A collaborative reuse method for a photovoltaic and energy storage system in the electricity spot market environment

Abstract

The present invention discloses a collaborative reuse optimization method for a photovoltaic and energy storage system in the environment of the electricity spot market. The reuse optimization of the photovoltaic and energy storage system is characterized as including a day-ahead bidding stage, a real-time decision-making stage, and a real-time frequency modulation stage. A three-stage stochastic optimization model for the collaborative reuse of the photovoltaic and energy storage system in the spot market is established respectively with the goal of maximizing the total revenue of the electricity energy and frequency modulation markets. Then, the stochastic dual dynamic programming algorithm is used to solve the above three-stage stochastic optimization model, so as to obtain the collaborative reuse scheme of the photovoltaic and energy storage system. In the electricity spot market, the present invention provides both electricity energy and frequency modulation services simultaneously, realizes the maximization of the overall revenue, and provides an effective and economic decision-making reference for the future joint participation of the photovoltaic and energy storage system in the electricity market.

Description

Technical Field

[0001] The present invention relates to the field of power spot market decision-making, and specifically to a spot market that simultaneously provides electric energy and frequency regulation auxiliary services. A photovoltaic storage combined system comprehensively improves the total revenue of the two markets while providing frequency regulation services, and specifically relates to a collaborative multiplexing method of a photovoltaic storage system in a power spot market environment. Background Art

[0002] Photovoltaics and energy storage have become increasingly important in the low-carbon transformation of energy systems and the construction of clean power systems due to their clean characteristics and regulatory performance. Therefore, in recent years, photovoltaics and energy storage have also ushered in a peak period of industrial development.

[0003] At this stage, how to fully and efficiently utilize large-scale photovoltaic and energy storage resources is an important issue that needs to be solved urgently. Studies have shown that the operating characteristics of photovoltaics and energy storage are complementary, and their joint operation can reduce the volatility of photovoltaic output to a certain extent and avoid frequent and deep charging and discharging of energy storage. In addition, the business model of photovoltaic and energy storage combined systems that benefit from participating in the electricity market will gradually become mainstream.

[0004] Chinese patent CN114493222A (publication date May 13, 2022) discloses a multi-market participation strategy for wind farm energy storage power stations that takes into account output forecasts and prices, and establishes an operating revenue model and operating constraints for the joint operation of wind and storage to participate in multiple markets, where multiple markets refer to medium- and long-term, day-ahead and real-time electricity energy markets, that is, the wind storage system only provides electricity energy services, and does not involve the reuse of services such as frequency regulation. Chinese patent CN114463132A (publication date May 10, 2022) discloses a method and system for energy storage to participate in electricity spot market transactions. It sets energy storage to participate in both the electricity energy and frequency regulation markets, but does not consider its synergy with photovoltaics, and the combination of electricity energy and frequency regulation services is only reflected in the capacity allocation of the day-ahead market, without considering the impact of real-time frequency regulation behavior on the revenue of the two markets.

[0005] When the photovoltaic and energy storage systems jointly participate in the electricity market, they usually have two revenue channels. One is to use the price volatility of the spot electricity market to arbitrage electricity energy; the other is to provide frequency regulation auxiliary services and obtain frequency regulation revenue by relying on short-term power following. The working mechanisms of the two revenue channels are different and have a certain degree of coupling. Overall, the existing technologies have failed to fully tap the profit potential of photovoltaics and energy storage in the electricity spot market environment. The strategy of providing electricity market reuse services by the photovoltaic and energy storage combined system urgently needs to be further optimized and improved.

[0006] Purpose of the Invention

[0007] The object of the present invention is to solve the problems faced in the prior art, realize the market decision-making problem of the optical storage system providing both electrical energy and frequency modulation multiplexing services under the background of the electricity spot market, propose an optimization method for the collaborative multiplexing of the optical storage system in the spot market environment, adopt multi-stage stochastic optimization, clarify the interaction relationship between the market decisions of the optical storage system at different stages, deeply explore the market profit space when the optical storage system provides both electrical energy and frequency modulation services, and maximize the overall revenue. The present invention can provide an effective and economic decision-making reference for the future joint participation of the optical storage system in the electricity market. Summary of the Invention

[0008] The present invention provides an optimization method for the collaborative multiplexing of an optical storage system in the electricity spot market environment. The multiplexing optimization of the optical storage system is characterized as including a day-ahead bidding stage, a real-time decision-making stage, and a real-time frequency modulation stage. With the goal of maximizing the total revenue of the electrical energy and frequency modulation markets, a three-stage stochastic optimization model for the collaborative multiplexing of the optical storage system in the spot market is established respectively, and then the stochastic dual dynamic programming algorithm is used to solve the above three-stage stochastic optimization model, including the following steps:

[0009] Step 1: Establish a day-ahead bidding stage stochastic optimization model, that is, determine the bidding quantities of the optical storage system in the day-ahead electrical energy market and the day-ahead frequency modulation market. Specifically, the optical storage system is regarded as a price taker to participate in the electricity market, assuming that its bidding quantities can all be successfully bid in the market clearing. The day-ahead electrical energy market divides the next operating day into 24 hours for bidding quantity declaration respectively; the day-ahead frequency modulation market declares the bidding quantity for the frequency modulation quantity every 5 minutes of the next operating day; the reference value + random quantity method is used to model the random variables of the day-ahead market prices and photovoltaic power outputs respectively. Among them, the reference value of each day-ahead market price is taken as the market clearing price of the day before the optimization. The market clearing price of the day before the optimization is known when the optimization work is carried out, and the random quantity is simulated by the Monte Carlo method; the reference value of the photovoltaic power output is the predicted value in the day-ahead stage, and its prediction deviation value, that is, the random quantity, satisfies the normal distribution, and the Monte Carlo sampling method is also used for simulation; a day-ahead electrical energy market revenue model, a day-ahead frequency modulation market revenue model, and a day-ahead market total revenue model are established respectively, and the modeling constraint conditions of the day-ahead bidding stage are established;

[0010] Step 2: Establish a real-time decision-making stage stochastic optimization model, specifically determine the respective power operation base points of the photovoltaic and energy storage in the optical storage system within each 5 minutes of the operating day. The operation base point is the reference value of the operating power for the energy arbitrage of the optical storage system and also the benchmark for providing frequency modulation service performance; by modeling two random variables including the spot electrical energy market price and the photovoltaic power output, the real-time electrical energy market revenue is modeled, and the modeling constraint conditions of the real-time decision-making stage are established;

[0011] Step 3: Establish an optimization model for the real-time frequency modulation stage. Specifically, determine the frequency modulation power of the photovoltaic and energy storage systems within each 5-minute interval of the operating day. By considering the impact of the frequency modulation power on the frequency modulation performance score and the power integration of the photovoltaic and energy storage systems within 5 minutes, determine the frequency modulation power of the photovoltaic and energy storage systems respectively with the goal of maximizing the total revenue of the two markets.

[0012] Step 4: Solve the model. Specifically, use the stochastic dual dynamic programming algorithm to solve the stochastic optimization model for the day-ahead bidding stage, the stochastic optimization model for the real-time decision-making stage, and the optimization model for the real-time frequency modulation stage, so as to obtain the collaborative reuse scheme of the photovoltaic and energy storage systems.

[0013] Preferably, step 1 further includes the following sub-steps:

[0014] Step S11: Characterize each day-ahead market price and photovoltaic output as shown in Equation (1):

[0015]

[0016] where π da , π cap , π per represent the day-ahead electricity energy market price, the frequency modulation capacity price and the frequency modulation performance price in the day-ahead frequency modulation market respectively; gπ da , gπ cap , gπ per represent the reference values of each market price respectively; Δπ da , Δπ cap , Δπ per represent the random quantities of each market price respectively; PV da,fore represents the predicted value of the day-ahead photovoltaic output; PV da,err represents the deviation of the predicted day-ahead photovoltaic output;

[0017] Step S12: Represent the day-ahead electricity energy market revenue model as shown in Equation (2):

[0018]

[0019] where represents the revenue of the day-ahead electricity energy market for a certain hour; represents the electricity price of the day-ahead electricity energy market for that hour; represents the bid volume for the day-ahead electricity energy market for that hour; the subscript h represents the hour, h = {1, 2,..., 24};

[0020] Step S13: Represent the day-ahead frequency modulation market revenue model as shown in Equation (3):

[0021]

[0022] where Denote the day-ahead frequency regulation market revenue within a certain five minutes of an hour; respectively denote the capacity price and performance price of the 5-minute frequency regulation market; MR is the frequency regulation mileage ratio, representing the ratio of the dispatching signal mileage allocated to the frequency regulation provider to the dispatching signal mileage allocated to traditional resources; Denote the day-ahead frequency regulation market bid volume for this 5 minutes; the subscript t represents the 5-minute time scale, t ∈ T, T = {Δt, 2Δt, 3Δt, … 12Δt}, Δt = 5 minutes;

[0023] Step S14: Represent the total day-ahead market revenue model as shown in Equation (4):

[0024]

[0025] where the subscript 1 represents the first stage, and x (×) represents the decision variable, and the decision variables in the first stage are and ξ (×) represents the random variable;

[0026] Step S15: Characterize the bidding stage constraint conditions as: the bid volume of the PV-ESS system shall not exceed the total charge-discharge amount it can provide. A positive bid volume indicates discharging, and a negative bid volume indicates charging. The bid volume in the frequency regulation market is always greater than 0, specifically represented as shown in Equation (5):

[0027]

[0028] where is the rated capacity of the photovoltaic; are the maximum charge-discharge powers of the energy storage respectively; is the day-ahead frequency regulation market bid volume, that is, the maximum frequency regulation capacity that can be provided;

[0029] Preferably, step 2 further includes the following sub-steps:

[0030] Step S21: Represent the spot electricity energy market price and photovoltaic output as shown in Equation (6):

[0031]

[0032] where π rt represents the real-time electricity energy market price; gπ rt represents its reference value; Δπ rt represents the price random variable; PV rt,fore is the real-time predicted value of the photovoltaic output; PV rt,err is the real-time prediction deviation of the photovoltaic output;

[0033] Step S22: Model the real-time electricity energy market revenue, expressed as shown in Equation (7):

[0034]

[0035] where, represents the total operating base point of the PV and energy storage system, which consists of the PV operating base point and the energy storage operating base point and is expressed as shown in Equation (8):

[0036]

[0037] where, is the decision variable of the second-stage optimization problem;

[0038] Step S23: Express the operating constraint conditions in the decision-making stage as shown in Equation (9):

[0039]

[0040] where, represents the rated capacity of the energy storage; SOC max and SOC min are the maximum and minimum SOC states of the energy storage respectively; E h,t represents the energy value stored in the energy storage; η is the comprehensive charge-discharge efficiency of the energy storage, Δt represents 5 minutes; ΔE h,t represents the change in the stored energy of the energy storage in the t time period of the hth hour, and E h,t-Δt represents the energy stored in the energy storage in the (t - Δt) time period of the hth hour.

[0041] Preferably, the said Step 3 further includes the following sub-steps:

[0042] Step S31: Set the frequency modulation signal r in the actual market as a random number in the range of [-1, 1], expressed as r ∈ [-1, 1];

[0043] Step S32: Establish a real-time reuse model of the PV and energy storage system, expressed as shown in Equation (10):

[0044]

[0045] where, the real-time market revenue difference W3(x3, ξ3) consists of the real-time electricity energy market revenue increment and the frequency modulation market revenue difference ; the real-time electricity energy market revenue increment is compared with the electricity energy market revenue in the second stage; the frequency modulation market revenue difference is compared with the day-ahead frequency modulation market revenue in the first stage For; G h,t Is the frequency modulation performance deviation value; ΔG h,t Represents the frequency modulation performance deviation value, which refers to the difference between the frequency modulation performance score and 1, that is, 1 + ΔG h,t = G h,t ; Represents the real-time frequency modulation power of the photovoltaic energy storage system, which is a theoretical decision variable; Is the introduced intermediate decision variable; r h,t Represents the frequency modulation signal;

[0046] Step S33: Represent the frequency modulation stage constraint conditions as shown in Equation (11):

[0047]

[0048] Among them, And Respectively represent the frequency modulation powers of the photovoltaic and energy storage; Is the actual output level of the photovoltaic, which is a known quantity in this stage; Represents the rated capacity of the energy storage; SOC max 、SOC min Are the maximum and minimum SOC states of the energy storage respectively; E h,t Represents the energy value stored in the energy storage; η is the comprehensive charge and discharge efficiency of the energy storage; Represents the real-time frequency modulation power of the photovoltaic energy storage system; Is the introduced intermediate decision variable; r h,t Represents the frequency modulation signal; Is the decision variable of the second-stage optimization problem; Are the maximum charge and discharge powers of the energy storage respectively.

[0049] Preferably, step 4 further includes:

[0050] Step S41: Establish a multi-stage stochastic optimization model, with the total revenue of the photovoltaic energy storage system participating in the market as the objective function, and the maximization as the optimization direction, expressed as shown in Equation (12):

[0051] min-W = -E(W1(x1,ξ1)+E(W2(x2,ξ2)+E(W3(x3,ξ3)))) (12);

[0052] Step S42: Apply the stochastic dual dynamic programming algorithm for optimization and solution, including: First, use the sample mean estimation principle to transform the original stochastic problem into a certain number of deterministic problems; Then, divide each single-stage problem into a master problem and a sub-problem, and use appropriate dual variables to carry out coordinated fusion in the order of stages; For each stage described in Step 1, Step 2, and Step 3, the master problem description refers to the current cost described by the objective function and the future cost approximately obtained by characterizing the corresponding dual problem in a series of scenarios; The stochastic dual dynamic algorithm continuously iterates between the predecessor process and the successor process. The predecessor process generates possible trial decisions for all stages and updates the upper bound of this multi-stage stochastic programming problem; Then, the successor generates Benders Cuts including optimal Cuts and feasibility Cuts, and before the next iteration, adds them as constraint conditions to the original stochastic problem. Each time after returning to the original stochastic problem and solving it, a new optimized lower bound is obtained. When the condition set between the upper bound and the lower bound is met, the iteration ends. Description of the Drawings

[0053] Figure 1 It is a comparison chart of power curves under the electricity spot market, without considering and considering the two cases of electricity energy and frequency modulation multiplexing.

[0054] Figure 2 It is a flow framework diagram of the collaborative multiplexing method of the optical storage system described in the present invention.

[0055] Figure 3 It is a solution flowchart of the stochastic dual dynamic programming algorithm in the method described in the present invention. Detailed Embodiment

[0056] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0057] Figure 1 It is a comparison chart of frequency modulation signals and power curves under the electricity spot market, without considering and considering the two cases of electricity energy and frequency modulation multiplexing. f(t) represents the curve of the frequency modulation signal changing with time, and x(t) represents the curve of the frequency modulation power changing with time. As Figure 1 shown, without considering the two cases of electricity energy and frequency modulation multiplexing, f(t) = x(t), and ∫ 5min f(t)dt = ∫ 5min∫x(t)dt = 0, while considering the two cases of electricity energy and frequency modulation multiplexing, f(t) ≠ x(t), and ∫ 5min f(t)dt = 0, ∫ 5min x(t)dt ≠ 0.

[0058] Figure 2 This is the flow framework diagram of the collaborative multiplexing method for the optical storage system described in the present invention. As shown in the figure, the present invention divides the decision-making problem of the optical storage system jointly participating in the power market into three stages: the day-ahead bidding stage, the real-time decision-making stage, and the real-time frequency modulation stage. Aiming at maximizing the total revenue of the electricity energy and frequency modulation markets, a three-stage stochastic optimization model for the collaborative multiplexing of the optical storage system in the spot market is established, which specifically includes the following steps:

[0059] Step 1: Day-ahead bidding stage. Determine the bidding quantities of the optical storage system in the day-ahead electricity energy market and the frequency modulation market. The optical storage system is regarded as a price taker to participate in the power market, that is, it is defaulted that all its bidding quantities can win the bid in the market clearing. The day-ahead electricity energy market divides the next day (operating day) into 24 hours for bidding quantity declaration respectively; while the day-ahead frequency modulation market declares the frequency modulation quantity every 5 minutes for the next day (operating day).

[0060] Step 1 can be further divided into the following sub-steps

[0061] 1-1) Stochastic variable modeling

[0062] In the present invention, the processing of stochastic variables in each stage adopts the method of reference value + random quantity. The stochastic variables in the day-ahead bidding stage (the first stage) include the day-ahead market prices and the photovoltaic output. Among them, the reference values of the day-ahead market prices are taken as the market clearing prices of the day before optimization (this price is known when conducting the optimization work), and the random quantities are simulated by the Monte Carlo method. The reference value of the photovoltaic output is the predicted value in the day-ahead stage, and its prediction deviation value (i.e., the random quantity) basically satisfies the normal distribution, with an error of about 20%. The Monte Carlo sampling method is also adopted. Each stochastic variable is respectively characterized as shown in Equation (1):

[0063]

[0064] Among them, π da , π cap , π per respectively represent the day-ahead electricity energy market price, the frequency modulation capacity price and the frequency modulation performance price in the day-ahead frequency modulation market; gπ da , gπ cap , gπ per respectively represent the reference values of each market price; Δπ da , Δπ cap , Δπ per respectively represent the random quantities of each market price; PVda,fore represents the predicted value of the PV output for the day-ahead; PV da,err represents the prediction deviation of the PV output for the day-ahead;

[0065] 1 - 2) Day-ahead electricity energy market revenue model, which is expressed as shown in Equation (2):

[0066]

[0067] where represents the revenue of the day-ahead electricity energy market for a certain hour; represents the electricity price of the day-ahead electricity energy market for that hour; represents the bid volume for the day-ahead electricity energy market for that hour; the subscript h represents the hour, h = {1, 2, …, 24}.

[0068] 1 - 3) Day-ahead frequency regulation market revenue model, which is expressed as shown in Equation (3):

[0069]

[0070] where represents the revenue of the day-ahead frequency regulation market for a certain five-minute period within an hour; respectively represent the capacity price and performance price of the 5-minute frequency regulation market; MR is the frequency regulation mileage ratio, representing the ratio of the mileage of the dispatch signal (such as signal D) allocated to the frequency regulation provider to the mileage of the dispatch signal (signal A) allocated to traditional resources; represents the bid volume for the day-ahead frequency regulation market for that 5-minute period; the subscript t represents the 5-minute time scale, t ∈ T, T = {Δt, 2Δt, 3Δt, … 12Δt}, Δt = 5 minutes.

[0071] The day-ahead bidding stage does not involve specific operating conditions. Therefore, for the convenience of calculation in the frequency regulation market revenue model, the frequency regulation performance score is set to 1, which may deviate from the actual situation. The deviation part will be taken into account in the third stage.

[0072] 1 - 4) Day-ahead market total revenue model, which is expressed as shown in Equation (4):

[0073]

[0074] where the subscript 1 represents the first stage, x (×) represents the decision variable, and the decision variables in the first stage are and ξ (×) represents the random variable.

[0075] 1 - 5) Bidding stage constraint conditions, which are expressed as shown in Equation (5):

[0076]

[0077] Among them, is the rated photovoltaic capacity; are the maximum charge and discharge powers of energy storage respectively. In the day-ahead bidding stage, there is no need to consider the actual output of photovoltaic and energy storage respectively, so they can be regarded as a whole for capacity declaration in the market. The basis for establishing the first two constraints is that the bidding volume of the photovoltaic-energy storage system shall not exceed the total charge and discharge amount it can provide. A positive bidding volume indicates discharge; a negative bidding volume indicates charge. The third constraint means that the bidding volume in the frequency regulation market represents the maximum frequency regulation capacity that can be provided, without direction, so it is always greater than or equal to zero.

[0078] Step 2, real-time decision-making stage. Determine the power operation base points of photovoltaic and energy storage respectively within each 5 minutes of the operating day. The operation base point is the reference value of the operating power for the photovoltaic-energy storage system to conduct energy arbitrage, and it is also the benchmark for examining its performance in providing frequency regulation services.

[0079] Step 2 is further divided into the following sub-steps:

[0080] 2-1) Random variable modeling

[0081] When the time develops to the real-time stage, the prices of each day-ahead market have become known quantities. Therefore, the random variables in the real-time decision-making stage include the spot electricity energy market price and photovoltaic output, which are expressed as shown in Equation (6):

[0082]

[0083] Among them, π rt represents the real-time electricity energy market price; gπ rt represents its reference value; Δπ rt represents the price random variable; PV rt,fore is the real-time predicted value of photovoltaic output; PV rt,err is the real-time prediction deviation of photovoltaic output. The processing method of random variables is basically the same as that in the first stage. The difference is that the reference value of photovoltaic output should be updated to the real-time predicted value (every 5 minutes) in a timely manner, and the real-time prediction deviation can be reduced to about 10%.

[0084] 2-2) Real-time electricity energy market revenue modeling, which is expressed as shown in Equation (7):

[0085]

[0086] The day-ahead and real-time electricity energy markets adopt a double-settlement mode (deviation settlement method). Among them, represents the total operation base point of the photovoltaic-energy storage system, which consists of the photovoltaic operation base point and the energy storage operation base point It consists of two parts, expressed as shown in Equation (8):

[0087]

[0088] Among them, is the decision variable of the second-stage optimization problem.

[0089] 2 - 3) Operating constraints in the decision-making stage, expressed as shown in Equation (9):

[0090]

[0091] Due to the actual operation issues, the constraints in the real-time decision-making stage need to consider the power output / operation of the photovoltaic and energy storage respectively, and also consider their external constraints as a whole. Among them, represents the rated capacity of the energy storage; SOC max , SOC min are the maximum and minimum SOC states of the energy storage respectively; E h,t represents the energy value stored in the energy storage; η is the comprehensive charge-discharge efficiency of the energy storage. The first operating constraint means that the operating base point of the photovoltaic cannot exceed its rated capacity; the second and third operating constraints stipulate that the operating base point of the energy storage cannot exceed its maximum charge-discharge power; the last four operating constraints are all for the state of charge of the energy storage.

[0092] Step 3, real-time frequency modulation stage. Determine the frequency modulation power of the photovoltaic and energy storage within each 5 minutes of the operating day. The purpose of this stage is to optimize the frequency modulation power of the photovoltaic-energy storage system (i.e., optimize the following of its frequency modulation signal), and it is achieved with the goal of maximizing the total revenue of the two markets by considering the impact of the frequency modulation power on the frequency modulation performance score and the power integration of the photovoltaic-energy storage system within 5 minutes.

[0093] Step 3 further includes the following sub-steps:

[0094] 3 - 1) Stochastic variable modeling

[0095] When the time progresses to the frequency modulation stage, the price in the real-time electricity energy market and the actual power output level of the photovoltaic have been realized. Therefore, the stochastic variable in this stage (the third stage) is only the frequency modulation signal. According to the research and statistics, the frequency modulation signal in the actual market is a random number within the interval [-1, 1] and there is no typical pattern. Therefore, in the present invention, no other constraints are imposed on it, and it is only processed as a random number.

[0096] r ∈ [-1, 1]

[0097] 3 - 2) Real-time reuse model of the photovoltaic-energy storage system

[0098] As Figure 1As shown, generally, the frequency regulation service provider receives a frequency regulation signal every 4 s, and the integral of the frequency regulation signal curve with respect to time within 5 min is zero. When the frequency regulation service provider fully responds to the frequency regulation signal, the integral of its frequency regulation power with respect to time within 5 min is also zero accordingly, and its frequency regulation behavior will not generate additional revenue or expenditure in the electricity energy market. At the same time, its frequency regulation performance score is the full value of 1. The multiplexing idea proposed in the present invention means that by optimizing the frequency regulation behavior of the optical storage system, it is not necessarily required to fully follow the frequency regulation signal (i.e., the frequency regulation power is determined by the optimization result and not completely determined by the frequency regulation signal), so as to increase its total revenue in the two markets. Theoretically, to optimize the real-time frequency regulation power of the optical storage system, a time scale of 4 s is required. However, since we take 24 h a day as the optimization interval, such a fine time scale will bring a large dimensional disaster. Therefore, it is considered to still establish a real-time multiplexing model of the optical storage system with a time scale of 5 min. In addition, the calculation of the frequency regulation performance score involves absolute value problems, which causes difficulties in solving the model. Therefore, without changing the essence of the problem, the present invention converts the original non-linear problem into a linear problem by introducing intermediate decision variables. The real-time multiplexing model of the optical storage system is expressed as shown in Equation (10):

[0099]

[0100] Among them, the real-time market revenue difference W3 is composed of the increment of the real-time electricity energy market revenue (compared with the electricity energy market revenue in the second stage) and the difference in the frequency regulation market revenue (compared with the day-ahead frequency regulation market revenue in the first stage ). G h,t is the frequency regulation performance deviation value; ΔG h,t represents the frequency regulation performance deviation value, which refers to the difference between the frequency regulation performance score and 1, that is, 1 + ΔG h,t = G h,t ; represents the real-time frequency regulation power of the optical storage system, which is a decision variable in theory; is the introduced intermediate decision variable; r h,t represents the frequency regulation signal.

[0101] 3-3) Frequency regulation stage constraint conditions, expressed as shown in Equation (11):

[0102]

[0103] Among them, and respectively represent the frequency regulation powers of the photovoltaic and energy storage; is the actual PV output level, which is a known quantity in this stage. The constraint conditions in the real-time frequency modulation stage have a series of frequency modulation power constraints more than those in the real-time decision-making stage, that is, the 4th to 6th constraints, which are expressed as follows: 1) The sum of the frequency modulation powers of PV and energy storage should be equal to the total frequency modulation power of the system to the outside; 2) The absolute values of the positive and negative frequency modulation amounts provided by the system should not exceed the frequency modulation bidding amounts in the day-ahead market.

[0104] 4) Model solution. The present invention uses the stochastic dual dynamic programming algorithm to solve the above model and uses the MATLAB software for modeling and solution.

[0105] 4-1) Multi-stage stochastic optimization model, which is expressed as shown in Equation (12)

[0106] min-W=-E(W1(x1,ξ1)+E(W2(x2,ξ2)+E(W3(x3,ξ3)))) (12),

[0107] Since the multi-stage stochastic optimization model often takes the cost as the objective function, the model defaults to minimization as the optimization direction. The objective function of the present invention is the total revenue when the PV and energy storage system participates in the market, and the optimization direction should be maximization.

[0108] 4-2) Stochastic dual dynamic programming algorithm

[0109] The stochastic dual dynamic programming algorithm is an effective method for solving multi-stage stochastic optimization problems. This method combines the dual theory, the sample average estimation algorithm, and the Benders decomposition. First, the original stochastic problem is transformed into a certain number of deterministic problems by using the sample mean estimation principle. Then, each single-stage problem is divided into a master problem and a sub-problem, and appropriate dual variables are used for coordinated integration in the order of stages. For each stage, the master problem precisely describes the current cost (objective function) and the future cost approximately obtained by characterizing the corresponding dual problem in a series of scenarios.

[0110] As Figure 2 shown, finally, the stochastic dual dynamic algorithm continues through iteration between the forward process and the backward process. It is essentially a process of continuous trial and revision. The forward process generates all possible trial decisions for all stages and updates the upper bound of this multi-stage stochastic programming problem. Then, the backward process generates Benders Cuts (including optimal Cuts and feasibility Cuts). Before the next iteration, the Benders Cuts will be added as constraints to the original problem. After each return to the original problem and its solution, a new optimized lower bound can be obtained. When the conditions set between the upper bound and the lower bound are met, the iteration can end.

[0111] Compared with the prior art, the present invention provides a multi-stage stochastic optimization method for solving the problem of the coordinated participation of a photovoltaic energy storage system in the electricity market and providing multiplexing services, having the following beneficial effects:

[0112] 1. The present invention takes into account the whole process from day-ahead bidding to real-time operation when the photovoltaic and energy storage jointly participate in the electricity spot market. Using the multi-stage stochastic optimization theory, the complex multi-time scale market decision-making problem is divided into three stages: day-ahead bidding, real-time decision-making, and real-time frequency modulation. Moreover, in the optimization process of the first two stages, the expected benefits that may be realized in the subsequent stage operations are considered, fully reflecting the interaction relationship between the decisions of each stage.

[0113] 2. The present invention focuses on the energy and power multiplexing problems when the photovoltaic energy storage system participates in the electricity energy market and the frequency modulation market simultaneously. Not only is the total capacity reasonably allocated according to the prices of the two markets in the day-ahead bidding stage, but also the possibility of improving the total revenue of the two markets by optimizing the actual frequency modulation power in the real-time frequency modulation stage is explored. It breaks the idea in traditional theory that the frequency modulation service provider fully responds to the frequency modulation signal, providing a greater decision-making space and profit space for the problem of the joint participation of the photovoltaic energy storage system in the spot market and providing multiplexing services.

Claims

1. A collaborative reuse optimization method for a photovoltaic and energy storage system in the power spot market environment, which characterizes the reuse optimization of the photovoltaic and energy storage system as including a day-ahead bidding stage, a real-time decision-making stage, and a real-time frequency modulation stage. With the goal of maximizing the total revenue of the electricity energy and frequency modulation markets, a three-stage stochastic optimization model for the collaborative reuse of the photovoltaic and energy storage system in the spot market is established respectively, and then the stochastic dual dynamic programming algorithm is used to solve the above three-stage stochastic optimization model. It is characterized in that, It includes the following steps: Step 1: Establish a stochastic optimization model for the day-ahead bidding stage, that is, determine the bidding volumes of the photovoltaic energy storage system in the day-ahead electric energy market and the day-ahead frequency regulation market. Specifically, regard the photovoltaic energy storage system as a price taker to participate in the power market, assuming that all its bidding volumes can win the bid in the market clearing. The day-ahead electric energy market divides the next operating day into 24 hours for bidding volume declaration respectively; the day-ahead frequency regulation market conducts bidding volume declaration for the frequency regulation volume every 5 minutes of the next operating day; adopt the method of reference value + random quantity to model the random variables of each day-ahead market price and photovoltaic output respectively. Among them, the reference value of each day-ahead market price is taken as the market clearing price of the day before the optimization day, and the market clearing price of the day before the optimization day is known when carrying out the optimization work. The random quantity is simulated by the Monte Carlo method; the reference value of the photovoltaic output is the predicted value in the day-ahead stage, and its prediction deviation value, that is, the random quantity, satisfies the normal distribution and is also simulated by the Monte Carlo sampling method; establish a day-ahead electric energy market revenue model, a day-ahead frequency regulation market revenue model and a day-ahead market total revenue model respectively, and establish the modeling constraint conditions for the day-ahead bidding stage; Step 2: Establish a stochastic optimization model for the real-time decision-making stage. Specifically, determine the respective power operation base points of the photovoltaic and energy storage in the photovoltaic energy storage system every 5 minutes of the operating day. The operation base point is the reference value of the operating power for the energy arbitrage of the photovoltaic energy storage system and is also the benchmark for examining the performance of the provided frequency regulation service; model two random variables including the spot electric energy market price and the photovoltaic output, so as to model the real-time electric energy market revenue, and establish the modeling constraint conditions for the real-time decision-making stage; Step 3: Establish an optimization model for the real-time frequency regulation stage. Specifically, determine the frequency regulation power of the photovoltaic and energy storage every 5 minutes of the operating day. By considering the influence of the frequency regulation power on the frequency regulation performance score and the 5-minute power integration of the photovoltaic energy storage system, determine the frequency regulation power of the photovoltaic and energy storage respectively with the maximization of the total revenue of the two markets as the goal; Step 4: Model solution. Specifically, use the stochastic dual dynamic programming algorithm to solve the stochastic optimization model for the day-ahead bidding stage, the stochastic optimization model for the real-time decision-making stage and the optimization model for the real-time frequency regulation stage, so as to obtain the collaborative reuse scheme of the photovoltaic energy storage system.

2. A collaborative reuse optimization method for a photovoltaic and energy storage system in a power spot market environment according to claim 1, characterized in that, The said Step 1 further includes the following sub-steps: Step S11: Characterize each day-ahead market price and photovoltaic output as shown in Equation (1): Among them, π da , π cap , π per respectively represent the price of the day-ahead electricity energy market, the price of the frequency regulation capacity and the price of the frequency regulation performance in the day-ahead frequency regulation market; gπ da , gπ cap , gπ per respectively represent the reference values of the prices of each market; Δπ da , Δπ cap , Δπ per respectively represent the random variables of the prices of each market; PV da,fore represents the predicted value of the day-ahead output of the photovoltaic power; PV da,err represents the deviation of the predicted day-ahead output of the photovoltaic power; Step S12: Represent the day-ahead electric energy market revenue model as shown in Equation (2): Among them, represents the revenue of the day-ahead electricity energy market before a certain hour; represents the electricity price of the day-ahead electricity energy market before that hour; represents the day-ahead electricity energy market bid volume for that hour; the subscript h represents the hour, h = {1, 2, …, 24}; Step S13: Represent the day-ahead frequency regulation market revenue model as shown in Equation (3): Among them, represents the day-ahead frequency regulation market revenue for a certain five-minute period within an hour; respectively represent the capacity price and performance price of the 5-minute frequency regulation market; MR is the frequency regulation mileage ratio, which represents the ratio between the mileage of the dispatch signal allocated to the frequency regulation provider and the mileage of the dispatch signal allocated to the traditional resources; represents the day-ahead frequency regulation market bid volume for this 5-minute period; the subscript t represents the 5-minute time scale, t ∈ T, T = {Δt, 2Δt, 3Δt, … 12Δt}, Δt = 5 minutes; Step S14: Represent the day-ahead market total revenue model as shown in Equation (4): Among them, the subscript 1 represents the first stage, and x (·) represents a decision variable. The decision variables in the first stage are and ξ (·) represents a random variable; Step S15: Characterize the bidding stage constraint conditions as: the bidding volume of the photovoltaic energy storage system shall not exceed the total charge and discharge amount it can provide. A positive bidding volume indicates discharge, and a negative bidding volume indicates charge. The bidding volume in the frequency regulation market is always greater than 0, and it is specifically expressed as shown in Equation (5): Among them, is the rated photovoltaic capacity; are the maximum charge and discharge powers of energy storage respectively; is the bidding volume in the day-ahead frequency regulation market, that is, the maximum frequency regulation capacity that can be provided.

3. A collaborative reuse optimization method for a photovoltaic and energy storage system in a power spot market environment according to claim 2, characterized in that The said Step 2 further includes the following sub-steps: Step S21: Represent the spot electric energy market price and photovoltaic output as shown in Equation (6): Among them, π rt represents the real-time electricity energy market price; gπ rt represents its reference value; Δπ rt represents the price random variable; PV rt,fore is the real-time predicted value of the photovoltaic output; PV rt,err is the real-time prediction deviation of the photovoltaic output; Step S22: Model the real-time electricity energy market revenue, expressed as shown in Equation (7): Among them, represents the total operating base point of the photovoltaic and energy storage system, which consists of the photovoltaic operating base point and the energy storage operating base point and is composed of two parts, expressed as shown in Equation (8): Among them, is the decision variable of the second-stage optimization problem; Step S23: Represent the operation constraint conditions in the decision-making stage as shown in Equation (9): Among them, represents the rated energy storage capacity; SOC max , SOC min are the maximum and minimum SOC states of the energy storage respectively; E h,t represents the energy value stored in the energy storage; η is the comprehensive charge-discharge efficiency of the energy storage, Δt represents 5 minutes; ΔE h,t represents the change value of the stored energy of the energy storage in the t time period of the hth hour, E h,t-Δt represents the energy value stored in the energy storage in the (t - Δt) time period of the hth hour.

4. A collaborative reuse optimization method for a photovoltaic and energy storage system in a power spot market environment according to claim 3, characterized in that, Step 3 further includes the following sub-steps: Step S31: Set the frequency modulation signal r in the actual market as a random number within the interval [-1, 1], expressed as r ∈ [-1, 1]; Step S32: Establish a real-time reuse model of the optical storage system, expressed as shown in Equation (10): Among them, the real-time market revenue difference W3(x3, ξ3) consists of the increment of the real-time electricity energy market revenue and the difference in the frequency regulation market revenue . The increment of the real-time electricity energy market revenue is compared with the revenue of the electricity energy market in the second stage; the difference in the frequency regulation market revenue is compared with the day-ahead frequency regulation market revenue in the first stage ; G h,t is the frequency regulation performance deviation value; ΔG h,t represents the frequency regulation performance deviation value, which refers to the difference between the frequency regulation performance score and 1, that is, 1 + ΔG h,t = G h,t ; represents the real-time frequency regulation power of the optical storage system, which is a theoretical decision variable; is the introduced intermediate decision variable; r h,t represents the frequency regulation signal; Step S33: Represent the constraint conditions in the frequency modulation stage as shown in Equation (11): Among them, and represent the frequency regulation power of photovoltaic and energy storage respectively; is the actual output level of photovoltaic, which is a known quantity at this stage; represents the rated capacity of energy storage; SOC max 、SOC min are the maximum and minimum SOC states of energy storage respectively; E h,t represents the energy value stored in energy storage; η is the comprehensive charge-discharge efficiency of energy storage; represents the real-time frequency regulation power of the photovoltaic-energy storage system; is the introduced intermediate decision variable; r h,t represents the frequency regulation signal; is the decision variable of the second-stage optimization problem; are the maximum charge-discharge powers of energy storage respectively.

5. A collaborative reuse optimization method for a photovoltaic and energy storage system in the electricity spot market environment according to claim 4, characterized in that, Step 4 further includes: Step S41: Establish a multi-stage stochastic optimization model, with the total revenue of the optical storage system participating in the market as the objective function, and maximizing as the optimization direction, expressed as shown in Equation (12): min-W=-E(W1(x1,ξ1)+E(W2(x2,ξ2)+E(W3(x3,ξ3)))) (12); Step S42: Apply the stochastic dual dynamic programming algorithm for optimization and solution, including: First, use the sample mean estimation principle to transform the original stochastic problem into a certain number of deterministic problems; Then, divide each single-stage problem into a master problem and a sub-problem, and use appropriate dual variables to coordinate and fuse in the stage order; For each stage described in Step 1, Step 2, and Step 3, the master problem description refers to the current cost described by the objective function and the future cost approximately obtained by characterizing the corresponding dual problem in a series of scenarios; The stochastic dual dynamic programming algorithm continues through iteration between the predecessor process and the successor process. The predecessor process generates possible trial decisions for all stages and updates the upper bound of this multi-stage stochastic programming problem; Then, the successor generates Benders Cuts including optimal Cuts and feasibility Cuts, and before the next iteration, adds them as constraint conditions to the original stochastic problem. Each time it returns to the original stochastic problem and solves it, a new optimized lower bound is obtained. When the conditions set between the upper bound and the lower bound are met, the iteration ends.

Images