Distributed optical storage aggregator multi-stage random optimization method

Abstract

According to the distributed optical storage aggregator multi-stage random optimization method, the price and photovoltaic output uncertainty are processed through the process of historical data statistics-Monte Carlo simulation-scene construction, a generated random sample conforms to actual fluctuation characteristics, and the practicability of decision making is improved; quantifying the bidirectional influence of frequency modulation power on the additional income of the electric energy market and the performance income of the frequency modulation market; the frequency modulation power is dynamically optimized by taking multi-market comprehensive income maximization as a target, the added value of the frequency modulation behavior on the electric energy market is quantified through the income linkage model, and the income of the aggregator is remarkably improved.

Description

Technical Field

[0001] This invention relates to a technology in the field of photovoltaic-storage microgrids, specifically a multi-stage stochastic optimization method for distributed photovoltaic-storage aggregators. Background Technology

[0002] The new energy industry, represented by distributed photovoltaic (PV), has experienced explosive growth, with its installed capacity and power generation continuing to climb, becoming a core force in the transformation of the energy structure. However, the inherent randomness and volatility of distributed PV output, along with its small individual scale and weak market bargaining power, pose serious challenges to the market-based consumption of large-scale PV power. Summary of the Invention

[0003] This invention addresses the problem of large errors in decision-making models due to the use of fixed prediction values ​​in existing technologies. It proposes a multi-stage stochastic optimization method for distributed photovoltaic-storage aggregators. This method handles the uncertainty of price and photovoltaic output through a process of historical data statistics, Monte Carlo simulation, and scenario construction. The generated random samples conform to actual fluctuation characteristics, improving the practicality of decision-making. It quantifies the bidirectional impact of frequency regulation power on the added revenue of the electricity market and the performance revenue of the frequency regulation market. With the goal of maximizing the comprehensive revenue of multiple markets, it dynamically optimizes frequency regulation power and quantifies the added value of frequency regulation behavior to the electricity market through a revenue linkage model, significantly improving the aggregator's revenue.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a multi-stage stochastic optimization method for distributed optical storage aggregators, comprising:

[0006] Step 1: Establish a two-stage joint operation mechanism for the electricity market and the frequency regulation market, and set up a dual-track settlement system;

[0007] Step 2: Calculate the frequency modulation market revenue, which includes capacity and performance revenue, to lay the foundation for multi-market revenue modeling.

[0008] Step 3: Using a multi-stage stochastic programming method, the aggregator decision-making process is decomposed into a three-stage time series framework of day-day bidding, real-time optimization, and real-time frequency adjustment. Each stage is connected to the objective function through stochastic parameter transfer to form a closed loop, with the overall goal of achieving the optimal global return.

[0009] In the aforementioned three-stage time-series framework, the first stage addresses the uncertainties of market prices and photovoltaic output by employing optimal estimation theory. It generates random samples through a process of historical data statistics → Monte Carlo simulation to generate deviations → benchmark value + deviation to construct a sample. This optimizes bidding capacity with the goal of maximizing day-ahead multi-market revenue, while satisfying the constraints of total bidding capacity and frequency regulation capacity limits. The second stage, based on intraday updated photovoltaic forecast data, determines the benchmark operating point P for photovoltaic and energy storage with the goal of maximizing real-time electricity market revenue. t,m,rt,pv P t,m,rt,es The third stage involves constructing a photovoltaic and energy storage SOC and market bidding deviation constraints; aiming to maximize the total revenue across multiple markets, this is achieved by optimizing the time scale to reduce computational complexity, and by using non-negative intermediate variables P. t,m,mid,fr By addressing the nonlinearity of the frequency regulation performance score, a simplified real-time electricity market revenue model R is obtained. t,m,rt,e and the revenue model of the frequency modulation market R t,m,fr .

[0010] Step 4: After obtaining the adjusted objective function of the third-stage temporal framework by combining the objective function, the SDDP algorithm is used to solve the model. The model is iterated until the upper and lower limits converge, and the global optimal decision is output.

[0011] Technical effect

[0012] This invention employs a random sample generation process involving historical data statistics, Monte Carlo simulation, and scenario construction. It generates market price and photovoltaic power output samples that conform to actual fluctuation characteristics by superimposing simulation deviations on benchmark values, specifically addressing the uncertainties of these two types of parameters. Secondly, it constructs a multi-market revenue coupling model that quantifies the bidirectional impact of frequency regulation power, clarifying the interaction between the additional revenue of frequency regulation power on the electricity market and the performance revenue of the frequency regulation market. Thirdly, it designs a three-stage closed-loop stochastic optimization framework of day-ahead bidding, real-time optimization, and real-time frequency regulation, achieving global synergy through the transfer of random parameters at each stage and the connection of the objective function. Fourthly, it proposes a modeling method combining a 4-second to 5-minute optimization timescale extension with non-negative intermediate variables to address the issues of nonlinearity in frequency regulation performance scores and high computational complexity. Compared with existing technologies, this invention can improve decision-making accuracy, avoid bias caused by fixed prediction values, and make the operating strategy more practical. It can also break the limitation of calculating the revenue of the electricity market and the frequency regulation market separately, accurately quantify the comprehensive value of frequency regulation behavior to multiple markets, and thus explore the potential of real-time frequency regulation revenue, maximizing the comprehensive revenue of multiple markets while meeting the physical constraints of the equipment. At the same time, by expanding the time scale and transforming nonlinear problems, it can improve the model solving efficiency and feasibility while ensuring the accuracy of modeling. Combined with the three-stage closed-loop architecture and the SDDP algorithm, it can effectively avoid stage-based greedy decision-making and ultimately achieve global revenue optimization across time scales and markets. Attached Figure Description

[0013] Figure 1 This is a flowchart of the present invention;

[0014] Figure 2 This is a schematic diagram of a three-stage stochastic optimization model. Detailed Implementation

[0015] like Figure 1 As shown in this embodiment, a multi-stage stochastic optimization method for distributed optical storage aggregators is provided, including:

[0016] Step 1: Construct a joint operation framework for the electricity market and the frequency regulation market, specifically including:

[0017] 1.1 Electricity Market: The electricity market mainly consists of two stages: day-ahead trading and real-time trading. In the day-ahead trading stage, market participants must submit their generation-price and demand-price data for the next 24 time periods before 11:00 AM the day before the start date. After the trading center aggregates all quantity-price data, it determines the trading volume and marginal price for each time period through market clearing calculations and announces the clearing results at 1:30 PM on the same day. In the real-time trading stage, market participants can adjust their quantity-price strategies based on the day-ahead market results starting at 6:30 PM. It is important to note that real-time trading uses a more refined 5-minute trading session. Market participants can modify their bids multiple times before the deadline (65 minutes before the start date), and the system will use the last valid bid as the final trading basis. Regarding the settlement mechanism, a dual-track settlement model is adopted: day-ahead trading volume is settled according to the day-ahead market price, while deviations between actual execution and the day-ahead plan are settled according to the real-time market price. Specifically: , where: R e For the electricity market revenue; t ∈ T, T={1,2,…,24}; m ∈ M, M={Δm, 2Δm,…,12Δm}, Δm = 5min; λ t,da,e P represents the clearing price in the day-ahead electricity market during period t; t,da,e λ represents the volume of electricity contracts won by aggregators in the day-ahead electricity market during period t; t,m,rt,e P represents the clearing price in the real-time electricity market for the aggregator during the (t,m) time period; t,m,rt,e This represents the amount of electricity won by the aggregator in the real-time energy market during the (t, m) time period.

[0018] 1.2 Frequency Regulation Market: Participants in the frequency regulation market must complete their applications by 14:15 the day before the operation date. The application information includes the frequency regulation capacity, mileage price, and performance price for each 5-minute period. Participants can still revise their bids after the deadline until 18:30. At 18:30, the trading center will release the final frequency regulation capacity allocation results, and the winning bids for each participant will be determined. During the real-time operation phase, the frequency regulation market and the electricity market will jointly clear, and the settlement price for each trading session (5 minutes) will be announced before the start of that session. During actual operation, the power dispatching agency continuously monitors the grid frequency, calculates the frequency deviation and the required regulation power, and issues control commands to the frequency regulation resources through AGC (Automatic Gain Control).

[0019] The revenue of the aforementioned FM service provider ,in: , , λ t,m,cap, , λ t,m,pi These represent the frequency modulation capacity and performance clearing price in the frequency modulation market during the time period (t, m); P t,m,fr r represents the amount of FM service providers won in the FM service during the time period (t,m); fr Frequency modulation mileage ratio; frequency modulation performance index M t,m,per It is a core parameter for evaluating the frequency modulation quality of market entities, consisting of delay M delay Correlation M rela And accuracy M acc It is composed of three sub-indicators.

[0020] The aforementioned sub-indicators are updated every 10 seconds, and their average value is calculated over a 5-minute period as the final performance score. Among them: delay Correlation ,accuracy α is the time extension when the correlation coefficient is at its maximum; COV(·) is the covariance; P reg P req These represent the frequency modulation demand power and response power, respectively; β(·) is the standard deviation.

[0021] Step 2, construct as follows Figure 2The three-stage stochastic optimization model for distributed photovoltaic-storage aggregators (PV-SGE) employs a multi-stage stochastic programming approach. It designs the decision-making process of PV-SGE aggregators participating in the spot and frequency regulation markets as a three-stage time-series framework. The first stage (day-ahead bidding) addresses capacity allocation optimization. The second stage (real-time optimization) determines the baseline operating point for the distributed PV and energy storage resources represented by the aggregator. The third stage (real-time frequency regulation) constructs a distributed PV-SGE collaborative reuse model, achieving optimal dual-market returns through dynamic adjustment of frequency regulation power. This segmented modeling strategy transforms the complex multi-timescale, multi-market decision-making problem into a series of interconnected sub-problems. Each stage has its own dedicated decision variables and local objective functions, ultimately achieving global optimization of the aggregator's total revenue. Where: R is the total revenue of the aggregator participating in multiple markets; R i (i=1,2,3) represents the objective function for each stage; x1, x2, x3 represent the optimization variables for each stage; δ1, δ2, δ3 represent the random parameters for each stage; E(·) represents the expected value calculation for all possible occurrences of the random parameters for the terms within the parentheses.

[0022] The aforementioned three-stage timing framework specifically includes:

[0023] 2.1 Day-ahead Bidding Stage: The model coefficients in the first stage are not entirely based on deterministic information, but rather incorporate uncertainties such as day-ahead electrical energy and frequency regulation market prices. To ensure the feasibility of solving the model and to meet the unpredictable requirements of stochastic optimization problems, optimal estimation theory is used, with the mean of the overall estimated sample being used as the overall estimate in the calculation.

[0024] The aforementioned overall estimated sample was obtained in the following manner:

[0025] 1) Data Acquisition Phase: Collect actual operating data of typical electricity markets, including time-of-use day-ahead electricity market prices, and frequency regulation market capacity clearing prices and performance clearing prices at 5-minute intervals;

[0026] 2) Statistical analysis stage: Calculate the variance of price fluctuations in each market based on massive historical data;

[0027] 3) Simulation generation stage: Using the Monte Carlo simulation method, a large number of random deviations are generated based on the characteristics of various price fluctuations, which serve as the fluctuation components of the price samples;

[0028] 4) Sample construction stage: Select the market price of a typical day as the benchmark value, and superimpose it with the random deviation generated by the simulation to finally form a complete population estimation sample.

[0029] Besides day-ahead electricity and frequency regulation market prices, the actual output level of distributed photovoltaic (PV) power is also uncertain during the day-ahead bidding stage. To address this issue, this application adopts a modeling method similar to that of the aforementioned market electricity prices: using the day-ahead forecast of PV output as a benchmark, and utilizing Monte Carlo simulation to generate the forecast error.

[0030] The aforementioned day-to-day bidding stage model ,in: , , x1∈{P t,da,e ,P t,m,da,fr}; δ1 includes stochastic parameters such as day-ahead electricity market price, frequency regulation market price, and actual output level of distributed photovoltaic power; R t,da,e For aggregators' day-ahead electricity market revenue; R t,m,da,fr For aggregators' revenue in the current frequency modulation market; λ t,m,cap, , λ t,m,pi These represent the frequency regulation capacity and performance clearing price for the day-ahead frequency regulation market (t, m) period, respectively; P t,m,da,fr This represents the amount of frequency modulation (FM) contracts won by aggregators in the day-ahead FM market during the (t, m) time period.

[0031] The constraints for the day-ahead bidding phase include:

[0032] 1) Total bidding capacity limit constraint: , , where: P rated,pv P represents the rated installed capacity of distributed photovoltaic power. dis,max P ch,max These represent the maximum values ​​of the energy storage charging and discharging power, respectively.

[0033] 2) Frequency Regulation Bidding Capacity Limit: In the electricity market environment, considering the inherent fluctuations in photovoltaic power output, to ensure the actual implementation effect of frequency regulation services, it is clearly required that the frequency regulation capacity declared by new energy entities must simultaneously meet the following requirements: not exceeding 24% of its rated capacity; and not exceeding its maximum ramp power within 5 minutes. Specifically: , , where: μ ramp This represents the power ramp-up rate per minute for distributed photovoltaic power generation.

[0034] 2.2 Real-time Optimization Phase: A baseline value plus random deviation method is adopted, with each random deviation corresponding to a scenario. The objective function of the dual-settlement mode is as follows: , Among them, the bidding volume of distributed photovoltaic (TP) in the real-time electricity market during the (t, m) period is P. t,m,rt,pv The bidding volume for energy storage (t, m) in the real-time electricity market is P. t,m,rt,es .

[0035] The constraints of the objective function of the dual-settlement mode include:

[0036] 1) Constraints of distributed photovoltaic power: ;

[0037] 2) Energy storage constraints: , , , , , , where: E t,m E represents the electrical energy stored during the time interval (t, m). rated,es S represents the rated capacity of the energy storage. min and S max These represent the minimum and maximum states of charge of the energy storage, respectively; η es This refers to the energy storage charging and discharging efficiency.

[0038] 3) Market Bidding Constraints: Market regulations established by the market management to avoid excessive differences between real-time bidding volume and day-ahead bidding volume, specifically: Where: Ψ is the difference coefficient between real-time bidding and day-ahead bidding, which is determined by the market.

[0039] 2.3 Real-time frequency regulation stage: By dynamically optimizing the frequency regulation power curve, the overall benefits of the electricity market and the frequency regulation market are maximized, specifically including:

[0040] 2.3.1 Constructing a Distributed Photovoltaic-Storage Collaborative Reuse Model: The distributed photovoltaic-storage collaborative reuse model is based on the aggregator's real-time electricity market revenue R during the time period (t,m). t,m,rt,e and FM market revenue R t,m,fr It consists of two parts, specifically: real-time total market revenue. Among them: revenue from the hourly electricity market FM market revenue Where: k ∈ K, K={Δk,2Δk,…,75Δk}, Δk = 4 s, used to characterize the response frequency of the distributed optical storage aggregator to the frequency-modulated signal; P t,m,k,fr,res Let P be the frequency modulation power of the distributed optical storage aggregator at time (t, m, k). t,m,k,fr,res When P > 0, the aggregator discharges to provide up-frequency modulation auxiliary services. t,m,k,fr,res When the frequency modulation is less than 0, the aggregator provides down-modulation auxiliary services, and the frequency modulation performance is rated. P t,m,k,fr,req Let (t, m, k) be the frequency modulation request received by the aggregator at time (t, m, k). , ζ t,m,k The frequency-modulated signal received by the distributed optical storage aggregator at time (t, m, k).

[0041] Preferably, this embodiment simplifies the revenue model by extending the optimization timescale of the frequency regulation phase from 4 seconds to 5 minutes. Firstly, the frequency regulation performance score is calculated based on the average frequency regulation accuracy over a 5-minute period; therefore, extending it to this timescale does not affect its evaluation basis. Secondly, the additional revenue in the electricity market is determined solely by the total integral within the 5-minute dispatch cycle, and is unrelated to instantaneous power fluctuations at the second level. Therefore, extending the optimization scale to 5 minutes effectively reduces the number of decision variables and significantly improves the solvability of the model while maintaining the essence of the problem and the accuracy of the model.

[0042] Preferably, this embodiment uses a non-negative intermediate decision variable P. t,m,mid,fr The characteristic of the frequency modulation performance function is as follows: when the frequency modulation power fully responds to the demand, i.e., P... t,m,k,fr,res =-P t,m,k,fr,req At the optimal point, the performance score is 1; as the actual power deviates from the demand, the performance score decreases, specifically as follows: The timescale of the optimization problem has been converted to 5 minutes, where: P t,m,fr,res This represents the total frequency modulation power of the distributed optical storage aggregator within the time period (t, m), reflecting its overall performance; while -r t,m This represents a typical frequency-modulated signal within the time interval (t, m), reflecting the overall system requirements. Constraint P t,m,mid,fr The constraint >=0 is intended to allow the model to avoid absolute value calculations. This constraint ensures that the frequency modulation power P... t,m,fr,res With signal -r t,m The direction is opposite, thus ensuring that the services provided by distributed optical storage aggregators are consistent with system requirements. The practical basis for this constraint is the performance threshold commonly found in the frequency modulation market; if the direction is inconsistent, the performance score will inevitably be lower than the threshold, which is unacceptable in this model. Therefore, the use of intermediate variables simultaneously solves the two major problems of model linearization and market compliance.

[0043] After the above simplification, the real-time electricity market revenue model is transformed into... Transform the FM market revenue model into .

[0044] 2.3.2 The constraints for constructing the distributed optical-storage collaborative reuse model specifically include:

[0045] 1) Constraints on distributed photovoltaic output: , , where: P t,m,fr,pv P represents the frequency-modulated power of photovoltaic power during the time period (t, m). t,m,sj,pv This represents the actual power output of the photovoltaic system during the time period (t, m).

[0046] 2) Energy storage operation constraints: , , , , , , where: P t,m,fr,es The frequency regulation power of the energy storage during the time period (t, m).

[0047] 3) Constraints on the operation of aggregator equipment: ;

[0048] 4) Frequency modulation power upper limit constraint: , ;

[0049] 5) Other constraints: .

[0050] 2.3.3 Adjusting the frequency modulation market revenue model in the distributed optical-storage collaborative reuse model for the objective function: In the day-ahead stage, the frequency modulation performance score is preset to 1, so the model only needs to focus on the deviation ΔM between the actual score and this preset value. t,m,per =M t,m,per The return corresponding to -1. Therefore, the objective function for the third stage is: , where: ΔR t,m,rt,e and ΔR t,m,fr These correspond to the incremental revenue of distributed photovoltaic-storage aggregators in the electricity market and their revenue deviation in the frequency regulation market, respectively, and are calculated in detail as follows:

[0051] .

[0052] 2.4 The three-stage stochastic optimization model for distributed optical storage aggregators can be simplified as follows: , , Where: z2 and z3 are defined as subproblems of the first and second stages, used to evaluate the expected returns of subsequent stages; Aⱼ(j=1,2,3) and Bⱼ(j=1,2,3) are the coefficient matrices of the corresponding stages; bⱼ(j=1,2,3) is the resource vector of the corresponding stage; fⱼ(j=1,2,3) is defined as the cumulative return from the j-th stage to the final stage, so f1 is the global total return.

[0053] 2.5 For the simplified three-stage stochastic optimization problem in step 2.4, the SDDP algorithm is used to obtain an optimized running strategy.

[0054] Compared to existing technologies, this invention addresses the uncertainty of price and photovoltaic output through a process of historical data statistics, Monte Carlo simulation, and scenario construction. The generated random samples conform to actual fluctuation characteristics, improving the practicality of decision-making. It quantifies the dual impact of frequency regulation power on the added revenue of the electricity market and the performance revenue of the frequency regulation market. With the goal of maximizing the comprehensive revenue of multiple markets, it dynamically optimizes frequency regulation power and quantifies the added value of frequency regulation behavior to the electricity market through a revenue linkage model, significantly improving the revenue of aggregators.

[0055] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A multi-stage stochastic optimization method for distributed optical storage aggregators, characterized in that, include: Step 1: Establish a two-stage joint operation mechanism for the electricity market and the frequency regulation market, and set up a dual-track settlement system; Step 2: Calculate the frequency modulation market revenue, which includes capacity and performance benefits, to lay the foundation for multi-market revenue modeling; Step 3: Using a multi-stage stochastic programming method, the aggregator decision-making process is decomposed into a three-stage time series framework of day-day bidding, real-time optimization, and real-time frequency adjustment. Each stage is connected to the objective function through stochastic parameter transfer to form a closed loop, and the overall goal is to achieve the optimal global return. Step 4: After obtaining the adjusted objective function of the third-stage temporal framework by combining the objective function, the SDDP algorithm is used to solve the model. The model is iterated until the upper and lower limits converge, and the global optimal decision is output.

2. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 1, characterized in that, The aforementioned dual-track settlement model refers to the following: day-ahead trading volume is settled based on the day-ahead market price, while the deviation between actual execution and the day-ahead plan is settled based on the real-time market price. Specifically: , where: R e For the electricity market revenue; t∈T, T={1,2,…,24}; m∈M, M={Δm,2Δm,…,12Δm}, Δm=5min; λ t,da,e P represents the clearing price in the day-ahead electricity market during period t; t,da,e λ represents the volume of electricity contracts won by aggregators in the day-ahead electricity market during period t; t,m,rt,e P represents the clearing price in the real-time electricity market for the aggregator during the (t,m) time period; t,m,rt,e This represents the amount of electricity won by the aggregator in the real-time energy market during the (t,m) time period.

3. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 1, characterized in that, The aforementioned FM market revenue refers to the revenue of FM service providers. ,in: , , λ t,m,cap, , λ t,m,pi These represent the FM capacity and performance clearing price in the FM market during the time period (t,m); P t,m,fr r represents the amount of FM service providers won in the FM service during the time period (t,m); fr Frequency modulation mileage ratio; frequency modulation performance index M t,m,per It is a core parameter for evaluating the frequency modulation quality of market entities, consisting of delay M delay Correlation M rela And accuracy M acc It is composed of three sub-indicators.

4. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 1, characterized in that, In the three-stage time series framework, the first stage addresses the uncertainties of market prices and photovoltaic output by adopting the optimal estimation theory. Random samples are generated through a process of historical data statistics → Monte Carlo simulation to generate deviations → benchmark value + deviation to construct samples. The bidding capacity is optimized with the goal of maximizing day-ahead multi-market revenue, while satisfying the upper limit constraints of total bidding capacity and frequency regulation capacity. The second stage is based on intra-day updated photovoltaic prediction data to determine a benchmark operating point P of photovoltaic and energy storage with the objective of maximizing real-time electricity market revenue t,m,rt,pv , P t,m,rt,es , photovoltaic, energy storage SOC and market bidding deviation constraints; the third stage constructs a photovoltaic and energy storage collaborative reuse model with the objective of maximizing total market revenue, reduces the calculation amount through optimization of time scale extension, solves the nonlinear problem of frequency modulation performance score through non-negative intermediate variable P t,m,mid,fr , and then obtains a simplified real-time electricity market revenue model R t,m,rt,e and a frequency modulation market revenue model R t,m,fr .

5. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 1 or 4, characterized in that, The aforementioned three-stage timing framework specifically includes: 2.1 Day-ahead Bidding Stage: The model coefficients in the first stage are not entirely based on deterministic information, but rather on uncertain factors such as day-ahead electrical energy and frequency regulation market prices. To ensure the feasibility of solving the model and to meet the unpredictable requirements of stochastic optimization problems, optimal estimation theory is used to process the model, and the mean of the overall estimated sample is used as the overall estimated value in the calculation. The aforementioned day-to-day bidding stage model ,in: , , x1∈{P t,da,e ,P t,m,da,fr }; δ1 includes stochastic parameters such as day-ahead electricity market price, frequency regulation market price, and actual output level of distributed photovoltaic power; R t,da,e For aggregators' day-ahead electricity market revenue; R t,m,da,fr For aggregators' revenue in the current frequency modulation market; λ t,m,cap, , λ t,m,pi These represent the frequency regulation capacity and performance clearing price for the day-ahead frequency regulation market (t,m) period; P t,m,da,fr This represents the volume of frequency modulation (FM) contracts won by aggregators in the day-ahead FM market during the (t, m) time period. The constraints for the day-ahead bidding phase include: 1) Total bidding capacity limit constraint: , , where: P rated,pv P represents the rated installed capacity of distributed photovoltaic power. dis,max P ch,max These represent the maximum values ​​of the energy storage charging and discharging power, respectively. 2) Frequency Regulation Bidding Capacity Limit: In the electricity market environment, considering the inherent fluctuations in photovoltaic power output, to ensure the actual implementation effect of frequency regulation services, it is clearly required that the frequency regulation capacity declared by new energy entities must simultaneously meet the following requirements: not exceeding 24% of its rated capacity; and not exceeding its maximum ramp power within 5 minutes. Specifically: , , where: μ ramp The power ramp-up rate per minute for distributed photovoltaic power generation; 2.2 Real-time Optimization Phase: A baseline value plus random deviation method is adopted, with each random deviation corresponding to a scenario. The objective function for the dual-settlement mode is as follows: , Among them, the bidding volume of distributed photovoltaic (t,m) in the real-time electricity market is P. t,m,rt,pv The bidding volume for energy storage (t, m) in the real-time electricity market is P. t,m,rt,es ; 2.3 Real-time frequency regulation stage: By dynamically optimizing the frequency regulation power curve, the overall benefits of the electricity market and the frequency regulation market are maximized, specifically including: 2.3.1 Constructing a Distributed Photovoltaic-Storage Collaborative Reuse Model: The distributed photovoltaic-storage collaborative reuse model is based on the aggregator's real-time electricity market revenue R during the time period (t,m). t,m,rt,e and FM market revenue R t,m,fr It consists of two parts, specifically: real-time total market revenue. Among them: revenue from the hourly electricity market FM market revenue Where: k∈K, K={Δk,2Δk,…,75Δk}, Δk=4s, used to characterize the response frequency of the distributed optical storage aggregator to the frequency-modulated signal; P t,m,k,fr,res Let P be the frequency modulation power of the distributed optical storage aggregator at time (t,m,k). t,m,k,fr,res When P > 0, the aggregator discharges to provide up-frequency modulation auxiliary services. t,m,k,fr,res When the frequency modulation is below 0, the aggregator provides down-modulation auxiliary services, and the frequency modulation performance is rated. P t,m,k,fr,req Let (t, m, k) be the frequency modulation request received by the aggregator at time (t, m, k). , ζ t,m,k The frequency-modulated signal received by the distributed optical storage aggregator at time (t,m,k); 2.3.2 The constraints for constructing the distributed optical-storage collaborative reuse model specifically include: 1) Constraints on distributed photovoltaic output: , , where: P t,m,fr,pv P represents the frequency-modulated power of photovoltaic power during the time period (t, m). t,m,sj,pv This represents the actual power output of the photovoltaic system during the time period (t, m). 2) Energy storage operation constraints: , , , , , , where: P t,m,fr,es The frequency regulation power of energy storage during the time period (t, m); 3) Constraints on the operation of aggregator equipment: ; 4) Frequency modulation power upper limit constraint: , ; 5) Other constraints: ; 2.3.3 Adjusting the frequency modulation market revenue model in the distributed optical-storage collaborative reuse model for the objective function: In the day-ahead stage, the frequency modulation performance score is preset to 1, so the model only needs to focus on the deviation ΔM between the actual score and this preset value. t,m,per =M t,m,per The objective function for the third stage, corresponding to a return of -1, is as follows: , where: ΔR t,m,rt,e and ΔR t,m,fr These correspond to the incremental revenue of distributed photovoltaic-storage aggregators in the electricity market and their revenue deviation in the frequency regulation market, respectively, and are calculated in detail as follows: ; 。 6. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 5, characterized in that, The aforementioned overall estimated sample was obtained in the following manner: 1) Data Acquisition Phase: Collect actual operating data of typical electricity markets, including time-of-use day-ahead electricity market prices, and frequency regulation market capacity clearing prices and performance clearing prices at 5-minute intervals; 2) Statistical analysis stage: Calculate the variance of price fluctuations in each market based on massive historical data; 3) Simulation generation stage: Using the Monte Carlo simulation method, a large number of random deviations are generated based on the characteristics of various price fluctuations, which serve as the fluctuation components of the price samples; 4) Sample construction stage: Select typical daily market prices as benchmark values, and superimpose them with the random biases generated by simulation to finally form a complete population estimation sample; In addition to day-ahead electricity and frequency regulation market prices, the actual output level of distributed photovoltaic power is also uncertain during the day-ahead bidding stage. To address this issue, this application adopts a modeling method similar to the aforementioned market electricity price: using the day-ahead forecast value of photovoltaic output as a benchmark, and using Monte Carlo simulation to generate forecast error.

7. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 5, characterized in that, The constraints of the objective function of the dual-settlement mode include: 1) Constraints of distributed photovoltaic power: ; 2) Energy storage constraints: , , , , , , where: E t,m E represents the electrical energy stored during the time interval (t, m). rated,es S represents the rated capacity of the energy storage. min and S max These represent the minimum and maximum states of charge of the energy storage, respectively; η es To improve the charging and discharging efficiency of energy storage; 3) Market Bidding Constraints: Market regulations established by the market management to avoid excessive differences between real-time bidding volume and day-ahead bidding volume, specifically: Where: Ψ is the difference coefficient between real-time bidding and day-ahead bidding, which is determined by the market.

8. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 5, characterized in that, Through a non-negative intermediate decision variable P t,m,mid,fr The characteristic of the frequency modulation performance function is as follows: when the frequency modulation power fully responds to the demand, i.e., P t,m,k,fr,res =-P t,m,k,fr,req At the optimal point, the performance score is 1; as the actual power deviates from the demand, the performance score decreases, specifically as follows: The timescale of the optimization problem has been converted to 5 minutes, where: P t,m,fr,res This represents the total frequency modulation power of the distributed optical storage aggregator within the time period (t,m), reflecting its overall performance; while -r t,m The signal is a typical frequency modulation signal within the time period (t,m), reflecting the overall requirements of the system, with constraint P. t,m,mid,fr The constraint of >=0 is intended to allow the model to avoid absolute value calculations, and this constraint ensures that the frequency modulation power P t,m,fr,res With signal -r t,m The direction is opposite, thus ensuring that the services provided by distributed optical storage aggregators are consistent with the system requirements. The basis for this constraint is the performance threshold that is common in the frequency modulation market. If the direction is inconsistent, the performance score will inevitably be lower than the threshold, which is not allowed by this model. Therefore, the use of intermediate variables solves the two major problems of model linearization and market compliance at the same time. After the above simplification, the real-time electricity market revenue model is transformed into... Transform the FM market revenue model into .

9. The multi-stage stochastic optimization method for distributed optical storage aggregators according to claim 8, characterized in that, The objective function of the adjusted third-stage temporal framework refers to: The three-stage stochastic optimization model of the distributed optical storage aggregator can be simplified as follows: , , Where: z2 and z3 are defined as subproblems of the first and second stages, used to evaluate the expected returns of subsequent stages; Aⱼ(j=1,2,3) and Bⱼ(j=1,2,3) are the coefficient matrices of the corresponding stages; bⱼ(j=1,2,3) is the resource vector of the corresponding stage; fⱼ(j=1,2,3) is defined as the cumulative return from the j-th stage to the final stage, so f1 is the global total return.

Images