating virtual power plants participating in day-ahead reserve ancillary services based on the feasible region of similar entropy convex hull. By constructing multidimensional similar entropy and convex hull feasible region constraints, a day-ahead optimization model is established under these constraints, and a sequence of power shortage and capacity penalty is introduced to output the day-ahead reserve adjustment application plan and regulation plan. This method optimizes under the given physical feasible region and historical data constraints, but in essence, it still focuses on the optimization scheduling within the feasible region and fails to output the quantitative application upper limit and risk margin based on the mapping relationship between application intensity and disturbance tolerance.

[0004] Reference [2] (CN111402015A) discloses a two-layer bidding method and system for virtual power plants based on purchase and sale risk. By establishing an operating cost model and a purchase and sale risk value model, the optimal solution set is obtained through multiple iterations and a cost risk map is drawn. The virtual power plant selects a bidding strategy according to its own risk preference. This method introduces risk measurement into the bidding decision, but its risk model focuses on the uncertainty of purchase and sale electricity prices, does not involve the physical realization risk in the standby ancillary service scenario, and does not consider the economic loss risk caused by the tail characteristics of the penalty rate in the deviation assessment.

[0005] Reference [3] (US20220284458A1 / US11887140B2) discloses a risk-constrained optimization method for virtual power plants. It uses Information Gap Decision Theory to handle the uncertainty of future market prices and second-order stochastic dominance constraints to handle the uncertainty of renewable energy and load, maximizing the expected total market revenue while minimizing the expected total energy cost. This method has advantages in modeling market price uncertainty, but it fails to distinguish the different impacts of wind power, photovoltaic and load deviation directions on the ability to fulfill commitments at the physical level, and it does not specifically handle the peak-heavy tail characteristics of penalty rates in deviation assessment scenarios. Reference [4] (CN117810991A) discloses a virtual power plant optimization scheduling method that considers conditional value at risk (CVaR) and carbon trading. It introduces CVaR into the optimization objective to measure the risk brought about by the uncertainty of renewable energy output and market electricity prices. Reference [5] (CN107464010A) discloses a virtual power plant capacity optimization configuration method that considers CVaR. It uses CVaR to measure the risk brought to planning by uncertainty and multiplies CVaR by a weighting coefficient to measure investor risk preference. The above methods use CVaR for planning or scheduling optimization of virtual power plants, but none of them model CVaR and physical disturbance boundaries in the standby ancillary service application scenario. They cannot answer the two key questions at the same time: "How wide is the application commitment range of source and load disturbance that can still be physically fulfilled?" and "Is the economic loss controllable under extreme penalty scenarios?"

[0006] Furthermore, existing technologies for handling source-load forecasting errors have significant shortcomings. Wind power, solar power, and load forecasting errors exhibit significant time-varying and directional asymmetry. For example, downward deviations in wind and solar power output, coupled with upward deviations in load, collectively increase net load and regulation pressure, posing a serious threat to the fulfillment of reserve commitments; while high renewable energy output coupled with low load demand leads to reverse regulation and curtailment pressures.

[0007] Existing methods mostly use a fixed proportion of symmetrical disturbance intervals, see references [1]-[5], which cannot fully characterize the differential impact of different source load variables and different deviation directions on the ability to fulfill commitments, resulting in bias in risk estimation.

[0008] Regarding deviation assessment, the deviation penalty rate in the settlement of standby ancillary services exhibits peak and heavy-tail characteristics. Although the probability of a few high penalty rate scenarios is low, once they are superimposed with large commitment deviations, they will cause significant economic losses. Existing methods mostly only consider expected costs (see references [1] and [2]), or only introduce CVaR at the planning and configuration level (see references [4] and [5]), lacking effective means to simultaneously retain high penalty rate tail scenarios and quantify economic tail risks at the day-ahead declaration level, resulting in the exposure of tail risks under high penalty rate scenarios. Considering only the physical disturbance boundary without identifying economic tail risks makes it difficult to assess the comprehensive risk level of the declaration strategy; considering only economic tail risks without identifying physical disturbance boundaries makes it difficult to determine whether the declaration commitment can be physically fulfilled.

[0009] In summary, existing methods for submitting virtual power plant standby ancillary service applications lack a technical solution that can simultaneously address the issues of physical feasibility and economic tail risks within a unified framework. There is an urgent need for a method for assessing and controlling the commitment boundaries of day-ahead submissions, quantifying the submission ceiling, risk margin, and corresponding dispatch strategies, and providing a clear basis for virtual power plants to select submission intensity under different market environments and risk appetites.

[0010] References: Reference [1]: "Control method and system for virtual power plant participation in day-ahead standby auxiliary services based on feasible domain of similar entropy convex hull", Chinese invention patent application, publication number CN122267720A, application date March 13, 2026, publication date June 23, 2026.

[0011] Reference [2]: "A two-layer bidding method and system for virtual power plants based on purchase and sale risk", Chinese invention patent application, publication number CN111402015A, application date March 17, 2020, publication date July 10, 2020, authorization announcement date June 27, 2023.

[0012] Reference [3]: SUN, HONGBO, MITSUBISHI ELECTRIC RESEARCH LABORATORIES, INC., "Risk-constrained Optimization of Virtual Power Plants in Pool and FutureMarkets", US Patent Application, Publication No. US20220284458A1, Application Date February 25, 2021, Publication Date September 8, 2022; Grant Patent No. US11887140B2, Grant Announcement Date January 30, 2024.

[0013] Reference [4]: ​​"A virtual power plant optimization scheduling method considering conditional risk value and carbon trading", Chinese invention patent application, publication number CN117810991A, application date January 2, 2024, publication date April 2, 2024.

[0014] Reference [5]: "A method for optimizing the configuration of virtual power plant capacity", Chinese invention patent application, publication number CN107464010A, application date June 29, 2017, publication date December 12, 2017; authorization announcement date October 29, 2021.

[0015] Reference [6]: Wei Zhinong, Chen Yu et al., “Multi-source capacity optimization configuration model of virtual power plant considering conditional risk value”, Automation of Electric Power Systems, 2018, Vol. 42, No. 4, pp. 40-47.

[0016] Reference [7]: Xu Hui, Jiao Yang et al., “Stochastic dispatch optimization model of wind-solar-gas-storage integrated virtual power plant considering uncertainty and demand response”, Power System Technology, 2017, Vol. 41, No. 11, pp. 3590-3597.

[0017] Literature [8]: Tan Z., Wang G., Ju L., et al., "Application of CVaR riskaversion approach in the dynamical scheduling optimization model for virtual power plant connected with wind-photovoltaic-energy storage system with uncertainty and demand response", Energy, 2017, Vol.124, pp.198-213. Summary of the Invention To address the problems of existing technologies that treat service commitments as exogenous inputs, lack coordinated assessment of physical fulfillment boundaries and economic tail risks, and struggle to output quantified declaration upper limits and risk margins, this invention provides a virtual power plant backup ancillary service declaration method that considers both physical fulfillment risks and economic tail risks. It characterizes the directional differences in uncertainty between the source and load sides through an asymmetric physical disturbance interval, captures the peak-and-fat tail characteristics of penalty rates using a tail-retention scenario set, and constructs a commitment boundary between the commitment coefficient and the maximum tolerable robust radius based on a two-layer optimization model. This transforms the declaration problem from operational optimization after a given declaration to an assessment of the safety commitment boundary under uncertain conditions, enabling the simultaneous output of quantified declaration upper limits and risk margins. It avoids the simple merging of physical fulfillment risks and economic tail risks, significantly improving the robustness and feasibility of the declaration strategy.

[0018] The technical solution adopted in this invention is as follows: A method for submitting applications for virtual power plant standby ancillary services that takes into account physical fulfillment risk and economic tail risk includes the following steps: S1: Obtain source and load forecast data, aggregated resource operation parameters, standby ancillary service application parameters, and deviation assessment parameters of the virtual power plant; S2: Construct an asymmetric physical perturbation interval that expands with the robust radius based on the source load prediction data, forming a set of physical uncertainties; S3: Generate a set of typical penalty rate scenarios that retain high penalty rate tail scenarios based on deviation assessment parameters; S4: Introduce a commitment coefficient, determine the day-ahead declared power based on the commitment coefficient, and establish a cost assessment model for the deviation between the day-ahead declared power and the real-time common connection point exchange power. S5: Construct a representative physical scenario that characterizes the key unfavorable realizations in the set of physical uncertainties, wherein the representative physical scenario includes a finite set of corner points; S6: Construct an upper-layer boundary identification model and a lower-layer risk scheduling model. Based on the representative physical scenario and the typical penalty rate scenario set, determine the maximum tolerable robust radius corresponding to different commitment coefficients under the unified budget tolerance upper limit constraint to form the commitment boundary. S7: Perform declaration control based on the stated commitment boundary, and output the day-ahead declared power or aggregated resource scheduling strategy.

[0019] In S1, The source-load forecast data includes wind power forecast, photovoltaic power forecast and load forecast for each scheduling period, and the time scale of the scheduling period includes 15 minutes, 30 minutes or 1 hour; The aggregated resource operating parameters include the upper and lower limits of output, ramping constraints and operating costs of controllable power generation resources, the charging and discharging power limits, upper and lower limits of energy state, charging and discharging efficiency, initial state and target terminal state of energy storage resources, as well as the power boundary and power purchase and sale state variables of the point of common coupling. The standby ancillary service application parameters include the baseline commitment capacity, commitment coefficient, capacity compensation price, and application period; the deviation assessment parameters include the deviation exemption threshold, the serious default threshold, the segmented penalty rate, and the assessment rules.

[0020] S2 includes: S2.1: For any source load variable in wind power, solar power, and load. Determine its point prediction value in each scheduling period. Lower deviation amplitude and upper deviation amplitude Construct the original prediction interval: ; in, Index for scheduling period; lower deviation magnitude and upper deviation amplitude The parameters are determined based on historical prediction error quantiles, conformal prediction intervals, operational experience thresholds, or prediction model outputs, and are set separately for different scheduling periods, different source load variables, and different deviation directions. S2.2: Introducing a robust radius , The original prediction interval is linearly expanded to obtain the lower bound of the expanded interval. and the upper bound of the interval : ; ; S2.3: Integrate the expanded intervals of the wind power, photovoltaic, and load source variables to obtain the physical uncertainty set. : ; in, , , Time periods Wind power, photovoltaic power and load power; , , These are the point prediction values ​​for the corresponding source load variables; , , , , , These represent the lower and upper deviation magnitudes of the corresponding source load variables, respectively. This is the set of scheduling time periods.

[0021] S3 includes: S3.1: Based on historical data of deviation penalty rates, ancillary service market assessment rules, or preset extreme penalty rate scenarios, generate initial penalty rate scenarios using probability density estimation and sampling methods; details are as follows: First, based on historical penalty rate samples for each scheduling period, kernel density estimation is used to obtain the corresponding cumulative distribution function. Then, stratified sampling is used to generate the initial scenario. During the period The penalty fee rate is: ; in, For the initial scene During the period Penalty rates; For time period The inverse function of the cumulative distribution function of penalty rates; This represents the initial number of scenes; The result is a random number within the interval (0,1). Stratified samples from different time periods are randomly paired to form the initial penalty rate scenario set. ,in, This is the initial set of scenes.

[0022] S3.2: Calculate the tail strength index for each initial penalty rate scenario, as follows: The tail strength of the scenario is calculated using a weighted sum of penalty rates for each time period: ; in, For the initial penalty rate scenario Tail strength index; For time period The weights can be set according to load peaks and valleys, the importance of reserve assessment, or a uniform weight; the meanings of the other symbols follow the definitions in S3.1 and S2.3.

[0023] The tail strength index is determined based on at least one of the following: the weighted sum of penalty rates for each time period in the scenario, the penalty rate range exceeding the preset quantile, and the duration of the high penalty rate period. The larger the tail strength index, the more significant the amplification effect of the scenario on the cost of deviation assessment. S3.3: Based on the tail strength index, retain several high-penalty-rate tail scenarios, and perform scenario reduction or clustering on the remaining scenarios to obtain representative scenarios; as follows: The initial scene is arranged according to Sort by size from largest to smallest, and keep the first few. One scenario is designated as a high-penalty-rate tail scenario; the remaining scenarios are clustered or reduced based on the vector composed of penalty rates for each time period, and the scenario closest to the center of each cluster is selected as the representative scenario. The probability quality corresponding to the deleted scenario is incorporated into the corresponding representative scenario.

[0024] in, The number of high-penalty-rate tail scenes to be directly retained can be determined based on the preset tail retention ratio or the upper limit of the number of scenes.

[0025] S3.4: Merge the high-penalty-rate tail scenarios with the representative scenarios, and normalize the scenario probabilities to obtain the typical penalty-rate scenario set. Specifically: The high-penalty-rate tail scenario set and the representative scenario set are combined into a typical penalty fee rate scenario set. The probability quality before normalization is determined according to the number of initial scenarios represented by each retained scenario. The normalized probability is: ; in, This is a typical set of penalty fee scenarios; For the scene Probability mass before normalization; For the scene The normalized probability satisfies and .

[0026] The processing described in S3.1 to S3.4 above corresponds to... Figure 3 The step of “constructing a typical penalty rate scenario” in the text.

[0027] S4 includes: S4.1: The optimal day-ahead power trajectory is obtained under the nominal scenario and used as a baseline commitment capability. The nominal scenario is one where the robust radius is zero and the penalty rate is the predicted value. S4.2: Based on the commitment coefficient Determine the power declared before the date : ; Wherein, the commitment coefficient A dimensionless parameter characterizing the intensity of the application; Abbreviated as ; S4.3: Power switching based on real-time common connection point Define commitment deviation and its absolute value : ; ; S4.4: Introduce piecewise deviation variable , , absolute value Perform linear decomposition and obtain the time period by weighted summation using segmented penalty rates. Deviation assessment cost : ; ; ; in, The threshold for exemption from liability for deviation. The threshold for severe default must be met. ; , , To achieve segmented penalty rates, satisfying .

[0028] S5 includes: S5.1: Constructing the basic corner set The basic corner point set This includes at least one of the following scenarios: nominal scenario, maximum net load scenario, minimum net load scenario, and single-source adverse scenario; wherein: The nominal scenario corresponds to point-based predicted values ​​for wind power, photovoltaic power, and load. The maximum net load scenario corresponds to the lower bound of the interval for wind power and photovoltaic, and the upper bound of the interval for load. The minimum net load scenario corresponds to the upper limit of the interval for wind power and photovoltaic, and the lower limit of the interval for load. The single-source adverse scenario corresponds to only wind power down-biasing, photovoltaic down-biasing, or load up-biasing; S5.2: Calculate the committed pressure index for each scheduling period. : ; in, For time period Net load power; For time period The energy available for energy storage; A preset positive number used to avoid a denominator of zero; It is a function with the deviation between the reported power and the net load power, the net load ramp-up rate, and the available energy storage as independent variables; S5.3: Selecting Commitment Pressure Indicators Time windows that continuously exceed the threshold During the time window The internal command takes the lower bound of the interval for wind power and photovoltaic power and the upper bound of the interval for load, generating local extreme disturbance scenarios and forming a commitment pressure-oriented enhanced corner point set. : ; in, A heuristic operator for encapsulating the process of time window filtering and interval boundary assignment; , , Time periods Forecast values ​​for wind power, solar power, and load points; This refers to the combined lower and upper deviations of each source load variable. S5.4: Merge the basic corner point sets With the aforementioned commitment pressure-guided enhanced corner point set , thus obtaining a finite set of corner points : .

[0029] In S6, the construction and solution of the lower-level risk scheduling model includes: S6.1: For a given commitment coefficient Robust radius and representative physical scenarios In typical penalty fee scenario sets Define risk adjustment costs The weighted sum of expected operating cost, economic tail risk term, and soft constraint penalty term: ; in, The expected operating costs include internal resource operation and maintenance costs, main grid electricity purchase and sales costs, and deviation assessment costs. This is the risk trade-off coefficient; Confidence level The conditional value at risk is used to measure the economic tail risk item. , Time periods The unit penalty price and reduction amount for flexible load shedding; For the set of scheduling periods; S6.2: Introducing Auxiliary Variables and super loss slack variables Conditional risk value The linear expression is: ; ; in, Confidence level Value at risk below; These are slack variables for the superloss; Index for penalty fee scenarios; For the scene Losses related to deviation assessments; For the scene The normalized probability; This is a typical set of penalty fee scenarios.

[0030] S6.3: With the objective of minimizing risk adjustment costs, a representative physical scenario is obtained by solving the following constraints: boundary constraints on power purchase and sale at the point of common coupling, mutual exclusion constraints on power purchase and sale, upper and lower limits and ramping constraints on controllable generation resource output, power constraints on energy storage charging and discharging, recursive constraints on energy storage state of mind, upper and lower limits constraints on energy storage state of mind, power balance constraints, deviation definition constraints, and step penalty linearization constraints. The optimal risk-adjusted cost is as follows: For a given commitment coefficient Robust radius and representative physical scenarios The lower-level decision variable set consists of internal distributed resource output, energy storage charging and discharging power and energy state, power purchased and sold at the point of common coupling, deviation segmentation variables, and soft constraint relaxation variables. Please provide a solution:

[0031] in, Representative physical scenarios The set of lower-level decision variables; Given decision variables, commitment coefficients, and robust radius, the risk adjustment cost in this physical scenario; The optimal risk adjustment cost is obtained to satisfy the operational constraints listed in this step. The penalty rate scenario only changes the economic settlement. The equipment output and energy storage status under the same representative physical scenario are shared among the penalty rate scenarios. The obtained optimal risk adjustment cost is used as the input of the upper-level boundary identification model in S6.4.

[0032] In step S6, the construction and solution of the upper boundary recognition model includes: S6.4: Define the worst-case risk adjustment cost The maximum value of the optimal risk adjustment cost under each representative physical scenario: ; in, A collection of representative physical scenarios; S6.5: Adjust costs based on reference risks With budget tolerance coefficient The unified budget tolerance limit is determined to be Among them, risk-adjusted costs are referenced. The nominal scenario risk adjustment cost is taken from the preset commitment coefficient, or it can be set by the user according to risk preference or market rules; S6.6: With the objective of maximizing the robust radius and the constraint that the adjustment cost of the worst-case risk does not exceed the unified budget tolerance limit, construct and solve the upper-level optimization model to obtain the commitment coefficient. The corresponding maximum robust radius : ; .

[0033] S6 further includes: S6.7: For a candidate robust radius, if the corresponding worst-case risk adjustment cost does not exceed the unified budget tolerance limit, then the virtual power plant is determined to be able to withstand source load disturbances under the candidate robust radius; otherwise, it is determined to exceed the tolerable boundary; specifically as follows: Define a budget discriminant function for candidate robust radii, and update the search interval using binary search as an example: ; in, For budget discriminant functions; and The first The lower and upper bounds of the search in each iteration; This is the candidate robust radius; and The definition in S6.5 is adopted. When At that time, update the lower bound of the search to Otherwise, update the upper search bound to... When the difference between the upper and lower bounds is not greater than the preset convergence accuracy. The iteration stops when the time is right, where, To determine the convergence accuracy of the robust radius search, output the last acceptable candidate value as... This process corresponds to Figure 4 The budget determination and search range update process is shown below.

[0034] Accordingly, an iterative search method is used to identify the maximum tolerable robust radius, including binary search, grid search, or heuristic search. S6.8: Perform the identification process of S6.7 on multiple commitment coefficients respectively to obtain the mapping relationship between the commitment coefficients and the maximum tolerable robust radius, forming the commitment boundary, and determine the upper limit of declaration and risk margin that meet the risk tolerance constraint based on the commitment boundary. Specifically as follows: S6.7 is applied to each commitment coefficient in the candidate commitment coefficient set to form discrete commitment boundaries, and the acceptable declaration range is determined accordingly: ; in, For the set of candidate commitment coefficients; The discrete commitment boundary is formed by the commitment coefficient and the maximum tolerable robust radius; To preset a safety margin; The set of acceptable commitment coefficients to meet safety margin requirements; The maximum commitment factor to meet the safety margin requirements; Candidate commitment coefficient The corresponding risk margin. When If the value is empty, it is determined that there is no claim strength that meets the safety margin under the current resource configuration; if it is not empty, it is determined that... The corresponding daily reported power is used as the reporting upper limit. The boundaries and area divisions are as follows: Figure 5 As shown.

[0035] In step S7, the maximum tolerable robust radius corresponding to the candidate commitment coefficient is compared with the preset safety margin; specifically as follows: First, obtain the set of acceptable commitment coefficients according to S6.8. .when When it is determined that there is no application intensity that meets the preset safety margin requirements under the current resource configuration, the aggregated resource reconfiguration or application suspension is triggered; when Then, the maximum tolerable robust radius corresponding to the candidate commitment coefficient is compared with the preset safety margin, and the declaration control is implemented in the following manner: When the maximum tolerable robust radius is lower than the preset safety margin, it is determined that the corresponding declaration intensity has a high risk. Based on the commitment boundary, the day-ahead declared power is reduced, the flexible resource reserved capacity is increased, or the energy storage operation strategy is adjusted. The adjusted day-ahead declared power or aggregated resource scheduling strategy is sent to the virtual power plant energy management system, the dispatch control system, or the ancillary service market declaration system. When the maximum tolerable robust radius is not lower than the preset safety margin, the day-ahead declared power corresponding to the candidate commitment coefficient is determined as the declared power within the acceptable declaration range, or the declared power with the better comprehensive evaluation result of benefit, risk and robust margin is selected from multiple candidate commitment coefficients.

[0036] The present invention provides a method for submitting virtual power plant backup ancillary services that takes into account physical fulfillment risk and economic tail risk, with the following technical effects: 1) Transform "operational optimization after a given declaration" into "commitment boundary assessment for safe declaration," achieving a quantitative output of the declaration ceiling and risk margin: A commitment coefficient is introduced to determine the day-ahead declared power. An upper-level boundary identification model and a lower-level risk scheduling model are constructed. The maximum tolerable robust radius is solved under a unified budget tolerance upper limit constraint. The upper-level model is further constrained to... Adjust costs for the worst-case scenario. To constrain this, an iterative search is performed on multiple commitment coefficients to form commitment boundaries and determine the upper limit of declarations and risk margins. Unlike existing technologies that use transaction plans as a prerequisite for optimization, the above-mentioned technology makes the declaration power itself the object of evaluation: each commitment coefficient necessarily corresponds to a quantifiable maximum tolerable disturbance range, thus solving the problems in the background technology of how much disturbance can be tolerated, whether the commitment can be fulfilled, and how much margin should be reserved.

[0037] 2) The asymmetric disturbance interval accurately characterizes the temporal differences and directional asymmetry of source load prediction errors, improving the accuracy of physical realization risk assessment: Determine the point prediction value for each source load variable. Lower deviation amplitude and upper deviation amplitude And with robust radius Linear extensions form physically uncertain sets The lower and upper deviation amplitudes are set separately for different scheduling periods, different source-load variables, and different deviation directions. This technical feature inevitably produces the following effects: the net load upward adjustment pressure formed by the superposition of the lower deviation of wind and solar power output and the upper deviation of load, as well as the reverse regulation and wind and solar curtailment pressure formed by high output of renewable energy and low load demand, can be characterized by the same framework in different directions and amplitudes. Compared with the fixed proportional symmetrical disturbance range of existing technologies, the identification accuracy of physical disturbance boundaries is significantly improved, avoiding the breach of commitment caused by overestimation of the disturbance range leading to conservative or underestimation of the declaration.

[0038] 3) The combination of a high-penalty-rate tail scenario retention mechanism and conditional risk value linearization effectively suppresses the economic tail risk exposure of deviation assessments: High-penalty-rate tail scenarios are directly retained based on the tail strength index. The remaining scenarios are reduced, clustered, merged, and their probabilities normalized, thus avoiding the meanization of high-penalty-rate scenarios from the perspective of scenario set construction. Furthermore, the conditional value-at-risk linear expression and its constraints described in S6.2 are adopted to allow the economic tail risk term to enter the objective function of the lower-level risk scheduling model. The algorithm's characteristics and the technical feature of the electricity market deviation assessment rules interact and work together. Although a few high-penalty-rate scenarios have low probabilities, their losses are explicitly amplified by the CVaR tail metric and participate in scheduling decisions. Compared with existing technologies that only optimize expected costs, the significant economic losses caused by the superposition of high-penalty-rate scenarios and large commitment deviations are suppressed in advance, and the exposure of deviation penalty risk is significantly reduced.

[0039] 4) Modeling the mechanisms of physical realization risk and economic tail risk, and conducting collaborative assessment under unified budget constraints, balancing assessment accuracy and solution efficiency: The risk of physical fulfillment is characterized by the asymmetric physical disturbance range and robust radius, while the economic tail risk is characterized by the tail retention scenario set and conditional risk value. Both are solved collaboratively under a unified budget tolerance ceiling constraint. The risks of these two different mechanisms are not simply merged into a single risk item; the "ability to physically fulfill" and "the economic cost of fulfillment being bearable" of the declared commitment are simultaneously assessed within the same framework. Furthermore, based on the basic corner point set... With commitment pressure-driven enhanced corner set The finite corner point set is used to represent key adverse realizations. While maintaining typical adverse directions (maximum net load, continuous high pressure period, etc.), it avoids the complete traversal of the continuous physical uncertainty set, reduces the computational scale of the two-layer model, and improves the solution efficiency. This is an improvement in the internal performance of the system brought about by the algorithm characteristics.

[0040] 5) The commitment boundary is linked to the closed-loop control of the declaration, providing a quantitative basis for the declaration intensity decision and improving the feasibility of the project: The mapping relationship between the commitment coefficient and the maximum tolerable robust radius forms the commitment boundary. This mapping directly reflects the compression effect of increased declaration intensity on the system's robustness margin, providing a quantitative decision-making basis for virtual power plants to select declaration intensity under different market environments and risk preferences. The dual-branch logic of the declaration control is as follows: when the maximum tolerable robust radius is lower than the preset safety margin, it triggers a reduction in day-ahead declared power, an increase in flexible resource reservation capacity, or an adjustment of energy storage operation strategy, and sends the adjustment results to the virtual power plant's energy management system, dispatch control system, or ancillary service market declaration system; when it is not lower than the safety margin, it directly determines the acceptable declared power. Thus, the evaluation results are not limited to offline analysis but have a closed-loop effect on the day-ahead declaration execution stage, improving the feasibility of the method in existing declaration control systems and forming a significant difference from the disconnect between existing technology optimization results and declaration decisions. Attached Figure Description

[0041] The present invention will be further described below with reference to the accompanying drawings and examples; Figure 1 This is a schematic diagram of the aggregated virtual power plant structure and the power exchange relationship at the point of common coupling of the present invention; Wherein: 10 represents the virtual power plant aggregation control center, 11 represents wind power resources, 12 represents photovoltaic resources, 13 represents controllable power generation resources, 14 represents energy storage resources, 15 represents aggregated load, 16 represents the ancillary service market declaration system, 17 represents the public connection point, and 18 represents the main grid.

[0042] Figure 2This is a schematic diagram of the cost assessment model for the deviation between the day-to-day reported power and the real-time point-of-compatibility switching power of the present invention.

[0043] Figure 3 This is a flowchart illustrating the application method for virtual power plant backup ancillary services that takes into account physical fulfillment risk and economic tail risk, as described in this invention.

[0044] Figure 4 This is a flowchart illustrating the solution process for the upper-layer boundary identification model and the lower-layer risk scheduling model of this invention.

[0045] Figure 5 This is a schematic diagram of the commitment boundary between the commitment coefficient and the maximum tolerable robust radius of the present invention.

[0046] Figure 6 This is a schematic diagram of the predicted power of a typical daily source load and the asymmetric physical disturbance range in Example 2 (corresponding to S2).

[0047] Figure 7 This is a schematic diagram of a typical penalty rate scenario set in Example 2 (corresponding to S3).

[0048] Figure 8 This is a schematic diagram (corresponding to S4) showing the variation of the most unfavorable risk adjustment cost with the robust radius and the determination of the maximum tolerable robust radius in Example 2.

[0049] Figure 9 This is a schematic diagram of the commitment boundary between the commitment coefficient and the maximum tolerable robust radius in Example 2 (corresponding to S5).

[0050] Figure 10 This is a schematic diagram of the risk adjustment cost composition under different commitment coefficients in Example 2 (corresponding to S6).

[0051] Figure 11 This is a schematic diagram comparing the declaration control execution before and after in Example 2 (corresponding to S7). Detailed Implementation

[0052] To facilitate understanding, the main terms of the present invention's method for submitting applications for virtual power plant backup ancillary services, which takes into account physical fulfillment risk and economic tail risk, are defined as follows: The “commitment coefficient” refers to a dimensionless parameter that characterizes the intensity of virtual power plant backup ancillary services application. The application power is obtained by multiplying the baseline commitment capacity by the commitment coefficient.

[0053] The "robust radius" is a non-negative parameter used to expand the range of source load disturbances. The larger the value, the wider the range of source load uncertainties considered.

[0054] "Maximum tolerable robust radius" refers to the maximum robust radius that can meet the risk adjustment cost requirements for a given commitment coefficient under the constraint of a uniform budget tolerance upper limit.

[0055] "Commitment boundary" refers to the mapping relationship between different commitment coefficients and the corresponding maximum tolerable robust radius, used to distinguish between the commitment fulfillment zone and the disturbance unbearable zone.

[0056] "Risk adjustment cost" refers to the comprehensive cost, including expected operating cost, economic tail risk, and soft constraint penalty, under a given physical implementation and penalty rate scenario. The economic tail risk is measured using conditional value at risk (VaR) to characterize the tail-expected level of deviation assessment losses in a few high-penalty-rate scenarios. Soft constraint penalty includes penalties for energy storage terminal state deviations, insufficient reserve margin penalties, or other flexible deviation penalties related to operational constraints.

[0057] "Representative physical scenarios" refers to a finite number of physical realizations selected from the set of physical uncertainties to characterize the pressure to fulfill commitments, including a finite set of adverse physical scenarios, a set of critical adverse scenarios, or a combination thereof.

[0058] The “Finite set of adverse physical scenarios” is realized by a finite set of corner points, including a finite number of typical implementations constructed by the combination of upper and lower bounds of each source load variable and continuous high-pressure periods.

[0059] The "uniform budget tolerance cap" refers to the upper limit of risk costs used to determine whether a candidate robustness radius is tolerable. It is determined by the reference risk-adjusted cost and the budget tolerance coefficient. The reference risk-adjusted cost can be the nominal scenario risk-adjusted cost corresponding to the preset commitment coefficient, or it can be set by the user based on risk preferences or market rules.

[0060] Example 1: This embodiment provides a method for submitting standby ancillary services for virtual power plants that takes into account physical fulfillment risk and economic tail risk. It can be applied to virtual power plants that aggregate wind power, photovoltaic power, controllable power generation resources, energy storage resources, and aggregated loads, and may also include other adjustable resources. Each step of the method in this embodiment can be implemented by a computer program executed by a processor in a virtual power plant aggregation control center, energy management server, or cloud computing platform.

[0061] like Figure 1 As shown, the virtual power plant aggregation control center 10 is connected to wind power resources 11, photovoltaic resources 12, controllable power generation resources 13, energy storage resources 14, and aggregated loads 15 via communication lines, and interacts with the ancillary service market declaration system 16 to exchange declaration information. The virtual power plant exchanges power with the main grid 18 through the common connection point 17, and the power exchanged through the common connection point is used to characterize the ability to fulfill the standby ancillary service commitments.

[0062] like Figure 2As shown, the virtual power plant generates day-ahead declared power during the day-ahead phase and real-time point-of-compatibility (POC) exchange power during the real-time operation phase. The difference between the two constitutes the commitment deviation, and its absolute value is incorporated into the tiered penalty rule to form the deviation assessment cost. This cost is further involved in the calculation of risk-adjusted costs.

[0063] like Figure 3 As shown, the virtual power plant backup ancillary service application method taking into account physical fulfillment risk and economic tail risk in this embodiment includes steps S1 to S7, which are executed sequentially in chronological order, as detailed below: S1: Obtain source-load forecast data, aggregated resource operation parameters, standby ancillary service application parameters, and deviation assessment parameters for the virtual power plant. The source-load forecast data includes wind power forecast, photovoltaic power forecast, and load forecast for each dispatch period. The dispatch period can be 15 minutes, 30 minutes, 1 hour, or other time scales suitable for day-ahead reporting. The units of each forecast power are power units (e.g., MW), consistent with the physical data corresponding to the input parameters in this application.

[0064] The aggregated resource operating parameters include the upper and lower limits of output, ramping constraints and operating costs of controllable power generation resources, the charging and discharging power limits, upper and lower limits of energy state, charging and discharging efficiency, initial state and target terminal state of energy storage resources, as well as the power boundary and power purchase and sale state variables of the point of common coupling.

[0065] The parameters for applying for backup ancillary services include the baseline commitment capacity, commitment coefficient, capacity compensation price, and application period. Deviation assessment parameters include the deviation exemption threshold, the serious default threshold, the tiered penalty rate, and the assessment rules; the tiered penalty rate is measured in units of electricity cost (e.g., yuan / MWh), while the deviation exemption threshold and the serious default threshold are measured in units of power or electricity, consistent with the format specified in the assessment rules.

[0066] S2: Construct an asymmetric physical perturbation interval that expands with the robust radius, forming a set of physical uncertainties: For any source-load variable among wind power, solar power, and load, determine the point forecast value, lower deviation magnitude, and upper deviation magnitude for each scheduling period. Let the variables be... During the period The predicted point value is The lower deviation is The upper deviation range is The original prediction interval is: (1); Introducing robust radius ( For the linear expansion of the above interval, the upper and lower bounds after expansion are as follows: (2); (3); By integrating the extended intervals of wind power, photovoltaic, and load variables, a global physical uncertainty set is obtained: (4); In equations (2) to (4), For scheduling period index, For the set of scheduling periods; , , Time periods Wind power, photovoltaic power and load power; , , These are the point prediction values ​​for the corresponding source load variables; , , , , , These represent the lower and upper deviation amplitudes for the corresponding source load variables, respectively, with dimensions consistent with the corresponding power. The lower and upper deviation amplitudes can be independently set for different source load variables, different time periods, and different deviation directions, and their values ​​vary with the robust radius. Linear expansion leads to the formation of an asymmetric physical uncertainty set.

[0067] It should be noted that the principle behind the above asymmetric setting is that the downward deviation in wind and solar power output and the upward deviation in load will jointly increase the net load and regulation pressure, while the high output and low demand of renewable energy will bring reverse regulation and wind and solar curtailment pressure. By adopting an asymmetric interval that can expand with the robust radius, it is possible to characterize the differentiated impact of different source and load variables and different deviation directions on the ability to fulfill commitments, and avoid the overestimation or underestimation of the ability to fulfill commitments by a fixed-proportion symmetrical disturbance interval.

[0068] S3: Construct a typical penalty fee rate scenario set: The uncertainty of deviation penalty rates is mainly reflected in their peak and heavy tail characteristics. To avoid high penalty rate scenarios being averaged out during the reduction of ordinary scenarios, this step uses a combination of tail retention and scenario reduction to construct a typical penalty rate scenario set.

[0069] Based on historical data of deviation penalty rates, ancillary service market assessment rules, or extreme penalty rate scenarios preset by the operator, initial penalty rate scenarios are generated using probability density estimation and sampling methods. A tail strength index is calculated for each scenario. This index can be determined based on the weighted sum of penalty rates for each time period within the scenario, the penalty rate margin exceeding a preset quantile, the duration of high penalty rate periods, or a combination thereof. A larger tail strength index indicates a more significant amplification effect of the scenario on deviation assessment costs.

[0070] Based on the tail strength index, several high-penalty-rate tail scenarios are directly retained; the remaining scenarios are clustered or otherwise reduced to select representative scenarios; finally, the high-penalty-rate tail scenarios and representative scenarios are merged, and the scenario probabilities are normalized to obtain a typical penalty rate scenario set. Thus, a few scenarios with low probability of occurrence but high penalty intensity are explicitly retained in the scenario set, providing a scenario basis for the accurate measurement of economic tail risk items in subsequent S6.

[0071] S4: Establish a cost model for assessing application power and deviation: Introducing a commitment coefficient The commitment coefficient represents the strength of the current declaration. Let the parameter be dimensionless. In the nominal scenario ( The optimal day-ahead power trajectory obtained through optimization (using the predicted penalty rate) serves as the baseline commitment capability. Then for a given The power rating declared recently is: (5); In equation (5), The following is abbreviated as Its dimension is power unit, which is one of the output results of this method.

[0072] During the real-time operation phase, the virtual power plant adjusts its internal resource output and the power purchased and sold from the main grid, thereby enabling the point of common coupling (PCC) to exchange power. Track the reported power as much as possible Define the commitment deviation and its absolute value: (6); (7); Deviation assessment costs are subject to tiered penalties: Let the exemption threshold be... The threshold for serious default is ,satisfy The corresponding penalty rates are as follows: , , ,satisfy Introducing segmented deviation variables , , right Linear decomposition: (8); (9); Time period The deviation assessment cost is: (10); Since the objective function is to minimize the total cost, the above linearized form automatically implements the priority filling of low-penalty segments, which is consistent with the tiered assessment rule, thereby transforming the nonlinear tiered penalty rule into a linear constraint that can be embedded in the optimization model.

[0073] S5: Construct a representative physical scenario: Complete traversal of a continuous set of physical uncertainties increases computational burden. To improve solution efficiency, a representative physical scenario is used to characterize the key adverse realizations in the set of physical uncertainties. This representative physical scenario is represented by a finite set of corner points.

[0074] Constructing a finite set of corner points It contains the basic corner set. Commitment pressure-driven enhancement corner set : (11); It is composed of the upper and lower bounds of each source load variable, including at least one of the following scenarios: nominal scenario, maximum net load scenario, minimum net load scenario, and single-source adverse scenario. The nominal scenario corresponds to the predicted values ​​of wind power, solar power, and load. The maximum net load scenario corresponds to the lower bound of wind power and solar power and the upper bound of load. The minimum net load scenario corresponds to the upper bound of wind power and solar power and the lower bound of load. The single-source adverse scenario corresponds to only causing wind power to be down-biased, solar power to be down-biased, or load to be up-biased.

[0075] Enhanced Corner Set Through heuristic operators Generate. First, calculate the commitment pressure index for each time period. , defined as a function of the deviation between the day-ahead declared power and net load, the net load ramp-up rate, and the available energy storage: (12); In equation (12), For time period Net load power, For time period The energy available for energy storage, This is a preset positive number used to avoid a denominator of zero. It is used to distinguish it from the deviation exemption threshold in equation (9). and severe default threshold In this invention, the aforementioned zero constant is denoted as... .

[0076] Select Time windows that continuously exceed the threshold Within this window, wind power and solar power are set to the lower bound of the interval, and load is set to the upper bound of the interval, forming a local extreme disturbance scenario. Operator Encapsulate the filtering and assignment process: (13); In equation (13), This refers to the combined lower and upper deviation amplitudes of each source load variable. The finite set of corner points obtained through equations (11) to (13) can reduce the computational scale while maintaining typical unfavorable directions, allowing the solution of the subsequent two-layer model to be completed within an engineering-acceptable timeframe.

[0077] S6: Construct an upper-level boundary identification model and a lower-level risk scheduling model, and determine the maximum tolerable robust radius under a unified budget tolerance upper limit constraint: For a given commitment coefficient Robust radius and representative physical scenarios The lower-level risk scheduling model in typical penalty fee scenario sets Solve for the risk-adjusted cost. Define the risk-adjusted cost. The weighted sum of expected operating costs, economic tail risk terms, and soft constraint penalties: (14); in, The expected operating costs include internal resource operation and maintenance costs, main grid electricity purchase and sales costs, and deviation assessment costs. This is the risk trade-off coefficient; Confidence level The conditional value at risk is used to measure the economic tail risk item. , These represent the unit penalty price and reduction amount for flexible load reduction, respectively.

[0078] Through auxiliary variables (VaR, or confidence level) Value at risk (VaR) and overloss slack variables Linear expression: (15); (16); in, Confidence level Value at risk below; These are slack variables for the superloss; Index for penalty fee scenarios; For the scene Losses related to deviation assessments; For the scene The normalized probability satisfies and ; This is a typical set of penalty fee scenarios.

[0079] The constraints of the lower-level risk scheduling model include: boundary constraints on power purchase and sale at the point of common coupling, mutual exclusion constraints between power purchase and sale, upper and lower limits and ramping constraints on controllable generation resource output, power constraints on energy storage charging and discharging, recursive constraints on energy storage state, upper and lower limits on energy storage state, power balance constraints, deviation definition constraints, and linearization constraints with step penalty. The lower-level model is solved with the objective of minimizing risk adjustment costs, yielding representative physical scenarios. The optimal risk adjustment cost.

[0080] like Figure 4 As shown, the upper-level boundary identification model is used to identify the maximum range of physical disturbances that a virtual power plant can withstand under a unified budget tolerance ceiling constraint, given a commitment coefficient. The worst-case risk adjustment cost is defined as the cost for each representative physical scenario. The maximum value of the risk adjustment cost at the next lower level: (17); in, It is a finite set of corner points or a representative set of physical scenes.

[0081] Let the reference risk adjustment cost be Budget tolerance factor is The unified budget tolerance limit is The upper-level optimization model is as follows: (18); (19); By solving equations (18) to (19), the commitment coefficient is obtained. Maximum tolerable robust radius The larger the value, the more source load disturbances the system can tolerate under the condition of uniform budget growth.

[0082] For each commitment coefficient An iterative search method is used to identify the maximum tolerable robust radius. If the worst-case risk adjustment cost corresponding to the candidate robust radius does not exceed the unified budget tolerance limit, then the virtual power plant can withstand source load disturbances under that robust radius; otherwise, it exceeds the tolerable boundary. As a specific binary search implementation: for a given... Initialize the lower bound of the search. Search upper bound Take the midpoint Substitute the values ​​into the lower-level risk scheduling model to solve for the optimal risk adjustment cost of each representative physical scenario, and take the maximum value according to equation (17); if the maximum value satisfies the constraint of equation (19), then let Otherwise Repeat the above process until the search interval width is less than the preset convergence accuracy, then output. .

[0083] Perform the above process on multiple commitment coefficients to obtain the mapping relationship between the commitment coefficients and the maximum tolerable robust radius, forming the virtual power plant backup ancillary service commitment boundary, such as... Figure 5 As shown. Figure 5 In the middle, the horizontal axis represents the commitment coefficient. The vertical axis represents the maximum robust radius that can be withstood. The commitment boundary curve divides the coordinate plane into a commitment fulfillable zone and a disturbance unacceptable zone: for a given commitment coefficient, when the actual considered source load disturbance range does not exceed the corresponding maximum tolerable robust radius, the declared commitment is in the commitment fulfillable zone; otherwise, it is in the disturbance unacceptable zone. As the commitment coefficient increases, the maximum tolerable robust radius generally shows a decreasing trend, reflecting the compression effect of increased declaration intensity on the system's robustness margin. The upper limit of declarations and the risk margin that satisfy the risk tolerance constraint are determined based on the commitment boundary.

[0084] S7: Implement declaration control based on the commitment boundary: When the maximum tolerable robust radius corresponding to the candidate commitment coefficient is lower than the preset safety margin, the declared strength is deemed to have a high risk. In this case, based on the commitment boundary, the day-ahead declared power is reduced, flexible resource reservation capacity is increased, or the energy storage operation strategy is adjusted. The adjusted day-ahead declared power or aggregated resource dispatch strategy is then sent to the virtual power plant energy management system, dispatch control system, or ancillary service market declaration system. Details are as follows: The set of acceptable commitment coefficients obtained according to S6.8 Make a judgment; when If, under the current resource configuration, it is determined that there is no application plan that meets the preset safety margin requirements, then aggregated resource reconfiguration or application suspension will be triggered; when And the current commitment coefficient does not belong to When that happens, reduce it to the smaller commitment coefficient adjacent to it within the set, or reduce it to... Based on this, the day-ahead declared power is recalculated, and the reduced declared amount is converted into ramp margin for controllable power generation resources, energy storage charge / discharge margin, or point of common coupling (PCC) power purchase / sale margin. The adjusted declared power and dispatch settings are sent to the corresponding systems through the declaration or dispatch interface of the virtual power plant energy management system.

[0085] When the maximum tolerable robust radius is not less than the preset safety margin, the day-ahead declared power corresponding to the candidate commitment coefficient is determined as the declared power within the acceptable declared range, or the declared power with the better comprehensive evaluation result of benefit, risk, and robust margin is selected from multiple candidate commitment coefficients. Specifically: If the current commitment coefficient belongs to When the corresponding day-ahead declared power can be directly used as an acceptable declaration scheme; if multiple acceptable schemes exist, the one that is adopted first should be used. The corresponding declared power, or based on preset comprehensive evaluation rules, in terms of expected return, risk-adjusted costs, and risk margin. Choose from the options.

[0086] Example 2: In one exemplary embodiment, the virtual power plant comprises wind power, photovoltaic power, controllable power generation resources, energy storage resources, and aggregated loads, with a scheduling period of 24 hours prior to the current day and a time resolution of 1 hour. This period and resolution are merely examples, and the invention is also applicable to shorter or longer scheduling periods.

[0087] Controllable power generation resources are configured with typical upper and lower limits for output and ramp-up capabilities; energy storage resources are configured with upper and lower limits for charge and discharge power, upper and lower limits for energy state, charge and discharge efficiency, initial energy state, and target terminal state; the point of common coupling (PCC) is configured with upper limits for purchased power, upper limits for sold power, and mutually exclusive state variables for power purchase and sale. These parameters can be adjusted according to different virtual power plant resource types and operating rules.

[0088] An optional quantitative implementation: scheduling time domain 24h, time resolution 1h; micro gas turbine output range 30MW~220MW; pumped storage power generation 180MW, pumping power 160MW, energy range 120MWh~900MWh; electrochemical energy storage charge / discharge power ±120MW, energy range 40MWh~320MWh; common connection point power purchase cap 50MW, power sales cap 30MW; wind power, photovoltaic and load disturbance amplitudes ±25%, ±35% and ±12% respectively; 15 typical penalty rate scenarios.

[0089] Budget tolerance coefficient CVaR confidence level Commitment coefficient The scan is performed in steps of 0.1 within the range of 0 to 1.2, meaning the scan points cover the two extreme values ​​of 0 and 1.2, as well as all intermediate values, thus completely depicting the overall shape of the commitment boundary. The above values ​​are for illustrative purposes only and do not constitute a limitation on the scope of protection.

[0090] The asymmetric physical disturbance range can be determined based on historical prediction error quantiles, conformal prediction ranges, manually set confidence intervals, or the uncertainty range output by the prediction model. Typical penalty rate scenarios can be determined based on historical market data, ancillary service market assessment rules, extreme scenarios set by the operator, or a combination of the above methods.

[0091] The maximum tolerable robust radius can be determined using binary search, grid search, heuristic search, or other iterative search methods. Representative physical scenarios can be constructed using a finite set of corner points, a set of critical adverse scenarios, a sampled scenario set, or a combination of the above methods.

[0092] In this embodiment, a simulation platform was built in the MATLAB R2021a environment. The optimized model was modeled using the YALMIP toolbox and solved by calling the GUROBI solver. The above exemplary parameters were verified by simulation.

[0093] like Figure 6 As shown, wind power forecast power exhibits a reverse peak-shaving characteristic of being high at night and low during the day, photovoltaic power forecast power exhibits a single-peak characteristic at midday, and load forecast power exhibits a double-peak characteristic in the morning and evening; the asymmetric disturbance range (robust radius) of wind power, photovoltaic power, and load is also shown. Time corresponds to , and The significant differences in the amplitude of the disturbance indicate that the prediction error levels of different source load variables vary over time and are asymmetrical in direction, verifying the necessity of setting the asymmetric physical disturbance interval in S2.

[0094] like Figure 7 As shown, among the 15 typical penalty rate scenarios generated by S3, the penalty rate levels of 12 regular scenarios are close to the benchmark value, while the penalty rates of the severe default segments of the 3 retained high penalty rate tail scenarios reach 3 to 8 times the benchmark value. The tail scenarios have a low probability of occurrence but are explicitly retained in the scenario set, avoiding the meanization caused by the reduction of ordinary scenarios, and providing a scenario basis for measuring economic tail risk items.

[0095] like Figure 8 As shown, when the commitment coefficient At that time, the most unfavorable risk adjustment cost With robust radius Monotonically increasing, with a uniform budget tolerance ceiling The intersection point is the maximum robust radius that can withstand. This intuitively demonstrates the solution mechanism of the upper boundary recognition model (18)(19) and the convergence objective of the binary search.

[0096] like Figure 9 As shown, for The commitment boundary obtained by scanning with a step size of 0.1 within the range of 0 to 1.2 decreases monotonically: when When the value increases from 0 to 1.2, The decrease from 2.50 to 0.40 quantitatively reflects the compression effect of increased reporting intensity on the system's robustness margin. A preset safety margin is used. The promised boundary intersects with the horizontal baseline at Therefore, the upper limit of the declared power is The area above the boundary curve is the zone where disturbances are unacceptable, and the area below it is the zone where commitments can be fulfilled. Specifically: The intermediate working point of the scan range is used to determine the reading result, which is related to the values ​​at both ends of the scan range. Maximum time margin The time margin is minimized, forming a complete control. The simulation data for the commitment boundary are shown in Table 1:

[0097] In Table 1, Follow Monotonically decreasing, verifying the compression effect of declaration intensity on robustness margin; and Comparison .

[0098] like Figure 10 As shown, with the increase of the commitment coefficient, the deviation assessment component and the economic tail risk item in the expected operating cost increase. Both soft constraint penalties and other terms increased significantly: when hour, Comparison The increase was about 10 times, and there was a clear soft constraint penalty, indicating that the pressure of physical realization of the system and the tail risk of the economy were simultaneously aggravated under the high intensity of declaration, which verifies the necessity of jointly considering the three types of cost items in S6.

[0099] like Figure 11 As shown, S7 declaration control is implemented: candidate commitment coefficient. correspond The assessment determined that there was a high risk, and the declared power was adjusted along the committed boundary. Corresponding to the level, the declared power for each time period is from Down to The adjusted declared power all fell within the commitment fulfillment zone, verifying the effectiveness of the declaration control closed loop.

[0100] Simulation results show that the method of the present invention can quantify the upper limit of the declaration and the risk margin, and coordinate the handling of physical realization risk and economic tail risk under a unified framework. The technical effect achieved is consistent with the beneficial effects described in the specification.

Claims

1. A method for submitting applications for virtual power plant standby ancillary services, taking into account physical fulfillment risk and economic tail risk, characterized in that... Includes the following steps: S1: Obtain source and load forecast data, aggregated resource operation parameters, standby ancillary service application parameters, and deviation assessment parameters of the virtual power plant; S2: Construct an asymmetric physical perturbation interval that expands with the robust radius based on the source load prediction data, forming a set of physical uncertainties; S3: Generate a set of typical penalty rate scenarios that retain high penalty rate tail scenarios based on deviation assessment parameters; S4: Introduce a commitment coefficient, determine the day-ahead declared power based on the commitment coefficient, and establish a cost assessment model for the deviation between the day-ahead declared power and the real-time common connection point exchange power. S5: Construct a representative physical scenario that characterizes the key unfavorable realizations in the set of physical uncertainties, wherein the representative physical scenario includes a finite set of corner points; S6: Construct an upper-layer boundary identification model and a lower-layer risk scheduling model. Based on the representative physical scenario and the typical penalty rate scenario set, determine the maximum tolerable robust radius corresponding to different commitment coefficients under the unified budget tolerance upper limit constraint to form the commitment boundary. S7: Perform declaration control based on the committed boundary and output the day-ahead declared power or aggregated resource scheduling strategy.

2. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 1, is characterized in that: In S1, The source-load forecast data includes wind power forecast, photovoltaic power forecast and load forecast for each scheduling period, and the time scale of the scheduling period includes 15 minutes, 30 minutes or 1 hour; The aggregated resource operating parameters include the upper and lower limits of output, ramping constraints and operating costs of controllable power generation resources, the charging and discharging power limits, upper and lower limits of energy state, charging and discharging efficiency, initial state and target terminal state of energy storage resources, as well as the power boundary and power purchase and sale state variables of the point of common coupling. The standby ancillary service application parameters include the baseline commitment capacity, commitment coefficient, capacity compensation price, and application period; the deviation assessment parameters include the deviation exemption threshold, the serious default threshold, the segmented penalty rate, and the assessment rules.

3. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 2, is characterized in that: S2 includes: S2.1: For any source load variable in wind power, solar power, and load. Determine its point prediction value in each scheduling period. Lower deviation amplitude and upper deviation amplitude Construct the original prediction interval: ； in, Index for scheduling period; lower deviation magnitude and upper deviation amplitude The parameters are determined based on historical prediction error quantiles, conformal prediction intervals, operational experience thresholds, or prediction model outputs, and are set separately for different scheduling periods, different source load variables, and different deviation directions. S2.2: Introducing a robust radius , The original prediction interval is linearly expanded to obtain the lower bound of the expanded interval. and the upper bound of the interval : ； ； S2.3: Integrate the expanded intervals of the wind power, photovoltaic, and load source variables to obtain the physical uncertainty set. : ； in, , , Time periods Wind power, photovoltaic power and load power; , , These are the point prediction values ​​for the corresponding source load variables; , , , , , These represent the lower and upper deviation magnitudes of the corresponding source load variables, respectively. This is the set of scheduling time periods.

4. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 3, is characterized in that: S3 includes: S3.1: Based on historical data of deviation penalty rates, ancillary service market assessment rules, or preset extreme penalty rate scenarios, generate initial penalty rate scenarios using probability density estimation and sampling methods; details are as follows: First, based on historical penalty rate samples for each scheduling period, kernel density estimation is used to obtain the corresponding cumulative distribution function. Then, stratified sampling is used to generate the initial scenario. During the period The penalty fee rate is: ； in, For the initial scene During the period Penalty rates; For time period The inverse function of the cumulative distribution function of penalty rates; This represents the initial number of scenes; The random number is within the interval (0,1); stratified samples from different time periods are randomly paired to form the initial penalty rate scenario set. ,in, This is the initial set of scenes; S3.2: Calculate the tail strength index for each initial penalty rate scenario, as follows: The tail strength of the scenario is calculated using a weighted sum of penalty rates for each time period: ； in, For the initial penalty rate scenario Tail strength index; For time period The weights; The larger the tail strength index, the more significant the amplification effect of the scenario on the cost of deviation assessment. S3.3: Based on the tail strength index, retain several high-penalty-rate tail scenarios, and perform scenario reduction or clustering on the remaining scenarios to obtain representative scenarios; as follows: The initial scene is arranged according to Sort by size from largest to smallest, and keep the first few. One scenario is designated as a high-penalty-rate tail scenario; the remaining scenarios are clustered or reduced according to the vector composed of penalty rates for each time period, and the scenario closest to the center of each class is selected as a representative scenario; the probability quality corresponding to the deleted scenario is incorporated into the corresponding representative scenario; in, The number of high-penalty-rate tail scenes to be directly retained can be determined based on the preset tail retention ratio or the upper limit of the number of scenes. S3.4: Merge the high-penalty-rate tail scenarios with the representative scenarios, and normalize the scenario probabilities to obtain the typical penalty-rate scenario set; specifically as follows: The high-penalty-rate tail scenario set and the representative scenario set are combined into a typical penalty fee rate scenario set. The probability quality before normalization is determined according to the number of initial scenarios represented by each retained scenario. The normalized probability is: ； in, This is a typical set of penalty fee scenarios; For the scene Probability mass before normalization; For the scene The normalized probability satisfies and .

5. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 4, is characterized in that: S4 includes: S4.1: The optimal day-ahead power trajectory is obtained under the nominal scenario and used as a baseline commitment capability. The nominal scenario is one where the robust radius is zero and the penalty rate is the predicted value. S4.2: Based on the commitment coefficient Determine the power declared before the date : ； Wherein, the commitment coefficient A dimensionless parameter characterizing the intensity of the application; Abbreviated as ; S4.3: Power switching based on real-time common connection point Define commitment deviation and its absolute value : ； ； S4.4: Introduce piecewise deviation variable , , absolute value Perform linear decomposition and obtain the time period by weighted summation using segmented penalty rates. Deviation assessment cost : ； ； ； in, The threshold for exemption from liability for deviation. The threshold for severe default must be met. ; , , To achieve segmented penalty rates, satisfying .

6. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 5, is characterized in that: S5 includes: S5.1: Constructing the basic corner set The basic corner point set This includes at least one of the following scenarios: nominal scenario, maximum net load scenario, minimum net load scenario, and single-source adverse scenario; wherein: The nominal scenario corresponds to point-based predicted values ​​for wind power, photovoltaic power, and load. The maximum net load scenario corresponds to the lower bound of the interval for wind power and photovoltaic, and the upper bound of the interval for load. The minimum net load scenario corresponds to the upper limit of the interval for wind power and photovoltaic, and the lower limit of the interval for load. The single-source adverse scenario corresponds to only wind power down-biasing, photovoltaic down-biasing, or load up-biasing; S5.2: Calculate the committed pressure index for each scheduling period. : ； in, For time period Net load power; For time period The energy available for energy storage; A preset positive number used to avoid a denominator of zero; It is a function with the deviation between the reported power and the net load power, the net load ramp-up rate, and the available energy storage as independent variables; S5.3: Selecting Commitment Pressure Indicators Time windows that continuously exceed the threshold During the time window The internal command takes the lower bound of the interval for wind power and photovoltaic power and the upper bound of the interval for load, generating local extreme disturbance scenarios and forming a commitment pressure-oriented enhanced corner point set. : ； in, A heuristic operator for encapsulating the process of time window filtering and interval boundary assignment; , , Time periods Forecast values ​​for wind power, solar power, and load points; This refers to the combined lower and upper deviations of each source load variable. S5.4: Merge the basic corner point sets With the aforementioned commitment pressure-guided enhanced corner point set , thus obtaining a finite set of corner points : 。 7. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 6, is characterized in that: In S6, the construction and solution of the lower-level risk scheduling model includes: S6.1: For a given commitment coefficient Robust radius and representative physical scenarios In typical penalty fee scenario sets Define risk adjustment costs The weighted sum of expected operating cost, economic tail risk term, and soft constraint penalty term: ； in, The expected operating costs include internal resource operation and maintenance costs, main grid electricity purchase and sales costs, and deviation assessment costs. This is the risk trade-off coefficient; Confidence level The conditional value at risk is used to measure the economic tail risk item. , Time periods The unit penalty price and reduction amount for flexible load shedding; For the set of scheduling periods; S6.2: Introducing Auxiliary Variables and super loss slack variables Conditional risk value The linear expression is: ； ； in, Confidence level Value at risk below; These are slack variables for the superloss; Index for penalty fee scenarios; For the scene Losses related to deviation assessments; For the scene The normalized probability; This is a typical set of penalty fee scenarios; S6.3: With the objective of minimizing risk adjustment costs, a representative physical scenario is obtained by solving the following constraints: boundary constraints on power purchase and sale at the point of common coupling, mutual exclusion constraints on power purchase and sale, upper and lower limits and ramping constraints on controllable generation resource output, power constraints on energy storage charging and discharging, recursive constraints on energy storage state of mind, upper and lower limits constraints on energy storage state of mind, power balance constraints, deviation definition constraints, and step penalty linearization constraints. The optimal risk-adjusted cost is as follows: For a given commitment coefficient Robust radius and representative physical scenarios The lower-level decision variable set consists of internal distributed resource output, energy storage charging and discharging power and energy state, power purchased and sold at the point of common coupling, deviation segmentation variables, and soft constraint relaxation variables. Please provide a solution: in, Representative physical scenarios The set of lower-level decision variables; Given decision variables, commitment coefficients, and robust radius, the risk adjustment cost in this physical scenario; The optimal risk adjustment cost is obtained to satisfy the operational constraints listed in this step; the obtained optimal risk adjustment cost is used as the input of the upper boundary identification model in S6.

4.

8. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 7, is characterized in that: In step S6, the construction and solution of the upper boundary recognition model includes: S6.4: Define the worst-case risk adjustment cost The maximum value of the optimal risk adjustment cost under each representative physical scenario: ； in, A collection of representative physical scenarios; S6.5: Adjust costs based on reference risks With budget tolerance coefficient The unified budget tolerance limit is determined to be Among them, risk-adjusted costs are referenced. The nominal scenario risk adjustment cost is taken from the preset commitment coefficient, or it can be set by the user according to risk preference or market rules; S6.6: With the objective of maximizing the robust radius and the constraint that the adjustment cost of the worst-case risk does not exceed the unified budget tolerance limit, construct and solve the upper-level optimization model to obtain the commitment coefficient. The corresponding maximum robust radius : ； 。 9. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 8, is characterized in that: S6 further includes: S6.7: For a candidate robust radius, if the corresponding worst-case risk adjustment cost does not exceed the unified budget tolerance limit, then the virtual power plant is determined to be able to withstand source load disturbances under the candidate robust radius; otherwise, it is determined to exceed the tolerable boundary; specifically as follows: Define a budget discriminant function for candidate robust radii, and update the search interval using binary search as an example: ； in, For budget discriminant functions; and The first The lower and upper bounds of the search in each iteration; This is the candidate robust radius; and Following the definition in S6.5; when At that time, update the lower bound of the search to Otherwise, update the upper search bound to... When the difference between the upper and lower bounds is not greater than the preset convergence accuracy. The iteration stops when the time is right, where, To determine the convergence accuracy of the robust radius search, output the last acceptable candidate value as... ; Based on this, an iterative search method is used to identify the maximum tolerable robust radius; S6.8: Perform the identification process of S6.7 on multiple commitment coefficients respectively to obtain the mapping relationship between the commitment coefficients and the maximum tolerable robust radius, forming the commitment boundary, and determine the upper limit of declaration and risk margin that meet the risk tolerance constraint based on the commitment boundary; as follows: S6.7 is applied to each commitment coefficient in the candidate commitment coefficient set to form discrete commitment boundaries, and the acceptable declaration range is determined accordingly: ； in, For the set of candidate commitment coefficients; The discrete commitment boundary is formed by the commitment coefficient and the maximum tolerable robust radius; To preset a safety margin; The set of acceptable commitment coefficients to meet safety margin requirements; The maximum commitment factor to meet the safety margin requirements; Candidate commitment coefficient The corresponding risk margin; when If the value is empty, it is determined that there is no claim strength that meets the safety margin under the current resource configuration; if it is not empty, it is determined that... The corresponding daily reported power is used as the upper limit for reporting.

10. The method for submitting a virtual power plant backup ancillary service application, taking into account physical fulfillment risk and economic tail risk, as described in claim 9, is characterized in that: In step S7, the maximum tolerable robust radius corresponding to the candidate commitment coefficient is compared with the preset safety margin; specifically as follows: First, based on the obtained set of acceptable commitment coefficients... ,when When it is determined that there is no application intensity that meets the preset safety margin requirements under the current resource configuration, the aggregated resource reconfiguration or application suspension is triggered; when Then, the maximum tolerable robust radius corresponding to the candidate commitment coefficient is compared with the preset safety margin, and the declaration control is implemented in the following manner: When the maximum tolerable robust radius is lower than the preset safety margin, it is determined that the corresponding declaration intensity has a high risk. Based on the commitment boundary, the day-ahead declared power is reduced, the flexible resource reserved capacity is increased, or the energy storage operation strategy is adjusted. The adjusted day-ahead declared power or aggregated resource scheduling strategy is sent to the virtual power plant energy management system, the dispatch control system, or the ancillary service market declaration system. When the maximum tolerable robust radius is not lower than the preset safety margin, the day-ahead declared power corresponding to the candidate commitment coefficient is determined as the declared power within the acceptable declaration range, or the declared power with the better comprehensive evaluation result of benefit, risk and robust margin is selected from multiple candidate commitment coefficients.

Images