An IGDT double-layer model-based wind-solar power station energy storage capacity optimization method and system

Abstract

The application discloses a wind and light power station energy storage capacity optimization method and system based on an IGDT double-layer model, relates to the technical field of new energy station energy storage planning optimization, and constructs an IGDT double-layer model with upper-layer annual net income maximization and lower-layer daily dispatching income maximization, generates wind and light typical scenes in combination with Copula, constructs energy and frequency modulation price uncertainty sets relying on the IGDT, and sets income bottom constraint; a complete set of engineering constraints such as power balance, SOC evolution and frequency modulation reachable domain are completely established, and full-cycle energy storage investment, operation and maintenance loss and grid-connected assessment income and expenditure are quantified synchronously; the system adopts a four-layer modular layered architecture, and each module is standardized and interconnected. The application solves the defects of traditional technology, such as planning and operation fragmentation, neglect of wind and light correlation, no income guarantee for price fluctuation and model constraint departure from engineering practice, can output robust energy storage capacity, power configuration and daily dispatching curve, improves economic stability of a scheme, and is suitable for wind and light station energy storage planning decision.

Description

Technical Field

[0001] This invention relates to the field of energy storage planning and optimization technology for new energy power plants, specifically to a method and system for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model. Background Technology

[0002] With the continuous and rapid expansion of domestic wind power and photovoltaic new energy installed capacity, wind and solar power naturally exhibit strong randomness and volatility. Furthermore, the nonlinear coupling relationship between wind and solar power output directly leads to difficulties in new energy consumption and insufficient grid connection and frequency regulation capabilities. Currently, the price volatility in the electricity spot market and frequency regulation ancillary service market continues to increase, and the assessment mechanism for grid-connected power generation prediction deviations is becoming increasingly stringent. Energy storage, as a core device for smoothing volatility, participating in market arbitrage, and providing frequency regulation support, directly determines the full-cycle economic benefits of power plants through its capacity and power configuration. Existing energy storage planning technologies generally separate energy storage capacity planning from intraday operation scheduling into independent models. Upper-level capacity configuration only uses simplified daily operation revenue calculations, failing to establish a two-layer optimization architecture with closed-loop linkage between the planning and operation layers. This fails to reflect the constraint effect of energy storage scale on intraday charging and discharging and frequency regulation strategies, resulting in significant deviations between actual and calculated benefits after the configuration scheme is implemented. Simultaneously, most existing schemes ignore the correlation between wind and solar power output, using only independent random scenario modeling, leading to distorted scenario samples and difficulty in reproducing the true fluctuation characteristics of wind and solar power output. Traditional robust optimization methods... The existing model requires precise probability distribution functions for electricity prices and wind and solar power output, which are difficult to obtain in engineering projects. Furthermore, it does not simultaneously construct uncertainty constraints for both electricity and frequency regulation market prices, making it impossible to set a unified revenue guarantee mechanism under multiple price fluctuations. When market electricity prices decline, energy storage investment faces significant loss risks. In addition, the existing model fails to comprehensively and uniformly characterize the three types of revenue: electricity sales, frequency regulation capacity, and frequency regulation mileage. It also does not simultaneously incorporate the full life cycle loss and maintenance costs of energy storage and the assessment costs of grid connection deviation. The constraint system lacks engineering constraints such as the three-component allocation of wind and solar power, multi-condition power decomposition of energy storage, daily SOC cycle, and charge-discharge mutual exclusion. The optimized dispatch scheme has many physically unrealizable operating conditions. The overall scheme has poor economic robustness and engineering feasibility, and cannot meet the current practical decision-making needs of refined planning and market risk resistance for wind and solar power plant energy storage. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model, thereby solving the defects of traditional technologies such as fragmented planning and operation, neglect of wind and solar correlation, lack of revenue guarantee due to electricity price fluctuations, and model constraints that are out of touch with engineering realities.

[0004] To achieve the above objectives, the embodiments of this invention provide the following technical solutions:

[0005] This application provides a method for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model, including the following steps: S1, constructing a two-layer model for optimizing the allocation of wind and solar power plant energy storage capacity, the two-layer model including an upper-layer model and a lower-layer model; the upper-layer model uses energy storage capacity and energy storage power as decision variables, and maximizes annual net income as the objective function; the lower-layer model, given the upper-layer decision variables and scenario and price set, optimizes the intraday operation and scheduling income of wind, solar and energy storage in the power grid market and frequency regulation market; S2, establishing model constraints based on the real-time or predicted operating status of wind and solar power plants and the configured energy storage, the constraints including power balance constraints, wind and solar power allocation and curtailment conservation constraints, energy storage charging and discharging power decomposition and rated power constraints, frequency regulation capacity reachability constraints, energy storage charge state evolution and upper and lower bound constraints, daily cycle consistency constraints, charging and discharging mutual exclusion constraints, and deviation assessment constraints; S3, constructing a robust solution method for the two-layer model based on IGDT, including: using kernel density estimation method to estimate edge density based on historical wind and solar data and using Copula... The function constructs a nonlinear correlation model of wind and solar power output, generates a set of typical scenarios and their weights; determines the electricity price benchmark sequence and sets the risk tolerance, and establishes the electricity price uncertainty set for two types of markets: electricity energy and frequency regulation; in each scenario, the most unfavorable electricity price is taken and a minimum return constraint is applied, and joint optimization is performed with the annual robust net return as the objective; S4, input the historical operating data and equipment parameters of wind power output, photovoltaic power output, load and electricity price, configure the key parameters of IGDT and the weights of typical scenarios, solve the two-layer model, and obtain the optimized configuration scheme of energy storage capacity and energy storage power and the corresponding daily operation plan curve.

[0006] Furthermore, the objective function of the upper-level model in S1 is to maximize the annual net return, expressed by the formula: ;in, Let the annual net income objective function be... A collection of indexes for typical scenarios; For the scene The annual weight is the proportion of times it appears; For the scene The optimal daily profit after the most unfavorable price is obtained by solving the lower-level model. The annualized cost of energy storage investment is calculated using a capital recovery factor, satisfying the following: ;in, The discount rate; For energy storage lifetime, the smaller value between cycle lifetime and float charge lifetime is taken, i.e.: ;in, For cycle life, For float charge life; This represents the initial total investment in energy storage; the initial total investment in energy storage consists of two parts: capacity investment and power investment, satisfying the following: ;in, The cost per unit capacity is expressed in ¥ / MWh. The cost per unit power is expressed in ¥ / MW. The energy storage capacity is expressed in MWh, and the variable is the decision variable of the upper-level model. , where is the energy storage capacity in MW, and is the decision variable of the upper-level model.

[0007] Furthermore, the lower-level model optimizes the intraday operation scheduling benefits, and its objective function is expressed as:

[0008] ;in, For intraday time step index, This represents the total number of time periods within a day. For the scene Down The price of electricity at any given time, expressed in ¥ / MWh; For the scene Down Power consumption at any given moment, measured in MW; For the scene Down The price of frequency regulation capacity at any given time, in ¥ / (MW·h); For the scene Down Frequency regulation mileage or performance price at any given time, in ¥ / (MW·h); This refers to the mileage intensity coefficient; For the scene Down Frequency modulation capacity at any given time, in MW; The equivalent turnover cost for degradation or maintenance is expressed in ¥ / MWh. For the scene Down Energy storage charging power at any given time, in MW; For the scene Down The energy storage and discharge power at any given moment, expressed in MW; The time step is expressed in hours (h). For the scene Down The cost of predictive deviation at any given time is expressed in ¥; the formula for calculating the electricity consumption online is as follows: ;in, The output of wind power sent to the electricity market, measured in MW; The output of photovoltaic power into the electricity market, measured in MW; The discharge power of energy storage in the energy market, measured in MW; The charging power of energy storage in the energy market, measured in MW.

[0009] Furthermore, in S2, the constraint conditions include: the formula for the wind and solar power allocation and curtailment conservation constraint is expressed as: ;in, For wind power in Available power at any given time, in MW; For wind power in The power used to charge energy storage at any given time, measured in MW; For wind power in Wind curtailment power at any given moment, in MW; For photovoltaics Available power at any given time, in MW; For photovoltaics The power used to charge energy storage at any given time, measured in MW; For photovoltaics The power of abandoned light at any given time is expressed in MW; the formula for the decomposition of energy storage charging and discharging power and the rated power constraint is as follows: ;in, This represents the portion of energy storage discharge power used in the energy market, measured in MW. This is the component of energy storage discharge power used for frequency modulation, measured in MW; The portion of energy storage charging power used in the energy market, measured in MW; The component used for frequency regulation in the energy storage charging power, measured in MW; This represents the upper limit of the rated power of energy storage, measured in MW.

[0010] Furthermore, the constraints also include: the formula for the frequency modulation capacity reachability domain constraint is expressed as: ;in, The available capacity for frequency regulation is expressed in MW; the formula for the energy storage state of charge (SOC) evolution constraint is as follows: ;in, Energy storage period In the scene The energy state under the following conditions is expressed in MWh. For charging efficiency; For discharge efficiency; the formulas for the upper and lower bound constraints of SOC and the daily cycle consistency constraint are expressed as follows: ;in, This represents the lower limit of SOC. This represents the upper limit ratio of SOC. Energy storage capacity, in MWh; The initial energy state of the sun. This represents the end-of-day energy state; the charge-discharge mutual exclusion constraint adopts the Big-M method, expressed by the following formula: ;in, It is a sufficiently large constant; For charging and discharging, there are two mutually exclusive variables. Indicates discharge. This indicates charging; the formula for the deviation assessment cost is expressed as: ;in, To assess penalty rates or deviation electricity price coefficients; For submitting or forecasting electricity generation plans; This represents the actual amount of electricity used for internet access.

[0011] Furthermore, the degradation or maintenance equivalent turnover cost The formula for calculating it is expressed as follows: ;in, The annualized total cost satisfies: ;in, Annualized operation and maintenance costs; This represents the operational coefficient for capacity. The maintenance factor is based on power; the total available energy turnover over the lifespan. The formula is expressed as: ;in, Equivalent cycle life, measured in cycles.

[0012] Further, in S3, the robust solution method of IGDT includes the following steps: S3.1: Obtain historical data of wind power and photovoltaic power, perform edge density estimation using kernel density estimation method and construct the correlation between wind and solar power output using Copula; sample on the joint distribution of wind and solar power, and obtain several typical scenario sets S and their weights through scenario reduction; S3.2: Determine the benchmark sequence of electricity price and frequency regulation price, set the risk tolerance and price deviation allocation weights, and bind the scenarios and weights obtained in S3.1; S3.3: Establish the price uncertainty set of the two types of markets, electricity and frequency regulation, and use the comprehensive deviation index to simultaneously describe the relative fluctuation of electricity price and frequency regulation price. Forming interval constraints; S3.4: For each scenario, under the lower-level daily operation constraints of S2, solve for the benchmark daily revenue using the benchmark electricity price; S3.5: Within the price uncertainty set, take the most unfavorable case for the electricity price and solve for the most unfavorable daily revenue under the lower-level daily operation constraints of S2; S3.6: Apply a revenue guarantee constraint to all scenarios. If the most unfavorable daily revenue does not meet the guarantee constraint, adjust the risk tolerance or upper-level decision variables until they are met; S3.7: Under the premise of satisfying S3.6, perform joint optimization with the annual robust net revenue as the objective. When the optimal value is reached or iterative convergence is achieved, output the optimal energy storage capacity and power configuration and the corresponding daily operation plan curve.

[0013] Furthermore, in S3.1, the kernel density estimation method (KDE) establishes marginal probability densities for wind power and photovoltaic power respectively, expressed by the following formula: ;in, This is an estimate of the edge probability density of wind power. This is an estimate of the photovoltaic edge probability density. The number of samples; For wind power nuclear power bandwidth; For photovoltaic core bandwidth; For kernel functions; For wind power sample points; These are photovoltaic sample points; the correlation between wind and solar output is constructed using Copula, employing the Frank-Copula cumulative distribution function, expressed as follows: ;in, The value of the Frank-Copula cumulative distribution function; For Copula-related parameters; This represents the cumulative distribution function of the edge distribution of wind power. Let be the cumulative distribution function of the photovoltaic edge distribution; After wind power edge distribution mapping variable; After mapping by photovoltaic edge distribution Variables; The scene reduction adopts a clustering method, and the formula for calculating the optimization objective and scene weights is expressed as follows: ;in, The i-th sampled wind and solar power output curve is the spliced ​​vector. This is the center of the kth typical scene; Number of typical scenarios; This represents the original number of samples; For the scene The weights; Let be the number of samples in the s-th cluster.

[0014] Furthermore, in S3.3, the establishment of price uncertainty sets for the two markets of electricity energy and frequency regulation establishes a mapping relationship between the comprehensive deviation and the sub-market deviation for the deviation between electricity energy price and frequency regulation price: ;in, The comprehensive deviation index serves as the decision variable for IGDT. This represents the relative deviation of electricity prices. This represents the relative deviation of the FM price. and These are the weights for the electricity price deviation and the frequency regulation price deviation, respectively; the formula for constructing the hourly price uncertainty set interval constraint is expressed as follows: ;in, This serves as a benchmark for predicting electricity prices. The benchmark price for FM mileage; The benchmark price for frequency regulation capacity; in S3.6, the formula for the guaranteed revenue constraint is expressed as: ;in, For the scene The optimal daily return after the most unfavorable price; Daily operating revenue under the benchmark electricity price; As for risk tolerance, A smaller value indicates greater conservatism; in S3.7, the formula for the joint optimization representation is expressed as: Under the conditions of satisfying the guaranteed return constraint and all constraints of S2, the optimal energy storage capacity is obtained by solving the problem. and optimal energy storage power And the corresponding daily operation plan curves for each scenario.

[0015] Accordingly, this application also provides a wind and solar power plant energy storage capacity optimization system based on an IGDT two-layer model. The system is structured in a layered architecture consisting of a data and configuration layer, an uncertainty layer, an optimization core layer, and a solution and output layer. It includes: a data and parameter management module, an uncertainty modeling module, a two-layer optimization model construction module, and a robust solution and plan output module. Specifically, the data and configuration layer includes the data and parameter management module, which is data-connected to both the uncertainty modeling module and the two-layer optimization model construction module. This module is used to uniformly access and store wind power output data, solar power output data, load data, electricity price data, and equipment parameters; manage scenario weights and IGDT parameters; and provide standard data interfaces and version traceability to each layer. The uncertainty layer includes the uncertainty modeling module, which comprises a typical scenario generation submodule and an IGDT uncertainty set submodule. The typical scenario generation submodule is data-connected to the data and parameter management module. The T-uncertainty set submodule is data-connected to the two-layer optimization model construction module; the typical scenario generation submodule is used to generate typical wind and solar scenarios and weights based on Copula; the IGDT uncertainty set submodule is used to map the benchmark electricity price and risk parameters into a price uncertainty set for optimization; the optimization core layer is equipped with the two-layer optimization model construction module, which includes a constraint submodule. The input of the two-layer optimization model construction module is connected to the data and parameter management module and the uncertainty modeling module, respectively, to establish an integrated model of upper-layer energy storage capacity and power configuration and lower-layer intraday scheduling, and to centrally register power balance constraints, SOC constraints, and frequency regulation capacity reachability constraints; the solution and output layer is equipped with the robust solution and plan output module, whose input is connected to the two-layer optimization model construction module, to perform robust verification and joint optimization under scenario and price range, output the optimal energy storage capacity and power configuration scheme and intraday operation plan curve, and archive them.

[0016] The beneficial effects of this invention are as follows: By constructing an IGDT double-layer nested optimization model of upper-layer capacity and power configuration and lower-layer intraday operation scheduling, and combining Copula wind and solar correlation scenario generation and robust minimum guarantee constraint for price uncertainty, an integrated closed-loop optimization of energy storage planning and operation is achieved. This solves the problems of separation of planning and operation, neglect of wind and solar output correlation, and lack of minimum guarantee of revenue under electricity price fluctuations in the existing technology. While ensuring technical feasibility, it improves the economic robustness of the configuration scheme and makes the planning results more in line with the actual operation scenario of the power market. Attached Figure Description

[0017] Figure 1A flowchart illustrating a method for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model, provided for an embodiment of this application;

[0018] Figure 2 A flowchart illustrating the robust solution method of IGDT in a wind and solar power plant energy storage capacity optimization method based on an IGDT two-layer model provided in this application embodiment;

[0019] Figure 3 A typical wind power output scenario diagram is shown in the copula method of the wind and solar power plant energy storage capacity optimization method based on the IGDT two-layer model provided in the embodiments of this application.

[0020] Figure 4 A typical photovoltaic power output scenario diagram is shown in the copula method of the wind and solar power plant energy storage capacity optimization method based on the IGDT two-layer model provided in the embodiments of this application.

[0021] Figure 5 This is a schematic diagram of a wind and solar power plant energy storage capacity optimization system based on an IGDT two-layer model, provided as an embodiment of this application. Detailed Implementation

[0022] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0023] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0024] Example 1:

[0025] like Figures 1-4As shown in the embodiment of this application, a method for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model is provided, including the following steps: S1, constructing a two-layer model for optimizing the allocation of wind and solar power plant energy storage capacity, the two-layer model including an upper-layer model and a lower-layer model; the upper-layer model uses energy storage capacity and energy storage power as decision variables, and maximizes annual net income as the objective function; the lower-layer model, given the upper-layer decision variables and scenario and price set, optimizes the intraday operation and scheduling income of wind, solar and energy storage in the power grid market and frequency regulation market; S2, based on the real-time or predicted operating status of wind and solar power plants and the configured energy storage, establishing model constraints, the constraints including power balance constraints, wind and solar power allocation and curtailment conservation constraints, energy storage charging and discharging power decomposition and rated power constraints, frequency regulation capacity reachability constraints, energy storage charge state evolution and upper and lower bound constraints, daily cycle consistency constraints, charging and discharging mutual exclusion constraints, and deviation assessment constraints; S3, constructing a robust solution method for the two-layer model based on IGDT, including: using kernel density estimation method to estimate edge density based on historical wind and solar data and using Copula... The function constructs a nonlinear correlation model of wind and solar power output, generates a set of typical scenarios and their weights; determines the electricity price benchmark sequence and sets the risk tolerance, and establishes the electricity price uncertainty set for two types of markets: electricity energy and frequency regulation; in each scenario, the most unfavorable electricity price is taken and a minimum return constraint is applied, and joint optimization is performed with the annual robust net return as the objective; S4, input the historical operating data and equipment parameters of wind power output, photovoltaic power output, load and electricity price, configure the key parameters of IGDT and the weights of typical scenarios, solve the two-layer model, and obtain the optimized configuration scheme of energy storage capacity and energy storage power and the corresponding daily operation plan curve.

[0026] In one possible embodiment, a two-layer model for optimizing the allocation of wind-solar-storage power plant capacity is first constructed. The two-layer model is a nested optimization architecture formed by an upper-layer model for annual-scale planning and a lower-layer model for daily-scale operation. The upper-layer model uses the rated capacity and rated power of energy storage as core decision variables and aims to maximize the annual net revenue weighted across all scenarios, thus determining the scale of energy storage construction. The lower-layer model is an intraday operation scheduling model. Given the capacity and power parameters output by the upper layer, as well as typical wind-solar scenarios and price fluctuation ranges, it optimizes the hourly operation strategy of the wind-solar-storage integrated system in the electricity market and frequency regulation ancillary service market to maximize daily operating revenue. The optimal operating revenue of the lower layer is fed back to the upper layer to verify the economics of capacity allocation, forming a closed-loop optimization of planning and operation. Secondly, based on real-time or short-term predicted operating data of wind and solar power plants and their configured energy storage, a complete model constraint system is established. This system specifically covers power balance constraints, wind and solar power allocation and curtailment conservation constraints, energy storage charging and discharging power decomposition and rated power constraints, frequency regulation capacity reachability constraints, energy storage charge state evolution and upper and lower bound constraints, daily cycle consistency constraints, charging and discharging mutual exclusion constraints, and deviation assessment constraints. The feasible domain for optimization is limited from three dimensions: the physical laws of the power system, the operating specifications of energy storage equipment, and the settlement rules of the electricity market, ensuring that the output results are engineering-executable. Then, a robust solution method for the two-layer model is constructed based on the information gap decision theory. Specifically, based on the historical output data of wind power and photovoltaic power, the kernel density estimation method is used to fit the output probability density of the two respectively. Then, the Copula function is used to characterize the nonlinear correlation between wind power and photovoltaic power output. Through sampling and scenario reduction, a representative set of typical scenarios and corresponding scenario weights are obtained. Next, the benchmark time series of electricity and frequency regulation price are determined, an acceptable risk tolerance parameter is set, and a price uncertainty set covering the range of electricity price fluctuations is constructed. Subsequently, under each typical scenario, the electricity price value with the most unfavorable return in the price uncertainty set is selected, and a return guarantee constraint is applied to ensure that the daily return under the most unfavorable situation is not lower than the set threshold. Finally, with the goal of maximizing the annual robust net return of all scenarios, the upper and lower layer models are jointly optimized. Finally, input historical operating data of wind power output time series, photovoltaic power output time series, load time series and electricity price, as well as parameters such as cost, efficiency and lifespan of energy storage equipment, configure the risk tolerance, deviation allocation weight and typical scenario weight of IGDT, solve the two-layer model through nested iteration or single-layer transformation, and finally output the optimal configuration scheme of energy storage capacity and energy storage power, as well as the hourly daily operation plan curves for each typical scenario.

[0027] By constructing a two-layer nested optimization model of upper-layer capacity and power configuration and lower-layer intraday operation scheduling using IGDT, and combining Copula wind-solar correlation scenario generation with robust minimum guarantee constraint for price uncertainty, integrated closed-loop optimization of energy storage planning and operation is achieved. This solves the problems of separation of planning and operation, neglect of wind and solar output correlation, and lack of minimum guarantee of revenue under electricity price fluctuations in existing technologies. While ensuring technical feasibility, it improves the economic robustness of the configuration scheme and makes the planning results more in line with the actual operation scenario of the electricity market.

[0028] In the embodiments of this application, the objective function of the upper-level model in S1 is to maximize the annual net income, expressed by the formula: ;in, Let the annual net income objective function be... A collection of indexes for typical scenarios; For the scene The annual weight is the proportion of times it appears; For the scene The optimal daily profit after the most unfavorable price is obtained by solving the lower-level model. The annualized cost of energy storage investment is calculated using a capital recovery factor, satisfying the following: ;in, The discount rate; For energy storage lifetime, the smaller value between cycle lifetime and float charge lifetime is taken, i.e.: ;in, For cycle life, For float charge life; This represents the initial total investment in energy storage; the initial total investment in energy storage consists of two parts: capacity investment and power investment, satisfying the following: ;in, The cost per unit capacity is expressed in ¥ / MWh. The cost per unit power is expressed in ¥ / MW. The energy storage capacity is expressed in MWh, and the variable is the decision variable of the upper-level model. , where is the energy storage capacity in MW, and is the decision variable of the upper-level model.

[0029] In one possible implementation, the optimization objective of the upper-level model is first defined as maximizing annual net profit, and the objective function expression is:

[0030] ;in, Let the annual net income objective function be... This is a collection of indexes for typical scenarios. For the scene The annual weight, that is, the proportion of days in the whole year corresponding to this typical scenario; For the scene The optimal daily operating revenue when the electricity price is at its most unfavorable value is obtained by solving the lower-level intraday operating model. This represents the annualized cost after discounting the total energy storage investment. Next, the annualized cost of energy storage investment is calculated using the capital recovery factor method, which distributes the initial one-time investment evenly across each year of the energy storage's entire lifespan. The calculation formula is as follows:

[0031] ;in The discount rate is used to reflect the time value of money. The actual service life of the energy storage is determined; then, the rule for determining the value of the energy storage life is established, taking the smaller value between the cycle life and the float charging life, i.e. The service life, converted from cycle life, is obtained by converting the total equivalent number of energy storage cycles to the annual number of cycles. The float life is the service life of the energy storage system in a static standby state. Taking the smaller of the two values ​​avoids overestimating the service life of the energy storage and ensures the conservatism and rationality of the cost calculation. Finally, the initial total investment in energy storage is calculated, which is divided into capacity investment and power investment, using the following formula:

[0032] ;in The unit cost of investment is expressed in yuan / MWh, which corresponds to the cost of energy-related components such as energy storage cells. The unit cost per unit power is expressed in yuan / MW, corresponding to the cost of power-related components such as the energy storage converter PCS; \(E_{ES}\) represents the rated capacity of the energy storage, in MWh. The rated power of the energy storage is expressed in MW, and both are core decision variables in the upper-level model.

[0033] By adopting a capital recovery factor to convert the initial investment in energy storage into an equivalent annualized cost, and by weighting and summing the daily operating revenue under the worst-case price scenario according to the annual weight of the scenario, the time caliber of investment cost and operating revenue is unified, solving the problem that the initial one-time investment and annual revenue cannot be directly compared, and achieving accurate quantification of annual net revenue. At the same time, the method of separately calculating capacity cost and power cost is in line with the cost structure characteristics of energy storage system that "cells determine capacity and converters determine power", making the economic calculation of capacity configuration decision more scientific.

[0034] In the embodiments of this application, the lower-level model optimizes the intraday operation scheduling benefits, and its objective function is expressed as:

[0035] ;in, For intraday time step index, This represents the total number of time periods within a day. For the scene Down The price of electricity at any given time, expressed in ¥ / MWh; For the scene Down Power consumption at any given moment, measured in MW; For the scene Down The price of frequency regulation capacity at any given time, in ¥ / (MW·h); For the scene Down Frequency regulation mileage or performance price at any given time, in ¥ / (MW·h); This refers to the mileage intensity coefficient; For the scene Down Frequency modulation capacity at any given time, in MW; The equivalent turnover cost for degradation or maintenance is expressed in ¥ / MWh. For the scene Down Energy storage charging power at any given time, in MW; For the scene Down The energy storage and discharge power at any given moment, expressed in MW; The time step is expressed in hours (h). For the scene Down The cost of predictive deviation at any given time is expressed in ¥; the formula for calculating the electricity consumption online is as follows: ;in, The output of wind power sent to the electricity market, measured in MW; The output of photovoltaic power into the electricity market, measured in MW; The discharge power of energy storage in the energy market, measured in MW; The charging power of energy storage in the energy market, measured in MW.

[0036] In one possible implementation, the optimization objective of the lower-level model is first defined as maximizing the intraday scheduling revenue. The objective function consists of total revenue minus total cost, and its expression is:

[0037] In the formula For intraday time step index, This represents the total number of time periods within a day, typically 24, corresponding to hourly scheduling. For the scene Down The price of electricity at any given time, expressed in ¥ / MWh; For the scene Down Power consumption at any given moment, measured in MW; For the scene Down The price of frequency regulation capacity at any given time, in ¥ / (MW·h); For the scene Down Frequency regulation mileage or performance price at any given time, in ¥ / (MW·h); This refers to the mileage intensity coefficient; For the scene Down Frequency modulation capacity at any given time, in MW; The equivalent turnover cost for degradation or maintenance is expressed in ¥ / MWh. For the scene Down Energy storage charging power at any given time, in MW; For the scene Down The energy storage and discharge power at any given moment, expressed in MW; The time step is expressed in hours (h). For the scene Down The cost of forecast deviation assessment at any given time is expressed in ¥. Next, the revenue components are broken down: The first part is the electricity market revenue, calculated by multiplying the grid-connected power by the time-of-use electricity price and accumulating hourly, reflecting the arbitrage profits from wind and solar power generation and energy storage charging and discharging in the electricity market; the second part is the frequency regulation ancillary service revenue, composed of capacity revenue and mileage revenue. Capacity revenue is settled based on the declared frequency regulation capacity, and mileage revenue is settled based on the actual regulated mileage. The two combined constitute the complete frequency regulation revenue. Then, the cost components are broken down: The first part is the equivalent turnover cost of energy storage degradation and operation and maintenance, calculated based on the electricity equivalent to the hourly charging and discharging power, reflecting the battery loss and operation and maintenance losses caused by each charging and discharging cycle; the second part is the forecast deviation assessment cost, corresponding to the penalty fees incurred due to grid-connected power generation deviation, reflecting the market rules' constraints on the accuracy of power generation forecasts. Finally, the calculation method for grid-connected electricity is clarified, the formula is:

[0038] ;in The output power of wind power transmitted to the electricity market. The output power of photovoltaic power transmitted to the electricity market. For the discharge power of energy storage in the energy market, The power to charge energy storage in the energy market is the power to feed into the grid. The power to feed into the grid is the sum of wind and solar power output and energy storage discharge, minus energy storage charging. At the same time, the power to feed into the grid is constrained to be non-negative to avoid the result of backfeeding power from the grid, which is inconsistent with the operation logic of the power station.

[0039] By uniformly characterizing electricity sales revenue, frequency regulation capacity revenue, and frequency regulation mileage revenue in the lower-level objective function, and simultaneously incorporating energy storage degradation operation and maintenance costs and grid connection deviation assessment costs, this method fully covers all revenue and cost items of the wind-solar-storage integrated system. This solves the problem of existing technologies only considering revenue from a single market and neglecting ancillary services and assessment costs. It ensures that the calculation of intraday operating revenue fully conforms to the settlement rules of the electricity spot market and the ancillary service market, thereby improving the accuracy of revenue after the actual implementation of the energy storage configuration scheme.

[0040] In the embodiments of this application, in step S2, the constraint conditions include: the formula for the wind and solar power allocation and curtailment conservation constraint is expressed as: ;in, For wind power in Available power at any given time, in MW; For wind power in The power used to charge energy storage at any given time, measured in MW; For wind power in Wind curtailment power at any given moment, in MW; For photovoltaics Available power at any given time, in MW; For photovoltaics The power used to charge energy storage at any given time, measured in MW; For photovoltaics The power of abandoned light at any given time is expressed in MW; the formula for the decomposition of energy storage charging and discharging power and the rated power constraint is as follows: ;in, This represents the portion of energy storage discharge power used in the energy market, measured in MW. This is the component of energy storage discharge power used for frequency modulation, measured in MW; The portion of energy storage charging power used in the energy market, measured in MW; The component used for frequency regulation in the energy storage charging power, measured in MW; This represents the upper limit of the rated power of energy storage, measured in MW.

[0041] In one possible embodiment, firstly, constraints on wind and solar power allocation and curtailment conservation are established, and power balance relationships are established separately for wind power and solar power. The constraints on the wind power side are as follows:

[0042] The constraints on the photovoltaic side are:

[0043] ;in , They are respectively Available output power of wind and solar power at any given time, in MW, given by wind and solar forecasts or measured data; , These represent the power output of wind power and photovoltaic power that is directly used to charge energy storage, respectively, in MW; , They are respectively The curtailment of wind and solar power at any given time is expressed in MW. This set of constraints ensures that all available wind and solar power output in each time period flows to the three destinations, with no additional power surplus or deficit, strictly adhering to the law of conservation of power. Secondly, a power decomposition constraint for energy storage charging and discharging is established, splitting the total energy storage discharge power into an energy market discharge component and a frequency regulation discharge component, namely:

[0044] The total energy storage charging power is broken down into energy market charging components, frequency regulation charging components, wind power DC charging power, and photovoltaic DC charging power, namely:

[0045] In the formula , These are the components of energy storage discharge power used for energy market arbitrage and for frequency regulation, respectively. , These represent the components of energy storage charging power used for energy market arbitrage and frequency regulation; the wind and solar direct power components correspond to the wiring method of directly connecting wind and solar power to the energy storage system at the power station, which conforms to the actual engineering architecture of energy storage distribution at new energy power stations. Then, upper and lower limits of the rated energy storage power are set to satisfy:

[0046] ,in The rated power limit of the energy storage system, in MW, is determined by the total capacity of the energy storage converter. This constraint ensures that the charging and discharging power of the energy storage system will not exceed the rated value of the equipment at any time, thus avoiding equipment overload operation. Finally, all the above constraints are incorporated into the feasible region of the lower-level operation model, so that all power allocation schemes in the optimization process meet the physical conservation and equipment rated value requirements, eliminating infeasible scheduling strategies from the root.

[0047] By establishing a three-component conservation constraint for wind and solar power output, the allocation of wind and solar power output in the three destinations of grid-connected electricity sales, energy storage charging, and curtailment is accurately quantified, realizing explicit modeling of wind and solar curtailment. Through multi-component decomposition constraint of energy storage charging and discharging power, the power allocation of energy storage in the energy market, frequency regulation market, and direct wind and solar charging is clearly separated, solving the problem of ambiguous power boundaries in multiple application scenarios of energy storage. At the same time, the rated power constraint ensures that it does not exceed the hardware capacity of the equipment under all operating conditions, making the model fully conform to the physical law of power conservation in the power system, and the output scheduling scheme has physical feasibility.

[0048] In embodiments of this application, the constraint further includes: the formula for the frequency modulation capacity reachability domain constraint is expressed as: ;in, The available capacity for frequency regulation is expressed in MW; the formula for the energy storage state of charge (SOC) evolution constraint is as follows: ;in, Energy storage period In the scene The energy state under the following conditions is expressed in MWh. For charging efficiency; For discharge efficiency; the formulas for the upper and lower bound constraints of SOC and the daily cycle consistency constraint are expressed as follows: ;in, This represents the lower limit of SOC. This represents the upper limit ratio of SOC. Energy storage capacity, in MWh; The initial energy state of the sun. This represents the end-of-day energy state; the charge-discharge mutual exclusion constraint adopts the Big-M method, expressed by the following formula: ;in, It is a sufficiently large constant; For charging and discharging, there are two mutually exclusive variables. Indicates discharge. This indicates charging; the formula for the deviation assessment cost is expressed as: ;in, To assess penalty rates or deviation electricity price coefficients; For submitting or forecasting electricity generation plans; This represents the actual amount of electricity used for internet access.

[0049] In one possible embodiment, a frequency modulation capacity reachability domain constraint is first established, constraining the frequency modulation capability boundaries under charging and discharging conditions respectively, satisfying:

[0050] ;in The available capacity for frequency regulation is expressed in MW. The physical meaning of this set of constraints is as follows: when the energy storage is in basic charging condition, the sum of the upward frequency regulation (discharge regulation) capacity and the basic charging power cannot exceed the rated power; when the energy storage is in basic discharging condition, the sum of the downward frequency regulation (charging regulation) capacity and the basic discharge power cannot exceed the rated power. This ensures that the declared frequency regulation capacity has a true power regulation capability and prevents the declared capacity from exceeding the adjustable range of the equipment. Secondly, constraints on the state of charge evolution of energy storage are established to describe the dynamic change of the remaining energy storage capacity over time. The formula is:

[0051] ;in For the scene Next period The final energy state of the stored energy, in MWh; For charging efficiency, only the amount of electricity multiplied by the efficiency is actually stored in the battery during the charging process. To achieve discharge efficiency, the battery charge (divided by the efficiency) needs to be consumed during discharge to output the corresponding power. This modeling method accurately reflects the energy loss during the charging and discharging process of energy storage, conforming to the actual operating characteristics of the battery. Then, upper and lower bound constraints on SOC and daily cycle consistency constraints are set to satisfy:

[0052] ;in , These are the lower and upper limits of the allowable SOC for energy storage, corresponding to the safe operating range of the battery. Setting the upper and lower limits is to avoid overcharging and over-discharging and to extend the battery's lifespan. This represents the stored energy state at the beginning of each day. This represents the energy storage status at the end of each day. Daily cycle constraints ensure that the initial and final energy levels are equal, allowing the scheduling plan to be executed on a rolling basis each day, preventing continuous accumulation or depletion of energy and ensuring the sustainability of multi-day operation. Finally, charging and discharging mutual exclusion constraints are established, using the Big-M method to implement binary logic constraints, with the following formula:

[0053] ;in To obtain a sufficiently large positive number, the value should be greater than the rated energy storage capacity; This is a binary variable for mutually exclusive charging and discharging. When the value is 1, it indicates that the energy storage is in a discharging state, at which time the charging constraint is relaxed and the discharging constraint is effective; when the value is 0, it indicates that the energy storage is in a charging state, at which time the discharging constraint is relaxed and the charging constraint is effective. This constraint physically eliminates the impossible condition of the energy storage charging and discharging at the same time, ensuring the engineering rationality of the scheduling scheme.

[0054] By constraining the reachability of frequency regulation capacity, the declared frequency regulation capacity is ensured to have actual power support, avoiding inflated frequency regulation capabilities. By using SOC evolution and upper and lower bound constraints to match the battery's charge and discharge loss characteristics and safe operating range, overcharging and over-discharging damage to the battery is avoided. By using daily cycle consistency constraints, the sustainability of power dispatch over multiple days is ensured. By using charge and discharge mutual exclusion constraints, false operating conditions of simultaneous charging and discharging at the same time are eliminated. These multi-dimensional constraints jointly ensure that the optimization results fully comply with the operating specifications and engineering realities of energy storage equipment, and the output dispatch scheme can be directly issued for execution.

[0055] In the embodiments of this application, the degradation or maintenance equivalent turnover cost The formula for calculating it is expressed as follows: ;in, The annualized total cost satisfies: ;in, Annualized operation and maintenance costs; This represents the operational coefficient for capacity. The maintenance factor is based on power; the total available energy turnover over the lifespan. The formula is expressed as: ;in, Equivalent cycle life, measured in cycles.

[0056] In one possible implementation, the degradation or maintenance equivalent turnover cost is first calculated. The core logic is to evenly distribute all costs throughout the energy storage's entire lifecycle across the total turnover of electricity that can be completed within that lifecycle, thus obtaining the loss cost corresponding to each kilowatt-hour of charging and discharging. The calculation formula is as follows:

[0057] This cost is factored into the operating cost with each charge and discharge cycle, allowing the model to automatically account for battery degradation during scheduling optimization and avoid uncontrolled frequent charging and discharging. Next, the annualized total cost is calculated. It consists of the annualized cost corresponding to the initial investment and the annualized operation and maintenance cost, and satisfies:

[0058] in The initial total investment for energy storage, This represents the annualized operation and maintenance cost; investment and operation and maintenance costs are combined and then allocated to fully cover the entire lifecycle cost of energy storage. The annualized operation and maintenance cost is then calculated separately for capacity and power dimensions, using the following formula:

[0059] ;in This is the annual maintenance coefficient for capacity, expressed in yuan / (MWh·year), corresponding to the maintenance costs of capacity-related components such as battery cells and battery management systems. This represents the annual maintenance factor based on power, expressed in yuan / (MW·year), corresponding to the maintenance costs of power-related components such as inductors and switchgear; separate calculations ensure that maintenance cost estimations are more closely aligned with equipment configuration. Finally, the total usable energy turnover over the equipment's lifespan is calculated. The formula is:

[0060] ;in The equivalent cycle life of energy storage, expressed in cycles; The available SOC range ratio for energy storage is multiplied by the rated capacity to obtain the available power in a single cycle; coefficient 2 corresponds to a complete charge-discharge cycle including two power turnover processes: charging and discharging; this formula calculates the total charge-discharge turnover power that energy storage can complete within the safe SOC range over its entire life cycle, which serves as the denominator for cost allocation.

[0061] By allocating the total investment and maintenance costs of energy storage throughout its entire lifecycle to the unit turnover volume, the equivalent turnover cost of degradation and maintenance is obtained. This enables precise accounting of energy storage losses in terms of electricity volume, solving the problem that fixed maintenance costs cannot reflect the impact of charging and discharging frequency on battery life. It allows lower-level scheduling decisions to endogenously weigh the benefits of charging and discharging against battery losses, avoiding excessive cycling in pursuit of short-term gains that shortens battery life and ensuring optimal overall benefits throughout the entire lifecycle.

[0062] In the embodiments of this application, in step S3, the robust solution method of IGDT includes the following steps: S3.1: Obtain historical data of wind power and photovoltaic power, perform edge density estimation using kernel density estimation method and construct the correlation between wind and solar power output using Copula; sample on the joint distribution of wind and solar power, and obtain several typical scenario sets S and their weights through scenario reduction; S3.2: Determine the benchmark sequence of electricity price and frequency regulation price, set the risk tolerance and price deviation allocation weights, and bind the scenarios and weights obtained in S3.1; S3.3: Establish the price uncertainty set of the two types of markets, electricity and frequency regulation, and use the comprehensive deviation index to simultaneously describe the relative volatility of electricity price and frequency regulation price. S3.4: For each scenario, under the lower-level daily operation constraints of S2, solve for the benchmark daily revenue using the benchmark electricity price; S3.5: Within the price uncertainty set, take the most unfavorable case for the electricity price and solve for the most unfavorable daily revenue under the lower-level daily operation constraints of S2; S3.6: Apply a revenue guarantee constraint to all scenarios. If the most unfavorable daily revenue does not meet the guarantee constraint, adjust the risk tolerance or upper-level decision variables until they are met; S3.7: Under the premise of satisfying S3.6, perform joint optimization with the annual robust net revenue as the objective. When the optimal value is reached or iterative convergence is achieved, output the optimal energy storage capacity and power configuration and the corresponding daily operation plan curve.

[0063] In one possible embodiment, step 3.1, generating typical wind and solar power scenarios, is first performed. Historical daily power output data for wind and solar power are obtained. The marginal probability densities of wind and solar power outputs are fitted using kernel density estimation. Then, a Copula connection function is used to construct a nonlinear correlation between wind and solar power outputs, resulting in a joint probability distribution of wind and solar power outputs. A large number of random samples are taken from the joint distribution to obtain an initial wind and solar power output sample set. Then, a clustering scenario reduction algorithm is used to aggregate the large number of samples into a small number of typical scenarios, resulting in a typical scenario set S and the annual weight corresponding to each scenario. Next, step 3.2, configuring benchmark parameters and risk parameters, is performed. The hourly benchmark sequences for electricity prices and frequency regulation prices are determined, typically using the historical average electricity price for the same period or the day-ahead forecast electricity price as the benchmark value. The acceptable risk tolerance parameter for decision-makers and the weighting of energy price deviation and frequency regulation price deviation are set, and these parameters are bound to the typical scenarios obtained in step 3.1, configuring corresponding price benchmarks and risk parameters for each scenario. Then, step 3.3 is executed to establish the price uncertainty sets for the two markets: electricity and frequency regulation. A comprehensive deviation index is used as a unified measure to describe the relative fluctuation range of electricity and frequency regulation prices, constraining prices within a range above and below the benchmark value, forming hourly price range constraints to characterize price uncertainty. Next, step 3.4 is executed to solve for the benchmark daily revenue. For each typical scenario, under all lower-level daily operational constraints in step 2, the benchmark electricity price sequence is input to solve the lower-level operational optimization model, obtaining the daily operational revenue corresponding to the benchmark electricity price in that scenario, which serves as a reference benchmark for subsequent robustness verification. Then, step 3.5 is executed to solve for the most unfavorable daily revenue. Within the range of the price uncertainty set, the electricity price value most unfavorable to operational revenue is selected, i.e., the lower limit of the electricity sales price and the lower limit of the frequency regulation price. Similarly, under all lower-level operational constraints, the lower-level model is solved to obtain the daily operational revenue corresponding to the most unfavorable price in that scenario. Next, step 3.6, minimum guarantee constraint verification, is performed. Minimum guarantee constraints are applied to each of the typical scenarios to verify whether the revenue on the worst-case day is not lower than the set minimum revenue limit. If any scenario fails to meet the minimum guarantee constraint, the risk tolerance parameter is adjusted, or the upper-level energy storage capacity and power decision variables are adjusted before resolving the lower-level model, until all scenarios meet the minimum guarantee revenue requirements. Finally, step 3.7, joint optimization solution, is performed. Under the premise that all scenarios meet the minimum guarantee constraint, with the goal of maximizing the upper-level annual robust net revenue, the energy storage capacity, energy storage power, and comprehensive deviation index are jointly optimized and iterated. When the objective function reaches its optimal value or the iteration converges, the optimal energy storage capacity and power configuration results are output, along with the hourly daily operation plan curve for each typical scenario.

[0064] By generating typical wind and solar power scenarios, constructing a price uncertainty set, solving for benchmark returns, solving for the worst-case return, verifying the minimum guarantee constraint, and performing joint optimization, a complete optimization framework covering the uncertainties of wind and solar power output and electricity prices is constructed. This solves the problems of existing technologies relying on precise probability distributions, being unable to quantify the minimum return, and the disconnect between planning and operation. This process does not require precise probability distributions of electricity prices, but only requires benchmark forecasts and fluctuation ranges for modeling, making it more applicable to engineering. At the same time, by locking in the returns under the worst-case scenario through the minimum guarantee constraint, it significantly improves the risk resistance of the configuration scheme in a volatile market environment.

[0065] In the embodiments of this application, in step S3.1, the kernel density estimation method (KDE) establishes marginal probability densities for wind power and photovoltaic power respectively, expressed by the following formula: ;in, This is an estimate of the edge probability density of wind power. This is an estimate of the photovoltaic edge probability density. The number of samples; For wind power nuclear power bandwidth; For photovoltaic core bandwidth; For kernel functions; For wind power sample points; These are photovoltaic sample points; the correlation between wind and solar output is constructed using Copula, employing the Frank-Copula cumulative distribution function, expressed as follows: ;in, The value of the Frank-Copula cumulative distribution function; For Copula-related parameters; This represents the cumulative distribution function of the edge distribution of wind power. Let be the cumulative distribution function of the photovoltaic edge distribution; After wind power edge distribution mapping variable; After mapping by photovoltaic edge distribution Variables; The scene reduction adopts a clustering method, and the formula for calculating the optimization objective and scene weights is expressed as follows: ;in, The i-th sampled wind and solar power output curve is the spliced ​​vector. This is the center of the kth typical scene; Number of typical scenarios; This represents the original number of samples; For the scene The weights; Let be the number of samples in the s-th cluster.

[0066] In one possible embodiment, the kernel density estimation method is first used to establish marginal probability densities for wind power and photovoltaic power output, respectively. The density estimation formula for the wind power side is as follows:

[0067] The formula for estimating the density on the photovoltaic side is:

[0068] ;in , These are the marginal probability density estimates for wind power and solar power output, respectively. The number of historical samples; , These are the core bandwidths for wind power and photovoltaic power, respectively, used to control the smoothness of density estimation; For kernel functions, the Gaussian kernel function is usually chosen; , These are historical power output sample points for wind and solar power, respectively. Kernel density estimation is a non-parametric method that does not require prior assumptions that the power output follows a normal or beta distribution. It can flexibly fit any distribution pattern of wind and solar power output, resulting in higher fitting accuracy. Secondly, the Frank-Copula cumulative distribution function is used to construct the correlation structure of wind and solar power output, expressed as:

[0069] ; The cumulative distribution function value of the Frank-Copula; These are the relevant parameters of Copula; the larger the absolute value of the parameter, the stronger the correlation between wind and solar power output. , These are the cumulative distribution functions of the edge distributions of wind power and photovoltaic power, respectively; , The variables, after being mapped by the marginal distribution, follow a uniform distribution between 0 and 1. The Frank-Copula model can characterize the symmetrical nonlinear correlation between variables, adapting to the coupling characteristics of wind and solar power output, and solving the problem that simple linear correlation cannot accurately describe the correlation between wind and solar power output. Then, Monte Carlo random sampling is performed on the wind-solar joint probability distribution to obtain a large number of daily power output curve samples containing the wind-solar correspondence. Each sample is a concatenated vector of hourly wind and solar power output within the day. Finally, a clustering method is used for scene reduction. The optimization objective of clustering is to minimize the sum of squared distances from all samples to their corresponding cluster centers, as shown in the formula:

[0070] The weight calculation formula for each typical scenario is as follows:

[0071] in The concatenation vector of the i-th sampled wind power output curve; This is the center output curve for the k-th typical scenario; The number of preset typical scenarios; This represents the total number of original samples. For the first Weights for a typical scenario; Let s be the number of samples contained in the s-th cluster. Through cluster reduction, tens of thousands of original samples are aggregated into several to dozens of typical scenarios. While preserving the distribution characteristics and correlations of wind and solar power output, the solution scale of the subsequent optimization model is greatly reduced, and the computational efficiency is improved.

[0072] By fitting the marginal distribution of wind and solar power output using kernel density estimation, without pre-setting parameterized assumptions such as normal distribution, it can accurately adapt to the actual multi-peak and asymmetric distribution characteristics of wind and solar power output. By characterizing the nonlinear correlation of wind and solar power output using the Frank-Copula function, it avoids scene distortion caused by independent assumptions, making the generated scene more consistent with the coupling law of actual wind and solar power output. By clustering reduction, massive samples are aggregated into a small number of typical scenes, which significantly reduces the solution scale while preserving statistical characteristics, balancing modeling accuracy and computational efficiency, making engineering-level annual-scale optimization problems solvable.

[0073] In the embodiments of this application, in step S3.3, the establishment of the price uncertainty set for the two markets of electricity energy and frequency regulation establishes a mapping relationship between the comprehensive deviation and the sub-market deviation for the deviation between electricity energy price and frequency regulation price: ;in, The comprehensive deviation index serves as the decision variable for IGDT. This represents the relative deviation of electricity prices. This represents the relative deviation of the FM price. and These are the weights for the electricity price deviation and the frequency regulation price deviation, respectively; the formula for constructing the hourly price uncertainty set interval constraint is expressed as follows: ;in, This serves as a benchmark for predicting electricity prices. The benchmark price for FM mileage; The benchmark price for frequency regulation capacity; in S3.6, the formula for the guaranteed revenue constraint is expressed as: ;in, For the scene The optimal daily return after the most unfavorable price; Daily operating revenue under the benchmark electricity price; As for risk tolerance, A smaller value indicates greater conservatism; in S3.7, the formula for the joint optimization representation is expressed as: Under the conditions of satisfying the guaranteed return constraint and all constraints of S2, the optimal energy storage capacity is obtained by solving the problem. and optimal energy storage power And the corresponding daily operation plan curves for each scenario.

[0074] In one possible implementation, a mapping relationship between the overall deviation and the market-specific deviation is first established, and weights are assigned to the deviation magnitudes of electricity prices and frequency regulation prices, satisfying the following:

[0075] ;in As a comprehensive deviation index, it is the core decision variable of the IGDT method, representing the overall level of price uncertainty; The relative deviation of electricity prices. This represents the relative deviation of the FM price. , These are the weights for electricity price deviation and frequency regulation price deviation, respectively, and their sum is 1. Adjusting these weights allows for flexible allocation of uncertainty levels; for example, when frequency regulation market prices fluctuate more, the weights can be increased. The value of is chosen to make the deviation range of frequency regulation prices larger, and the model is more realistic. Secondly, an hourly price uncertainty set interval constraint is constructed, and the fluctuation range of electricity prices is:

[0076] The price fluctuation range for FM mileage is as follows:

[0077] The price fluctuation range for frequency regulation capacity is as follows: ;in This serves as a benchmark for predicting electricity prices. The benchmark price for FM mileage. Let the frequency regulation capacity be the benchmark price; this uncertainty set is centered on the benchmark price, forming a symmetrical floating range above and below it according to the relative deviation magnitude, within which all possible electricity price values ​​fall. Then, a minimum revenue constraint is set, expressed as:

[0078] ;in For the scene The optimal daily return corresponding to the most unfavorable price. Daily operating revenue under the benchmark electricity price, Risk tolerance; A smaller value indicates a smaller allowable decline in returns, a more conservative decision-making approach, and a higher lower bound on returns; conversely, a larger value indicates a more aggressive decision-making approach, with higher expected returns but also greater risks. Decision-makers can flexibly adjust this parameter according to their own risk tolerance. Finally, the objective function for robust joint optimization is established, expressed as:

[0079] Under the condition of meeting the minimum revenue constraints in all scenarios and all operational constraints at the lower level, the energy storage capacity... Energy storage capacity Comprehensive Deviation Index The three variables are jointly optimized to obtain the optimal energy storage capacity, energy storage power configuration parameters, and the hourly daily operation plan curves for each typical scenario.

[0080] By mapping the comprehensive deviation index to the weights of different markets, a unified quantification and flexible allocation of uncertainty in the energy market and frequency regulation market is achieved. The deviation weights can be adjusted according to the actual fluctuation characteristics of the two types of markets, making the uncertainty set more closely match the real market fluctuation patterns. Through explicit minimum return constraints, the abstract risk tolerance is transformed into an intuitive minimum return, making it convenient for decision-makers to adjust parameters according to their own risk preferences. The final joint optimization uses capacity, power, and robust deviation as collaborative decision variables to pursue maximum net return while ensuring a minimum return, achieving the optimal balance between economy and robustness.

[0081] Example 2:

[0082] Reference Figure 5This application also provides a wind and solar power plant energy storage capacity optimization system based on an IGDT two-layer model. The system is structured in a layered architecture: a data and configuration layer, an uncertainty layer, an optimization core layer, and a solution and output layer. It includes: a data and parameter management module, an uncertainty modeling module, a two-layer optimization model construction module, and a robust solution and plan output module. Specifically, the data and configuration layer includes the data and parameter management module, which is data-connected to both the uncertainty modeling module and the two-layer optimization model construction module. This module is used to uniformly access and store wind power output data, solar power output data, load data, electricity price data, and equipment parameters; manage scenario weights and IGDT parameters; and provide standard data interfaces and version traceability to each layer. The uncertainty layer includes the uncertainty modeling module, which comprises a typical scenario generation submodule and an IGDT uncertainty set submodule. The typical scenario generation submodule is data-connected to the data and parameter management module. The data connection between the determination sub-module and the two-layer optimization model construction module is determined; the typical scenario generation sub-module is used to generate typical wind and solar scenarios and weights based on Copula; the IGDT uncertainty set sub-module is used to map the benchmark electricity price and risk parameters into a price uncertainty set for optimization; the optimization core layer is equipped with the two-layer optimization model construction module, which includes a constraint sub-module. The input of the two-layer optimization model construction module is connected to the data and parameter management module and the uncertainty modeling module, respectively, to establish an integrated model of upper-layer energy storage capacity and power configuration and lower-layer intraday scheduling, and to centrally register power balance constraints, SOC constraints, and frequency regulation capacity reachability constraints; the solution and output layer is equipped with the robust solution and plan output module, whose input is connected to the two-layer optimization model construction module, to perform robust verification and joint optimization under scenario and price range, output the optimal energy storage capacity and power configuration scheme and intraday operation plan curve, and archive them.

[0083] In one possible embodiment, the system's overall layered architecture is first established, divided into four layers according to data flow and logical hierarchy: a data and configuration layer, an uncertainty layer, an optimization core layer, and a solution and output layer. The system is configured with four main functional modules: a data and parameter management module, an uncertainty modeling module, a two-layer optimization model construction module, and a robust solution and plan output module. The uncertainty modeling module embeds a typical scenario generation submodule and an IGDT uncertainty set submodule, while the two-layer optimization model construction module embeds a constraint submodule. Next, the data and parameter management module of the data and configuration layer is deployed, connecting its output to the inputs of both the uncertainty modeling module and the two-layer optimization model construction module. This module uniformly accesses and stores wind power output time-series data, photovoltaic output time-series data, load data, electricity price data, and energy storage device parameters. It also manages typical scenario weights and IGDT risk parameters, provides standardized data reading interfaces to other modules, and supports parameter version traceability to ensure that the data source for optimization calculations is traceable and reproducible. Then, the uncertainty modeling module of the uncertainty layer is deployed. The typical scenario generation submodule inside the module is connected to the data and parameter management module at its input end. It receives historical wind and solar power output data and generates a set of typical wind and solar scenarios and corresponding weights based on Copula correlation modeling and scenario reduction algorithm. Its output end is connected to the input end of the two-layer optimization model construction module, which inputs the scenario data into the optimization model. The IGDT uncertainty set submodule inside the module is connected to the data and parameter management module at its input end. It receives electricity price benchmark data and risk configuration parameters and maps the benchmark electricity price and risk parameters into a price uncertainty range. Its output end is connected to the input end of the robust solution and planned output module, which provides price fluctuation boundaries for robust solution. A two-layer optimization model construction module is then deployed in the core layer, with a constraint submodule within it. This module connects to the data and parameter management module and the uncertainty modeling module at its inputs, receiving equipment parameters, scenario data, and price benchmark data. It is responsible for building an integrated two-layer optimization model encompassing upper-layer energy storage capacity and power configuration, and lower-layer intraday operation scheduling. The constraint submodule centrally registers and manages all constraints, including power balance constraints, SOC constraints, and frequency regulation reachability constraints, providing a unified constraint set to the optimization model. This module establishes a bidirectional data connection with the robust solution and planning output module, outputting the complete optimization model and constraints to the solution module while receiving iteratively updated decision variables from the solution module to dynamically adjust the model's feasible region. Finally, a robust solution and planning output module is deployed in the solution and output layer. This module receives the two-layer optimization model and price uncertainty set data, performs robust verification and joint optimization calculations under typical scenarios and price ranges, and after iterative solving, outputs the optimal energy storage capacity, energy storage power, comprehensive deviation index parameters, and intraday operation plan curves for each scenario. All results are also structured and archived for retrieval by the planning entity.

[0084] Through a four-layer architecture consisting of a data and configuration layer, an uncertainty layer, an optimization core layer, and a solution and output layer, the functions of data management, uncertainty modeling, model building, and solution output are decoupled. Each module has a clear responsibility and standardized interfaces, which facilitates modular development, debugging, and subsequent functional expansion, while also ensuring the consistency and traceability of data flow. At the same time, the clear connection relationships between modules correspond to the complete logical flow of the technical solution. When implementing the project, it can be deployed step by step according to the layered architecture, which reduces the complexity of system implementation and improves the reusability and engineering practicality of the system.

[0085] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0086] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0087] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A method for optimizing the energy storage capacity of wind and solar power plants based on an IGDT two-layer model, characterized in that, Includes the following steps: S1. Construct a two-layer model for optimizing the allocation of wind and solar power plant storage capacity. The two-layer model includes an upper-layer model and a lower-layer model. The upper-layer model uses storage capacity and storage power as decision variables and maximizes annual net income as the objective function. The lower-layer model optimizes the intraday operation and scheduling income of wind, solar and energy storage participating in the power energy market and frequency regulation market, given the upper-layer decision variables, scenarios and price sets. S2. Based on the real-time or predicted operating status of wind and solar power stations and the configured energy storage, establish model constraints. The constraints include power balance constraints, wind and solar power allocation and curtailment conservation constraints, energy storage charging and discharging power decomposition and rated power constraints, frequency regulation capacity reachability constraints, energy storage charge state evolution and upper and lower bound constraints, daily cycle consistency constraints, charging and discharging mutual exclusion constraints, and deviation assessment constraints. S3. A robust solution method for the two-layer model based on IGDT includes: estimating edge density using kernel density estimation based on historical wind and solar data and constructing a nonlinear correlation model of wind and solar output using the Copula function to generate a set of typical scenarios and their weights; determining the electricity price benchmark sequence and setting a risk tolerance, and establishing an electricity price uncertainty set for the two markets of electricity energy and frequency regulation; taking the most unfavorable case for electricity price in each scenario and applying a minimum revenue constraint, and performing joint optimization with annual robust net revenue as the objective. S4. Input historical operating data and equipment parameters of wind power output, photovoltaic power output, load and electricity price, configure key parameters and typical scenario weights of IGDT, solve the two-layer model, and obtain the optimized configuration scheme of energy storage capacity and energy storage power as well as the corresponding daily operation plan curve.

2. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 1, characterized in that, The objective function of the upper-level model in S1 is to maximize annual net income, expressed by the formula: ； in, Let the annual net income objective function be... A collection of indexes for typical scenarios; For the scene The annual weight is the proportion of times it appears; For the scene The optimal daily profit after the most unfavorable price is obtained by solving the lower-level model. Annualized cost of energy storage investment; The annualized cost of energy storage investment is calculated using a capital recovery factor, satisfying the following: ； in, The discount rate; For energy storage lifetime, the smaller value between cycle lifetime and float charge lifetime is taken, i.e.: ； in, For cycle life, For float charge life; This represents the initial total investment for energy storage; The initial total investment for energy storage consists of two parts: capacity investment and power investment, satisfying the following: ； in, The cost per unit capacity is expressed in ¥ / MWh. The cost per unit power is expressed in ¥ / MW. The energy storage capacity is expressed in MWh, and the variable is the decision variable of the upper-level model. , where is the energy storage capacity in MW, and is the decision variable of the upper-level model.

3. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 1, characterized in that, The objective function of the lower-level model that optimizes intraday scheduling revenue is expressed as: ;in, For intraday time step index, This represents the total number of time periods within a day. For the scene Down The price of electricity at any given time, expressed in ¥ / MWh; For the scene Down Power consumption at any given moment, measured in MW; For the scene Down The price of frequency regulation capacity at any given time, in ¥ / (MW·h); For the scene Down Frequency regulation mileage or performance price at any given time, in ¥ / (MW·h); This refers to the mileage intensity coefficient; For the scene Down Frequency modulation capacity at any given time, in MW; The equivalent turnover cost for degradation or maintenance is expressed in ¥ / MWh. For the scene Down Energy storage charging power at any given time, in MW; For the scene Down The energy storage and discharge power at any given moment, expressed in MW; The time step is expressed in hours (h). For the scene Down Cost of prediction deviation assessment at any given time, in ¥; The formula for calculating internet usage electricity consumption is expressed as follows: ； in, The output of wind power sent to the electricity market, measured in MW; The output of photovoltaic power into the electricity market, measured in MW; The discharge power of energy storage in the energy market, measured in MW; The charging power of energy storage in the energy market, measured in MW.

4. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 1, characterized in that, In S2, the constraints include: The formula for the wind and solar power allocation and curtailment conservation constraint is expressed as follows: ； in, For wind power in Available power at any given time, in MW; For wind power in The power used to charge energy storage at any given time, measured in MW; For wind power in Wind curtailment power at any given moment, in MW; For photovoltaics Available power at any given time, in MW; For photovoltaics The power used to charge energy storage at any given time, measured in MW; For photovoltaics The power of light discarded at any given moment, expressed in MW; The formula for the decomposition of energy storage charging and discharging power and the rated power constraint is expressed as follows: ； in, This represents the portion of energy storage discharge power used in the energy market, measured in MW. This is the component of energy storage discharge power used for frequency modulation, measured in MW; The portion of energy storage charging power used in the energy market, measured in MW; The component used for frequency regulation in the energy storage charging power, measured in MW; This represents the upper limit of the rated power of energy storage, measured in MW.

5. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 4, characterized in that, The constraints also include: The formula for the frequency modulation capacity reachability domain constraint is expressed as follows: ； in, Available capacity for frequency modulation, in MW; The formula for the evolution constraint of the energy storage state of charge (SOC) is expressed as follows: ； in, Energy storage period In the scene The energy state under the following conditions is expressed in MWh. For charging efficiency; For discharge efficiency; The formulas for the upper and lower bound constraints of SOC and the daily cycle consistency constraint are expressed as follows: ； in, This represents the lower limit of SOC. This represents the upper limit ratio of SOC. Energy storage capacity, in MWh; The initial energy state of the sun. This represents the end-of-day energy state. The charge-discharge mutual exclusion constraint adopts the Big-M method, and the formula is expressed as follows: ； in, It is a sufficiently large constant; For charging and discharging, there are two mutually exclusive variables. Indicates discharge. Indicates charging; The formula for the deviation assessment cost is expressed as follows: ； in, To assess penalty rates or deviation electricity price coefficients; For submitting or forecasting electricity generation plans; This represents the actual amount of electricity used for internet access.

6. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 3, characterized in that, The degradation or maintenance equivalent turnover cost The formula for calculating it is expressed as follows: ； in, The annualized total cost satisfies: ； in, Annualized operation and maintenance costs; This represents the operational coefficient for capacity. The operation and maintenance coefficient is the power dimension; Total available energy turnover during lifespan The formula is expressed as: ； in, Equivalent cycle life, measured in cycles.

7. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 1, characterized in that, In S3, the robust solution method for IGDT includes the following steps: S3.1: Obtain historical data for wind power and photovoltaic power, estimate edge density using kernel density estimation, and construct the correlation between wind and solar output using Copula; sample the joint wind and solar distribution, and obtain several typical scenario sets through scenario reduction. and their weights; S3.2: Determine the benchmark sequence of electricity price and frequency regulation price, set the risk tolerance and price deviation allocation weights, and bind the scenarios and weights obtained in S3.1; S3.3: Establish the price uncertainty set for the two markets of electricity energy and frequency regulation, and use the comprehensive deviation index to describe the relative fluctuation of electricity energy price and frequency regulation price simultaneously, forming an interval constraint; S3.4: For each scenario, under the lower-level daily operating constraints of S2, the benchmark daily revenue is obtained by solving with the benchmark electricity price; S3.5: Take the most unfavorable case for electricity price within the set of price uncertainties, and solve for the most unfavorable daily revenue under the lower-level daily operation constraints of S2; S3.6: Apply a minimum return constraint to all scenarios. If the minimum return for the worst-case day does not meet the minimum return constraint, adjust the risk tolerance or upper-level decision variables until they are met. S3.7: Under the premise of satisfying S3.6, joint optimization is performed with the annual robust net income as the objective. When the optimal value is reached or the iteration converges, the optimal energy storage capacity and power configuration and the corresponding daily operation plan curve are output.

8. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 7, characterized in that, In S3.1, the kernel density estimation method (KDE) establishes marginal probability densities for wind power and photovoltaic power respectively, expressed by the following formula: ； in, This is an estimate of the edge probability density of wind power. This is an estimate of the photovoltaic edge probability density. The number of samples; For wind power nuclear power bandwidth; For photovoltaic core bandwidth; For kernel functions; For wind power sample points; Photovoltaic sample points; The correlation between wind and solar output is constructed using Copula, employing the Frank-Copula cumulative distribution function, expressed as follows: ； in, The value of the Frank-Copula cumulative distribution function; For Copula-related parameters; This represents the cumulative distribution function of the edge distribution of wind power. Let be the cumulative distribution function of the photovoltaic edge distribution; After wind power edge distribution mapping variable; After mapping by photovoltaic edge distribution variable; The scene reduction method employs clustering, and the formula for calculating the optimization target and scene weights is expressed as follows: ； in, The i-th sampled wind and solar power output curve is the spliced ​​vector. This is the center of the kth typical scene; Number of typical scenarios; This represents the original number of samples; For the scene The weights; Let be the number of samples in the s-th cluster.

9. The method for optimizing the energy storage capacity of wind and solar power plants based on the IGDT two-layer model according to claim 7, characterized in that, In step S3.3, the establishment of price uncertainty sets for the two markets—electric energy and frequency regulation—establishes a mapping relationship between the overall deviation and the market-specific deviations for the deviation between electricity prices and frequency regulation prices. ； in, The comprehensive deviation index serves as the decision variable for IGDT. This represents the relative deviation of electricity prices. This represents the relative deviation of the FM price. and These are the weightings for the electricity price deviation and the frequency regulation price deviation, respectively. The formula for constructing the hourly price uncertainty set interval constraint is expressed as follows: ； in, This serves as a benchmark for predicting electricity prices. The benchmark price for FM mileage; The benchmark price for frequency modulation capacity; In S3.6, the formula for the guaranteed return constraint is expressed as follows: ； in, For the scene The optimal daily return after the most unfavorable price; Daily operating revenue under the benchmark electricity price; As for risk tolerance, The smaller the value, the more conservative the designation. In S3.7, the formula for the joint optimization representation is expressed as: ； Under the conditions of satisfying the guaranteed return constraint and all constraints of S2, the optimal energy storage capacity can be obtained by solving the problem. and optimal energy storage power And the corresponding daily operation plan curves for each scenario.

10. A wind and solar power plant energy storage capacity optimization system based on an IGDT two-layer model, characterized in that, The wind and solar power plant energy storage capacity optimization method based on the IGDT two-layer model as described in any one of claims 1-9 is adopted. The system is configured with a layered architecture consisting of a data and configuration layer, an uncertainty layer, an optimization core layer, and a solution and output layer, including: a data and parameter management module, an uncertainty modeling module, a two-layer optimization model construction module, and a robust solution and plan output module; wherein: The data and configuration layer is equipped with a data and parameter management module. The data and parameter management module is connected to the uncertainty modeling module and the two-layer optimization model construction module respectively. It is used to uniformly access and store wind power output data, photovoltaic output data, load data, electricity price data and equipment parameters, manage scenario weights and IGDT parameters, and provide standard data interfaces and version traceability to each layer. The uncertainty layer includes an uncertainty modeling module, which comprises a typical scenario generation submodule and an IGDT uncertainty set submodule. The typical scenario generation submodule is data-connected to the data and parameter management module, and the IGDT uncertainty set module is data-connected to the two-layer optimization model construction module. The typical scenario generation submodule is used to generate typical wind and solar scenarios and weights based on Copula. The IGDT uncertainty set module is used to map the benchmark electricity price and risk parameters into a price uncertainty set for optimization. The optimization core layer is equipped with the dual-layer optimization model construction module, which includes a constraint sub-module. The input terminals of the dual-layer optimization model construction module are respectively connected to the data and parameter management module and the uncertainty modeling module, which are used to establish an integrated model of upper-layer energy storage capacity and power configuration and lower-layer intraday scheduling, and centrally register power balance constraints, SOC constraints, and frequency regulation capacity reachability constraints. The solution and output layer is equipped with a robust solution and planning output module. The input of the robust solution and planning output module is connected to the two-layer optimization model construction module, which is used to perform robust verification and joint optimization under different scenarios and price ranges, output the optimal energy storage capacity and power configuration scheme and the daily operation plan curve, and archive them.

Images