A method for joint bidding of energy and reserve market of a light storage system considering uncertainty of photovoltaic power generation and uncertainty of electricity price

Abstract

A kind of photovoltaic power generation uncertainty and electricity price uncertainty are considered the energy-reserve market joint bidding method of photovoltaic storage system, it includes: determining photovoltaic storage system bidding strategy of day-ahead stage electric energy market and reserve market, formulating real-time stage response photovoltaic output uncertainty energy storage power regulation scheme.Based on the typical time sequence scene of photovoltaic output of historical data is generated by means of Latin hypercube sampling and scene reduction method, the day-ahead electricity price scene is formed using multidimensional scene generation;A two-stage bidding-operation optimization model with the goal of maximizing total revenue is constructed;Solving the optimization model obtains the optimal bidding electric quantity, reserve capacity and real-time stage energy storage power regulation result under each type of electricity price scenario in each period.The method can effectively deal with multiple uncertainties, coordinate and optimize the bidding behavior of photovoltaic storage system in different markets;And through reasonable real-time regulation strategy, the deviation risk is reduced, so as to maximize the overall revenue of photovoltaic storage system.

Description

Technical Field

[0001] This invention relates to the field of power system and power market technology, and specifically to a joint bidding method for photovoltaic-storage system energy-reserve market that takes into account the uncertainty of photovoltaic power generation and the uncertainty of electricity prices. Background Technology

[0002] With the deepening of energy transition and the ongoing reform of the electricity market, the penetration rate of distributed photovoltaic (PV) power generation in distribution networks is steadily increasing. This promotes the large-scale utilization of clean energy, but also brings new challenges to the safe and stable operation of distribution networks. The inherent intermittency and uncertainty of PV power generation can lead to voltage exceeding limits and power fluctuations, while its lack of self-regulation capability also limits its competitiveness in the electricity market.

[0003] Against this backdrop, the combined operation of energy storage systems and photovoltaic (PV) power generation has become an effective way to solve the aforementioned problems. Leveraging the characteristics of energy spatial-temporal shifting and rapid power response, PV-storage systems can not only mitigate fluctuations in PV output and improve power quality, but also offer new possibilities for participating in electricity market transactions. In recent years, with the gradual opening of the distribution market, especially the establishment of the electricity market and the backup ancillary service market, conditions have been created for PV-storage systems to participate in market transactions and obtain multiple benefits.

[0004] However, participating in the distribution-side energy-storage market by photovoltaic (PV) and energy storage systems still faces many challenges: First, the uncertainty of PV output and the volatility of market prices complicate bidding decisions, requiring a balance between returns and risks. Second, there is a strong coupling between the electricity market and the backup market; PV and energy storage system bidding for electricity and backup power are mutually constrained, necessitating consideration of synergistic optimization. Furthermore, the connection between the current market and the real-time market further complicates decision-making; PV and energy storage systems need to develop reasonable current bidding strategies and possess the ability to cope with real-time deviations.

[0005] Existing technologies still have several shortcomings in participating in market bidding for photovoltaic and energy storage systems: First, they only consider participation in a single market and cannot fully utilize the synergistic value of photovoltaic and energy storage systems in multiple markets; second, although they consider joint bidding in multiple markets, they ignore the power regulation capability during real-time operation; and finally, they oversimplify uncertainty modeling and do not accurately reflect the actual fluctuation characteristics of photovoltaic output and market prices.

[0006] In conclusion, although some progress has been made in the bidding decision-making method for photovoltaic and energy storage systems, further improvement and refinement are still needed to achieve a globally optimal dynamic balance between time-series foresight and real-time adaptability, energy revenue and ancillary service revenue. Summary of the Invention

[0007] This invention proposes a joint bidding method for photovoltaic-storage system energy-reserve market that takes into account the uncertainties of photovoltaic power generation and electricity price. This method can effectively handle multiple uncertainties, coordinate and optimize the bidding behavior of photovoltaic-storage system in different markets, and reduce deviation risk through reasonable real-time adjustment strategies, thereby maximizing the overall benefits of photovoltaic-storage system and promoting the efficient utilization of distributed energy in the electricity market environment.

[0008] The technical solution adopted in this invention is as follows:

[0009] A joint bidding method for the photovoltaic-storage system energy-standby market that takes into account the uncertainties of photovoltaic power generation and electricity price, includes the following steps: Step 1: Determine the bidding strategy for photovoltaic and energy storage systems in the day-ahead electricity market and the standby market, including photovoltaic power forecast, energy storage charging and discharging plan and standby capacity arrangement, and formulate an energy storage power adjustment scheme that can cope with the uncertainty of photovoltaic power output in the real-time stage.

[0010] Step 2: Typical time-series photovoltaic output scenarios based on historical data are generated using Latin hypercube sampling and scenario reduction methods. A day-ahead electricity price scenario is generated using a multi-dimensional scenario generation method as input for optimization modeling. Step 3: Construct a day-ahead and real-time two-stage bidding-operation optimization model with the goal of maximizing total revenue. This model comprehensively considers electricity market revenue, standby market expenditure, and real-time deviation settlement, and also includes energy storage operation constraints and real-time adjustment constraints.

[0011] Step 4: Solve the bidding-operation optimization model constructed in Step 3 to obtain the optimal bidding amount, reserve capacity, and real-time energy storage power adjustment results for each time period under various electricity price scenarios.

[0012] In step 1, the day-ahead forecast value of photovoltaic power generation at time t is... The lower boundary of the uncertainty range of photovoltaic power generation The difference is defined as the maximum day-ahead forecast deviation for photovoltaic power generation, i.e. ; The energy price, the increase in reserve electricity price, and the decrease in reserve electricity price at time t are respectively expressed as follows: , , ; In the day-ahead energy-reserve market on the distribution side, photovoltaic power generators report day-ahead forecast information, including day-ahead forecast values ​​and the uncertainty range of photovoltaic power generation. The current forecast value falls within the envelope of the uncertainty range for photovoltaic power generation. ; Then the revenue from photovoltaic power generation at time t can be expressed as: Photovoltaic operators need to purchase reserve capacity in the distribution-side reserve market equal to the maximum daily forecast deviation of photovoltaic power generation. The reserve expenditure of photovoltaic power generators is... .

[0013] In step 1, when bidding separately for photovoltaic power, the bid volume in the day-ahead electricity market is the day-ahead forecast value of photovoltaic power generation. The bid volume in the standby market represents the largest forecast deviation. .

[0014] When photovoltaics and energy storage participate in the market together, energy storage arranges day-ahead charging plans. and discharge plan Compared with the day-ahead forecast of photovoltaic power generation Together, they constitute the day-ahead electrical energy bidding information for the photovoltaic-storage system; In addition, energy storage arrangements have increased reserve capacity. and reduce reserve capacity Maximum prediction deviation Together, these constitute the backup purchase bidding information for the photovoltaic energy storage system.

[0015] In step 1, during the real-time phase, the actual output of the photovoltaic-storage system... There is a deviation between the current day's electricity energy bid and the actual bid amount; this deviation will be denoted as... ; When the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system will be charged at a price lower than the day-ahead electricity price. Electricity sold; When the actual output of the photovoltaic and energy storage system is lower than the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system is priced higher than the day-ahead electricity price. The price, accept the punishment.

[0016] Constrained by the reserve capacity determined in the day-ahead market for energy storage, the real-time charging power of energy storage is adjusted. and real-time discharge power This reduces losses caused by deviations in the real-time phase of the photovoltaic-storage system.

[0017] When energy storage is not adjusted in real time If the actual output of the photovoltaic storage system The bid amount exceeded the day-ahead electricity volume, while energy storage capacity was reduced during the day-ahead phase. ;like The real-time charging power of energy storage ;like The real-time charging power of energy storage ; If energy storage does not have a plan to reduce reserve capacity. In this case, energy storage will not be adjusted in real time; If the actual output of the photovoltaic storage system The amount was lower than the day-ahead electricity bid, while energy storage capacity was increased during the day-ahead phase. and The real-time discharge power of the stored energy ; like The real-time discharge power of the stored energy ; If energy storage does not plan to increase reserve capacity. In this case, energy storage will not be adjusted in real time.

[0018] In step 2, the uncertainties faced by the bidding-operation optimization problem of the photovoltaic-storage system include the uncertainty of photovoltaic power generation and the uncertainty of electricity prices. The uncertainty of photovoltaic power generation is described by uncertainty intervals in the day-ahead phase and characterized by multi-scenario technology in the real-time phase.

[0019] In the day-ahead phase, when the day-ahead forecast value of photovoltaic power generation is obtained... Then, based on the day-ahead forecast deviation of photovoltaic power generation... Historical data is used to obtain the uncertainty interval at a given confidence level. Specifically, this includes: First, collect the day-ahead forecasts for the same time period t over the past N days. Actual output of photovoltaic and energy storage systems Calculate the daily forecast deviation. And remove outliers; Secondly, the prediction bias before the fitting date probability distribution: Photovoltaic prediction biases typically follow a normal distribution. The following example uses a normal distribution to estimate the distribution parameters using historical bias data: ①. Calculate the day-ahead forecast deviation Expectations : (1); ②. Calculate the day-ahead forecast deviation Standard deviation : (2); The KS test is used to verify whether the deviation conforms to a normal distribution. If the test passes, the normal distribution is adopted. Otherwise, the t-distribution or mixed distribution can be used.

[0020] Next, based on the preset confidence level α, the corresponding quantiles are obtained from the quantile table to determine the interval boundaries: Taking the normal distribution as an example, the range of quantiles for the deviation is determined as follows: ; Then combined with the previous day's forecast value The uncertainty range of photovoltaic output is derived as follows: ; Finally, adjust the interval: Because photovoltaic power output has a physical upper limit, namely "installed capacity" The interval derived needs to be truncated and corrected to include the lower limit "0". (3).

[0021] To model the uncertainties of photovoltaic (PV) power generation in the real-time phase, we first generate typical PV output scenarios for each time period based on the Latin hypercube sampling (LHS) method and scenario reduction techniques. Then, we combine the single-time-period scenarios into typical time-series PV output scenarios through an iterative reduction and fusion method. The specific steps are as follows: First, random sampling of photovoltaic power output samples during the real-time phase is performed using Latin hypercube sampling (LHS); specifically including: a1: Define the photovoltaic power output distribution model: Based on historical real-time photovoltaic power output data Here, t represents the time period and ξ represents the scenario, and we will perform fitting work on its probability distribution; a2: Operations related to stratification and sampling point generation: The cumulative probability function of the distribution Divided into equal parts There are 3 non-overlapping sub-intervals, each with a width of 1. As shown in equation (4).

[0022] (4); a3: Perform relevant calculations for the sampled values: Generate a random number within each subinterval. To obtain the cumulative probability value Using the inverse function of the distribution Calculate the sample value As shown in equation (5): (5); Secondly, a scenario reduction method was used to reduce the photovoltaic output samples obtained from each time period to identify several representative scenarios; specifically including: b1: Scene preprocessing: After filtering out invalid samples from the sampled scenarios that do not match actual working conditions, the valid scenario set is obtained: .

[0023] b2: Initialization parameter settings: Set the number of target scenes To keep the initial probabilities of each scenario equal, that is , i =1,2,...,N s ′, and define the reduction set J at the same time. t and the reserved set R t ={1,2,...,Ns′}.

[0024] b3: Inter-scene distance calculation: Calculate any two scenarios and The Euclidean distance, used to measure scene similarity, is calculated using the following formula: (6); b4: Remove redundant scenes. For each scene... i First find the scene with the smallest distance. j Then calculate the product of their probabilities. Then remove In the smallest scenario, the reduction set J is updated synchronously. t and the reserved set R t The elements that make it up.

[0025] Finally, the scenarios from each time period are stitched together to form a final set of typical photovoltaic output scenarios for each time period. The specific steps are: repeat the operation until R... t The number of scenes is To obtain representative scenarios .

[0026] To address the uncertainty modeling of electricity prices, the day-ahead electricity price, the increase in reserve electricity price, and the decrease in reserve electricity price all exhibit uncertainty. Based on historical electricity price data, a multi-dimensional typical scenario generation method is used to generate day-ahead electricity price scenarios. The specific steps are as follows: c1: First, historical electricity price data is cleaned to remove outliers and multidimensional features are extracted to ensure that the features reflect the fluctuation patterns and correlations of electricity prices. Let the analysis period be T, and the electricity price at each time t be a 3-dimensional vector. This forms the initial electricity price dataset. T represents the number of samples, and 3 represents the dimension. Let be the electricity price at time t; The reserve electricity price is increased at time t; The reserve electricity price is reduced at time t.

[0027] To eliminate the impact of different electricity price levels on clustering, the electricity price for each dimension is subjected to "mean-removal and modulus normalization", as shown in the following formula: (7); in, This is the original electricity price data; m=1 corresponds to the electricity price, m=2 corresponds to the increase in the reserve electricity price, and m=3 corresponds to the decrease in the reserve electricity price. Let be the sample mean of the electricity price in the m-th dimension; The standardized electricity price characteristic values ​​ensure the comparability of fluctuation patterns across various dimensions.

[0028] c2: Using the weighted K-means clustering algorithm, the standardized electricity price feature matrix is ​​processed. Compression is performed to reduce the number of scenarios. This is achieved by using weighted Euclidean distance to characterize the differences between scenarios, ensuring that the clustering results reflect the weighted impact of electricity prices on photovoltaic and energy storage bidding. (1) Define time i and j The difference in electricity price scenarios is represented by a weighted Euclidean distance, denoted as... See formula (8): (8); in, For the first m The weighting coefficient for electricity-like prices is determined by the degree of impact of electricity prices on the revenue of photovoltaic and energy storage. , They are respectively i time, j Time of the first m Characteristic values ​​after standardization of electricity prices. yes i The time-standardized electricity price vector includes i Three types of electricity pricing after time-of-use standardization: . for i Electricity price at any given moment; for i The reserve electricity price will be raised at any time; for i The reserve electricity price will be lowered at any time.

[0029] (2) Sort by weighted distance between the scene and the origin, as shown in formula (9). For a moment i The weighted distance between the electricity price scenario and the origin.

[0030] (9); Uniformly selected initial centers, totaling t The 2nd (g)th element is selected using the 'g' flag. 1) Use T / (2t) scenarios as the initial centers of the g-th class (g=1,2,...,t) to ensure the uniformity of the initial centers.

[0031] (3) Assign the electricity price scenario at each time point to the corresponding cluster group based on the minimum weighted distance. And update the cluster centers using the within-group mean: (10); in: Let g be the number of samples in the g-th group. This is the center price vector of the g-th cluster, containing the center values ​​of the prices for the three clusters: . This represents the central value of the electricity price for the g-th category; This is the central value for the increased reserve electricity price in category g; This is the central value for the reduced standby electricity price in category g.

[0032] (4) Repeat the grouping and center update until the convergence condition is met: (11); in, , These are the centers of the g-th cluster before and after the update. =0.0001, to ensure the stability of cluster centers.

[0033] c3: After clustering converges, the final number of typical scenarios is determined through error verification, and the probability of occurrence of each scenario is calculated to form a set of electricity price scenarios that can be directly used for bidding optimization.

[0034] (1) First, for each cluster g, g=1,2,...,t, calculate all original standardized electricity price scenarios within the group. With the center of this group The sum of weighted Euclidean distances Divide by the total number of original samples T to obtain the standardized global average difference. Finally, with precision threshold If the condition of being less than the accuracy threshold is not met, the number of cluster centers t is increased, and step 2 is repeated until the error meets the standard.

[0035] In summary, the criteria for determining whether the typical scenarios after clustering can reflect the probability distribution of the original electricity price are as follows: (12); Where: Gg is the scene set of the g-th group, EM( Quantify the differences between a single scenario and a central one.

[0036] (2) Typical scene reconstruction and probability allocation: Standardized cluster centers By destandardizing and restoring to the actual electricity price scenario , The electrical energy price of cluster group g. The reserve electricity price will be increased for cluster group g. The reserve electricity price for cluster group g is reduced.

[0037] (3) Calculate the probability of the scenario occurring based on the proportion of each group's sample size: (13); in: Let g be the probability of the occurrence of the g-th typical scenario, satisfying .

[0038] Finally, t typical electricity price scenarios are generated, each scenario containing the specific values ​​of 3 types of electricity prices and their probability of occurrence.

[0039] In step 3, the photovoltaic-storage system participates in day-ahead market bidding with the goal of maximizing total revenue and performs real-time phase optimization operation. The objective function of the day-ahead and real-time two-stage bidding-operation optimization model is expressed as: (14); In equation (14): , These represent the probabilities for day-ahead electricity price scenarios and real-time photovoltaic power scenarios, respectively. , These are the corresponding scene numbers; Set as time interval ; , , These are the number of day-ahead electricity price scenarios, the number of day-ahead bidding periods, and the number of real-time photovoltaic power generation output scenarios, respectively. and These represent the electricity price and the increased reserve electricity price under the day-ahead electricity price scenario at time t; , These represent the ratios of the settlement price for the portion of electricity delivered by the photovoltaic-storage system that is greater than or less than the day-ahead tendered electricity, respectively, to the day-ahead electricity price. , ; and These are the portions of the electricity generated by the photovoltaic and energy storage system that are greater than and less than the day-ahead electricity bid. , Let represent the day-ahead discharge power plan and day-ahead charging power plan of the photovoltaic-storage system at time t and the ω-th day-ahead electricity price scenario, respectively. This represents the lower boundary value of the uncertainty interval of photovoltaic power generation output at time t.

[0040] Equation (14) includes: Item 1 This indicates the revenue obtained by the photovoltaic and energy storage system from day-ahead electricity market bidding; Item 2 This indicates the expenditure on purchasing spare parts in the spare parts market for photovoltaic energy storage systems; Item 3 This indicates the revenue generated from the excess power generated when the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid during the real-time phase. Item 4 This indicates the penalty imposed on the under-generated electricity when the actual output of the photovoltaic-storage system is less than the day-ahead electricity bid during the real-time phase.

[0041] The constraints of the optimization model include: current energy storage charge and discharge plan constraints, reserve capacity constraints, real-time adjustment power constraints, and real-time deviation power calculation constraints. 1) Current constraints on energy storage charging and discharging plans: The discharge power and charging power constraints are as follows: (15); (16); In the above formula, , Let represent the day-ahead discharge power plan and day-ahead charging power plan of the photovoltaic-storage system at time t and the ω-th day-ahead electricity price scenario, respectively. This represents the charging and discharging plan status of the photovoltaic energy storage system at time t. A value of 1 indicates discharging, and a value of 0 indicates charging. , These represent the maximum discharge power limit and the maximum discharge power limit of the photovoltaic-energy storage system, respectively; t represents the time scale. This indicates the number of the current day electricity price scenario.

[0042] The remaining energy equality constraints and upper and lower bound constraints for energy storage are as follows: (17); (18); In the above formula, This represents the remaining energy stored at time t after the scheduled charge / discharge plan has been arranged. This indicates the remaining energy stored at time t-1 before the current reserve capacity is allocated. , These represent the charge and discharge efficiencies, respectively. , These represent the lower and upper limits of the remaining energy stored, respectively. The energy storage system has equal remaining energy at the beginning and end of the charging and discharging schedule before the scheduled date, expressed as: (19); Equation (19), This indicates the remaining energy at the initial moment after the scheduled charge and discharge plan is completed before the arrangement date; This indicates the remaining energy at the end of the charging and discharging schedule before the scheduled date.

[0043] 2) Energy storage backup capacity constraints: The increase in energy storage reserve capacity should meet the energy storage capacity constraint, and the increase in energy storage reserve capacity should be less than or equal to the difference between the photovoltaic power generation capacity and the photovoltaic day-ahead forecast value. (20); (twenty one); In the above formula, This indicates that the energy storage arrangement will increase the reserve capacity under the ω-th day-ahead electricity price scenario; This indicates the maximum discharge power limit of the energy storage; This indicates the reserve arrangement status of energy storage under the day-ahead electricity price scenario at time t. A value of 1 indicates an upward adjustment of the reserve, which is used for discharging during the real-time phase, while a value of 0 indicates a downward adjustment of the reserve, which is used for charging during the real-time phase. This indicates the installed capacity of photovoltaics.

[0044] The reduction in energy storage reserve capacity should meet the energy storage capacity constraint, and the reduction in energy storage reserve capacity should be less than or equal to the day-ahead forecast value of photovoltaic power.

[0045] (twenty two); (twenty three); In the above formula, This represents the reduced reserve capacity of energy storage arrangement during time period t under the ω-th day-ahead electricity price scenario; This indicates the maximum charging power limit for energy storage; Taking into account the day-ahead charge / discharge plan and reserve capacity arrangement of energy storage, the remaining energy equation constraints and the upper and lower limits of the remaining energy are as follows: (twenty four); (25); In the above formula, This represents the remaining energy stored at time t after the reserve capacity has been allocated. This indicates the remaining energy stored at time t-1 before the current reserve capacity is allocated. After the scheduled charge / discharge plan and reserve capacity are set before the energy storage date, the remaining energy at the beginning and end times are equal: (26); 3) Real-time power constraint adjustment for energy storage: (27); (28); (29); (30); (31); In the above formula, This indicates whether the real-time actual output of the photovoltaic-storage system is lower than its day-ahead electricity bid amount when there is no real-time adjustment of energy storage, under the day-ahead electricity price scenario at time t and the ωth day-ahead. A value of 1 indicates yes, and a value of 0 indicates no. This indicates the maximum output limit of the photovoltaic power generation system; Represents the scenario at time t The actual output of photovoltaic power generation; , They represent the scenarios at time t ( , Real-time discharge and charging power of energy storage; This indicates that the energy storage arrangement will increase the reserve capacity during time period t; This indicates that the energy storage plan will reduce the reserve capacity during time period t.

[0046] 4) Real-time deviation power constraint of photovoltaic-storage system: After real-time energy storage adjustment, the deviation of the actual power generation of the photovoltaic-storage system in the real-time stage from the day-ahead tendered power energy needs to meet the following constraints: (32); (33); In the above formula, , They represent the scenarios at time t ( , After real-time energy storage adjustment, the actual output of the photovoltaic-storage system is higher or lower than the day-ahead electricity bid portion.

[0047] In step 4, the specific process of solving the bidding-operation optimization model is as follows: First, continuous variables in the model, including charging and discharging power, need to be included. and Energy storage energy status and Deviation in charge, etc., and integer variables including state of charge and discharge. Backup arrangement status These should be clearly defined, and their upper and lower limits should be set; Secondly, the probability of combining electricity price scenarios with photovoltaic power output scenarios. , As a weighting factor, it is introduced into the objective function to reflect the expected returns under different scenarios; Next, the Cplex solver is used to call the branch and bound algorithm to solve the mixed-integer linear programming model.

[0048] After the solution is completed, Cplex will output the optimal decision variable values ​​for each time period under various electricity price scenarios. This yields the optimal bidding volume for each time period under various electricity price scenarios. standby capacity , and the power regulation results of real-time stage energy storage , .

[0049] This invention provides a joint bidding method for the energy-storage market of photovoltaic power generation and electricity price uncertainties, with the following technical advantages: 1) The method proposed in this invention designs a collaborative bidding and real-time adjustment mechanism for photovoltaic-storage systems in the day-ahead power and reserve markets on the distribution side in step 1. Its core advantage lies in constructing an integrated decision-making framework that combines market collaboration, time-series coordination, and risk mitigation. This step integrates the day-ahead charging and discharging plan of energy storage with the reserve capacity arrangement, enabling energy storage to simultaneously serve energy transfer and auxiliary response, breaking through the limitations of traditional single-market bidding and effectively expanding revenue streams. Furthermore, regarding the uncertainty of photovoltaic power generation, step 1 not only leverages energy storage to provide reserve capacity in the day-ahead phase to offset the risk of forecast deviations, but also innovatively designs a real-time power adjustment strategy constrained by the day-ahead reserve arrangement. When the actual photovoltaic output deviates from the forecast, this strategy can be used to quickly and orderly utilize the remaining adjustment capacity of energy storage to absorb power generation deviations in real time, thereby significantly reducing the penalty cost of the real-time market and improving the overall economic efficiency and operational robustness of the photovoltaic-storage system in multi-timescale market environments.

[0050] 2) To address the dual uncertainties of photovoltaic (PV) power and electricity prices, a "LHS sampling + scenario reduction + time series splicing" approach is used to generate PV scenarios, balancing randomness and temporal correlation. Furthermore, "multidimensional standardization + weighted K-means clustering" is employed to characterize the multidimensional correlation of electricity prices and dynamically optimize the number of scenarios. Weighting is further used to highlight the differences in the impact of different electricity price dimensions on bidding decisions. This data-driven uncertainty modeling framework effectively avoids the strong assumptions about probability distributions found in traditional stochastic programming, significantly improving the model's ability to represent real market environments and its robustness to decision-making. Breaking through the limitations of traditional single-dimensional or unrelated scenario modeling, this is the first time a targeted scenario generation scheme designed for dual uncertainties has been developed. It accurately captures the characteristics of PV time series and the coupling relationship with electricity prices, providing high-quality input for subsequent optimization and demonstrating significant innovation.

[0051] 3) The advantage of step 3 of this invention lies in constructing a day-ahead and real-time two-stage stochastic optimization model. This model coordinates multi-timescale decisions for photovoltaic-storage systems within a unified mathematical framework. This model takes maximizing the expected total revenue across multiple scenarios as the objective function, while also considering day-ahead electricity market revenue, reserve capacity purchase costs, and deviation settlement revenue during real-time operation, comprehensively covering the entire lifecycle economic flow of photovoltaic-storage systems participating in the joint market. In terms of constraint system design, the model system integrates physical constraints on energy storage equipment operation, market bidding coupling constraints, and real-time power adjustment constraints based on reserve arrangements to ensure that optimization decisions are physically feasible and compatible with market rules. This modeling method achieves risk-aware and robust decision-making by embedding uncertain scenarios into constraints and objectives. Furthermore, its mixed-integer linear programming form ensures that the model can obtain high-quality feasible solutions within a finite time using commercial solvers, thus providing a theoretically rigorous and engineering-practical optimization tool for collaborative bidding and real-time control of photovoltaic-storage systems in a market environment.

[0052] 4) This method can effectively handle multiple uncertainties, coordinate and optimize the bidding behavior of photovoltaic and energy storage systems in different markets, and reduce deviation risks by using reasonable real-time adjustment strategies, thereby maximizing the overall benefits of photovoltaic and energy storage systems and promoting the efficient utilization of distributed energy in the electricity market environment. Attached Figure Description

[0053] The present invention will be further described below with reference to the accompanying drawings and examples; Figure 1 A flowchart outlining the bidding process for photovoltaic-storage systems to participate in the distribution-side energy-sustainment market.

[0054] Figure 2 This is a schematic diagram of the day-ahead bidding and real-time phase strategies for a photovoltaic-storage system.

[0055] Figure 3 This is a schematic diagram of the real-time stage strategy for energy storage.

[0056] Figure 4 Generate an overall process flowchart for a typical scenario with dual uncertainties in photovoltaics and electricity prices.

[0057] Figure 5 The results are the day-ahead bidding results, including the day-ahead electricity market bidding results and the standby market bidding results.

[0058] Figure 6 A schematic diagram of the day-ahead charge / discharge schedule and reserve capacity arrangement for energy storage.

[0059] Figure 7 This represents the real-time stage results of energy storage, including real-time adjusted output and real-time total output. Detailed Implementation

[0060] This invention proposes a joint bidding method for photovoltaic-storage systems (PV-SES) in the energy-reserve market, taking into account the uncertainties of PV power generation and electricity price. The method first clarifies the bidding strategies of PV-SES in the day-ahead energy market and the reserve market, encompassing PV power output forecasting, energy storage charging and discharging planning, and reserve capacity configuration, while simultaneously formulating energy storage power adjustment schemes for the real-time phase. Next, relying on Latin hypercube sampling combined with scenario reduction technology, typical time-series scenarios of PV power output are generated, and day-ahead electricity price scenarios are formed through a multi-dimensional scenario generation method. Subsequently, a two-stage bidding-operation optimization model is built, with the goal of maximizing total revenue. This model comprehensively considers energy market revenue, reserve market costs, and real-time deviation settlement, while incorporating energy storage operation constraints and real-time adjustment constraints throughout the entire process. Finally, the optimization model is solved to obtain the optimal bid volume, reserve capacity, and real-time energy storage adjustment strategy for each time period. This invention, through collaborative optimization of the joint bidding strategy for PV-SES in the energy-reserve market, effectively improves the system's revenue level in the day-ahead market, reduces deviation penalty losses in the real-time operation phase, and significantly increases the overall revenue of PV-SES participating in the electricity market.

[0061] A joint bidding method for the photovoltaic-storage system energy-standby market that takes into account the uncertainties of photovoltaic power generation and electricity price, characterized by the following steps: Step 1: Determine the bidding strategy for photovoltaic and energy storage systems in the day-ahead electricity market and the standby market, including photovoltaic power forecast, energy storage charging and discharging plan and standby capacity arrangement, and formulate an energy storage power adjustment scheme that can cope with the uncertainty of photovoltaic power output in the real-time stage.

[0062] Step 2: Typical time-series photovoltaic output scenarios based on historical data are generated using Latin hypercube sampling and scenario reduction methods. The day-ahead electricity price scenario generated by the multi-dimensional scenario generation method is used as the input for optimization modeling.

[0063] Step 3: Construct a day-ahead and real-time two-stage bidding-operation optimization model with the goal of maximizing total revenue. This model comprehensively considers electricity market revenue, standby market expenditure, and real-time deviation settlement, and also includes energy storage operation constraints and real-time adjustment constraints.

[0064] Step 4: Solve the bidding-operation optimization model constructed in Step 3 to obtain the optimal bidding amount, reserve capacity, and real-time energy storage power adjustment results for each time period under various electricity price scenarios.

[0065] In step 1, the day-ahead forecast value of photovoltaic power generation at time t is... The lower boundary of the uncertainty range of photovoltaic power generation The difference is defined as the maximum day-ahead forecast deviation for photovoltaic power generation, i.e. ; The energy price, the increase in reserve electricity price, and the decrease in reserve electricity price at time t are respectively expressed as follows: , , ; In the day-ahead energy-reserve market on the distribution side, photovoltaic power generators report day-ahead forecast information, including day-ahead forecast values ​​and the uncertainty range of photovoltaic power generation. The current forecast value falls within the envelope of the uncertainty range for photovoltaic power generation. ; Then the revenue from photovoltaic power generation at time t can be expressed as: Photovoltaic operators need to purchase reserve capacity in the distribution-side reserve market equal to the maximum daily forecast deviation of photovoltaic power generation. The reserve expenditure of photovoltaic power generators is... .

[0066] In step 1, when bidding separately for photovoltaic power, the bid volume in the day-ahead electricity market is the day-ahead forecast value of photovoltaic power generation. The bid volume in the standby market represents the largest forecast deviation. .

[0067] When photovoltaics and energy storage participate in the market together, energy storage arranges day-ahead charging plans. and discharge plan Compared with the day-ahead forecast of photovoltaic power generation Together, they constitute the day-ahead electrical energy bidding information for the photovoltaic-storage system; In addition, energy storage arrangements have increased reserve capacity. and reduce reserve capacity Maximum prediction deviation Together, these constitute the backup purchase bidding information for the photovoltaic energy storage system.

[0068] In step 1, during the real-time phase, the actual output of the photovoltaic-storage system... There is a deviation between the current day's electricity energy bid and the actual bid amount; this deviation will be denoted as... ; When the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system will be charged at a price lower than the day-ahead electricity price. Electricity sold; When the actual output of the photovoltaic and energy storage system is lower than the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system is priced higher than the day-ahead electricity price. The price, accept the punishment.

[0069] This invention uses the reserve capacity of energy storage determined in the day-ahead market as a constraint, and adjusts the real-time charging power of the energy storage. and real-time discharge power This reduces losses caused by deviations in the real-time phase of the photovoltaic-storage system.

[0070] When energy storage is not adjusted in real time If the actual output of the photovoltaic storage system The bid amount exceeded the day-ahead electricity volume, while energy storage capacity was reduced during the day-ahead phase. ;like The real-time charging power of energy storage ;like The real-time charging power of energy storage ; If energy storage does not have a plan to reduce reserve capacity. In this case, energy storage will not be adjusted in real time; If the actual output of the photovoltaic storage system The amount was lower than the day-ahead electricity bid, while energy storage capacity was increased during the day-ahead phase. and The real-time discharge power of the stored energy ; like The real-time discharge power of the stored energy ; If energy storage does not plan to increase reserve capacity. In this case, energy storage will not be adjusted in real time.

[0071] In step 2, the uncertainties faced by the bidding and operation optimization problem of the photovoltaic-storage system mainly include the uncertainty of photovoltaic power generation and the uncertainty of electricity prices. The uncertainty of photovoltaic power generation is described by uncertainty intervals in the day-ahead stage and characterized by multi-scenario technology in the real-time stage.

[0072] In the day-ahead phase, when the day-ahead forecast value of photovoltaic power generation is obtained... Then, based on the day-ahead forecast deviation of photovoltaic power generation... Historical data is used to obtain the uncertainty interval at a given confidence level. Specifically, this includes: First, collect the day-ahead forecasts for the same time period t over the past N days (e.g., N=365). Actual output of photovoltaic and energy storage systems Calculate the daily forecast deviation. And remove outliers; Secondly, the prediction bias before the fitting date probability distribution: Photovoltaic prediction biases typically follow a normal distribution. The following example uses a normal distribution to estimate the distribution parameters using historical bias data: ①. Calculate the day-ahead forecast deviation Expectations : (1); ②. Calculate the day-ahead forecast deviation Standard deviation : (2); The KS test is used to verify whether the deviation conforms to a normal distribution. If the test passes, the normal distribution is adopted. Otherwise, the t-distribution or mixed distribution can be used.

[0073] Next, based on the preset confidence level α, the corresponding quantiles are obtained from the quantile table to determine the interval boundaries: Taking the normal distribution as an example, the range of quantiles for the deviation is determined as follows: ; Then combined with the previous day's forecast value The uncertainty range of photovoltaic output is derived as follows: ; Finally, adjust the interval to ensure its reasonableness: Because photovoltaic power output has a physical upper limit, namely "installed capacity" The interval derived needs to be truncated and corrected to include the lower limit "0". (3).

[0074] To model the uncertainties of photovoltaic (PV) power generation in the real-time phase, we first generate typical PV output scenarios for each time period based on the Latin hypercube sampling (LHS) method and scenario reduction techniques. Then, we combine the single-time-period scenarios into typical time-series PV output scenarios through an iterative reduction and fusion method. The specific steps are as follows: First, random sampling of photovoltaic power output samples during the real-time phase is performed using Latin hypercube sampling (LHS); specifically including: a1: Define the photovoltaic power output distribution model: Based on historical real-time photovoltaic power output data Here, t represents the time period and ξ represents the scenario, and we will perform fitting work on its probability distribution; a2: Operations related to stratification and sampling point generation: The cumulative probability function of the distribution Divided into equal parts There are 3 non-overlapping sub-intervals, each with a width of 1. As shown in equation (4).

[0075] (4); a3: Perform relevant calculations for the sampled values: Generate a random number within each subinterval. To obtain the cumulative probability value Using the inverse function of the distribution Calculate the sample value As shown in equation (5): (5); Secondly, a scenario reduction method was used to reduce the photovoltaic output samples obtained from each time period to identify several representative scenarios; specifically including: b1: Scene preprocessing: After filtering out invalid samples from the sampled scenarios that do not match the actual working conditions (such as scenarios where the output is less than 0), the valid scenario set is obtained. .

[0076] b2: Initialization parameter settings: Set the number of target scenes To keep the initial probabilities of each scenario equal, that is , i =1,2,...,N s ′, and define the reduction set J at the same time. t and the reserved set R t ={1,2,...,Ns′}.

[0077] b3: Inter-scene distance calculation: Calculate any two scenarios and The Euclidean distance, used to measure scene similarity, is calculated using the following formula: (6); b4: Remove redundant scenes. For each scene... iFirst find the scene with the smallest distance. j Then calculate the product of their probabilities. Then remove In the smallest scenario, the reduction set J is updated synchronously. t and the reserved set R t The elements that make it up.

[0078] Finally, the scenarios from each time period are stitched together to form a final set of typical photovoltaic output scenarios for each time period. The specific steps are: repeat the operation until R... t The number of scenes is To obtain representative scenarios .

[0079] To address the uncertainty modeling of electricity prices, the day-ahead electricity price, the increase in reserve electricity price, and the decrease in reserve electricity price all exhibit uncertainty. These three types of prices are obtained through a coordinated optimization scheduling model, thus showing a clear correlation between them. Based on historical electricity price data, a multi-dimensional typical scenario generation method is used to generate day-ahead electricity price scenarios. The specific steps are as follows: c1: First, historical electricity price data is cleaned to remove outliers and multidimensional features are extracted to ensure that the features reflect the fluctuation patterns and correlations of electricity prices. Let the analysis period be T (e.g., 8760 hours in a year), and the electricity price at each time t be a 3-dimensional vector. This forms the initial electricity price dataset. T represents the number of samples, and 3 represents the dimension. Let be the electricity price at time t; The reserve electricity price is increased at time t; The reserve electricity price is reduced at time t.

[0080] To eliminate the impact of different electricity price levels on clustering, the electricity price for each dimension is subjected to "mean-removal and modulus normalization", as shown in the following formula: (7); in, This is the original electricity price data; m=1 corresponds to the electricity price, m=2 corresponds to the increase in the reserve electricity price, and m=3 corresponds to the decrease in the reserve electricity price. Let be the sample mean of the electricity price in the m-th dimension; The standardized electricity price characteristic values ​​ensure the comparability of fluctuation patterns across various dimensions.

[0081] c2: Using the weighted K-means clustering algorithm, the standardized electricity price feature matrix is ​​processed. Compression is performed to reduce the number of scenarios. This is achieved by using weighted Euclidean distance to characterize the differences between scenarios, ensuring that the clustering results reflect the weighted impact of electricity prices on photovoltaic and energy storage bidding. (1) Define time i and j The difference in electricity price scenarios is represented by a weighted Euclidean distance, denoted as... See Formula (8); Formula (8) highlights the impact of key electricity price dimensions through weight allocation, and avoids clustering results from deviating from actual market priorities.

[0082] (8); in, For the first m The weighting coefficient for electricity-like prices is determined by the degree of impact of electricity prices on the revenue of photovoltaic and energy storage. , They are respectively i time, j Time of the first m Characteristic values ​​after standardization of electricity prices. yes i The time-standardized electricity price vector includes i Three types of electricity pricing after time-of-use standardization: . for i Electricity price at any given moment; for i The reserve electricity price will be raised at any time; for i The reserve electricity price will be lowered at any time.

[0083] (2) Sort by weighted distance between the scene and the origin, as shown in formula (9). For a moment i The weighted distance between the electricity price scenario and the origin.

[0084] (9); Uniformly selected initial centers, totaling t The 2nd (g)th element is selected using the 'g' flag. 1) Use T / (2t) scenarios as the initial centers of the g-th class (g=1,2,...,t) to ensure the uniformity of the initial centers.

[0085] (3) Assign the electricity price scenario at each time point to the corresponding cluster group based on the minimum weighted distance. And update the cluster centers using the within-group mean: (10); in: Let g be the number of samples in the g-th group. This is the center price vector of the g-th cluster, containing the center values ​​of the prices for the three clusters: . This represents the central value of the electricity price for the g-th category; This is the central value for the increased reserve electricity price in category g; This is the central value for the reduced standby electricity price in category g.

[0086] (4) Repeat the grouping and center update until the convergence condition is met: (11); in, , These are the centers of the g-th cluster before and after the update. =0.0001, to ensure the stability of cluster centers.

[0087] c3: After clustering converges, the final number of typical scenarios is determined through error verification, and the probability of occurrence of each scenario is calculated to form a set of electricity price scenarios that can be directly used for bidding optimization.

[0088] (1) First, for each cluster g (g=1,2,...,t), calculate all original standardized electricity price scenarios within the group. With the center of this group The sum of weighted Euclidean distances Divide by the total number of original samples T to obtain the standardized global average difference. Finally, with precision threshold If the condition of being less than the accuracy threshold is not met, the number of cluster centers t is increased, and step 2 is repeated until the error meets the standard.

[0089] In summary, the criteria for determining whether the typical scenarios after clustering can reflect the probability distribution of the original electricity price are as follows: (12); Where: Gg is the scene set of the g-th group, EM( Quantify the differences between a single scenario and a central one.

[0090] (2) Typical scene reconstruction and probability allocation: Standardized cluster centers By destandardizing and restoring to the actual electricity price scenario , The electrical energy price of cluster group g. The reserve electricity price will be increased for cluster group g. The reserve electricity price for cluster group g is reduced.

[0091] (3) Calculate the probability of the scenario occurring based on the proportion of each group's sample size: (13); in: Let g be the probability of the occurrence of the g-th typical scenario, satisfying .

[0092] Finally, t typical electricity price scenarios are generated, each scenario containing the specific values ​​of 3 types of electricity prices and their probability of occurrence.

[0093] In step 3, the photovoltaic-storage system participates in day-ahead market bidding with the goal of maximizing total revenue and performs real-time phase optimization operation. The objective function of the day-ahead and real-time two-stage bidding-operation optimization model can be expressed as: (14); In equation (14): , These represent the probabilities for day-ahead electricity price scenarios and real-time photovoltaic power scenarios, respectively. , These are the corresponding scene numbers; Set as time interval ; , , These are the number of day-ahead electricity price scenarios, the number of day-ahead bidding periods, and the number of real-time photovoltaic power generation output scenarios, respectively. and These represent the electricity price and the increased reserve electricity price under the day-ahead electricity price scenario at time t; , These represent the ratios of the settlement price for the portion of electricity delivered by the photovoltaic-storage system that is greater than or less than the day-ahead tendered electricity, respectively, to the day-ahead electricity price. , They were set to 0.9 and 1.1 respectively; and These are the portions of the electricity generated by the photovoltaic and energy storage system that are greater than and less than the day-ahead electricity bid. , Let represent the day-ahead discharge power plan and day-ahead charging power plan of the photovoltaic-storage system at time t and the ω-th day-ahead electricity price scenario, respectively. This represents the lower boundary value of the uncertainty interval of photovoltaic power generation output at time t.

[0094] Equation (14) includes: Item 1 This indicates the revenue obtained by the photovoltaic and energy storage system from day-ahead electricity market bidding; Item 2 This indicates the expenditure on purchasing spare parts in the spare parts market for photovoltaic energy storage systems; Item 3 This indicates the revenue generated from the excess power generated when the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid during the real-time phase. Item 4 This indicates the penalty imposed on the under-generated electricity when the actual output of the photovoltaic-storage system is less than the day-ahead electricity bid during the real-time phase.

[0095] The constraints of the optimization model include: current energy storage charge and discharge plan constraints, reserve capacity constraints, real-time adjustment power constraints, and real-time deviation power calculation constraints. 1) Current constraints on energy storage charging and discharging plans: The discharge power and charging power constraints are as follows: (15); (16); In the above formula, , Let represent the day-ahead discharge power plan and day-ahead charging power plan of the photovoltaic-storage system at time t and the ω-th day-ahead electricity price scenario, respectively. This represents the charging and discharging plan status of the photovoltaic energy storage system at time t. A value of 1 indicates discharging, and a value of 0 indicates charging. , These represent the maximum discharge power limit and the maximum discharge power limit of the photovoltaic-energy storage system, respectively; t represents the time scale. This indicates the number of the current day electricity price scenario.

[0096] The remaining energy equality constraints and upper and lower bound constraints for energy storage are as follows: (17); (18); In the above formula, This represents the remaining energy stored at time t after the scheduled charge / discharge plan has been arranged. This indicates the remaining energy stored at time t-1 before the current reserve capacity is allocated. , These represent the charge and discharge efficiencies, respectively. , These represent the lower and upper limits of the remaining energy stored, respectively. The energy storage system has equal remaining energy at the beginning and end of the charging and discharging schedule before the scheduled date, expressed as: (19); Equation (19), This indicates the remaining energy at the initial moment after the scheduled charge and discharge plan is completed before the arrangement date; This indicates the remaining energy at the end of the charging and discharging schedule before the scheduled date.

[0097] 2) Energy storage backup capacity constraints: The increase in energy storage reserve capacity should meet the energy storage capacity constraint, and the increase in energy storage reserve capacity should be less than or equal to the difference between the photovoltaic power generation capacity and the photovoltaic day-ahead forecast value. (20); (twenty one); In the above formula, This indicates that the energy storage arrangement will increase the reserve capacity under the ω-th day-ahead electricity price scenario; This indicates the maximum discharge power limit of the energy storage; This indicates the reserve arrangement status of energy storage under the day-ahead electricity price scenario at time t. A value of 1 indicates an upward adjustment of the reserve, which is used for discharging during the real-time phase, while a value of 0 indicates a downward adjustment of the reserve, which is used for charging during the real-time phase. This indicates the installed capacity of photovoltaics.

[0098] The reduction in energy storage reserve capacity should meet the energy storage capacity constraint, and the reduction in energy storage reserve capacity should be less than or equal to the day-ahead forecast value of photovoltaic power.

[0099] (twenty two); (twenty three); In the above formula, This represents the reduced reserve capacity of energy storage arrangement during time period t under the ω-th day-ahead electricity price scenario; This indicates the maximum charging power limit for energy storage; Taking into account the day-ahead charge / discharge plan and reserve capacity arrangement of energy storage, the remaining energy equation constraints and the upper and lower limits of the remaining energy are as follows: (twenty four); (25); In the above formula, This represents the remaining energy stored at time t after the reserve capacity has been allocated. This indicates the remaining energy stored at time t-1 before the current reserve capacity is allocated. After the scheduled charge / discharge plan and reserve capacity are set before the energy storage date, the remaining energy at the beginning and end times are equal: (26); 3) Real-time power constraint adjustment for energy storage: (27); (28); (29); (30); (31); In the above formula, This indicates whether the real-time actual output of the photovoltaic-storage system is lower than its day-ahead electricity bid amount when there is no real-time adjustment of energy storage, under the day-ahead electricity price scenario at time t and the ωth day-ahead. A value of 1 indicates yes, and a value of 0 indicates no. This indicates the maximum output limit of the photovoltaic power generation system; Represents the scenario at time t The actual output of photovoltaic power generation; , They represent the scenarios at time t ( , Real-time discharge and charging power of energy storage; This indicates that the energy storage arrangement will increase the reserve capacity during time period t; This indicates that the energy storage plan will reduce the reserve capacity during time period t.

[0100] 4) Real-time deviation power constraint of photovoltaic-storage system: After real-time energy storage adjustment, the deviation of the actual power generation of the photovoltaic-storage system in the real-time stage from the day-ahead tendered power energy needs to meet the following constraints: (32); (33); In the above formula, , They represent the scenarios at time t ( , After real-time energy storage adjustment, the actual output of the photovoltaic-storage system is higher or lower than the day-ahead electricity bid portion.

[0101] In step 4, the bidding-operation optimization model constructed in step 3 is solved using an optimization solver to obtain the optimal bidding amount, reserve capacity, and real-time energy storage power adjustment results for each time period under various electricity price scenarios.

[0102] In some embodiments, the commercial solver Cplex can be invoked in the MATLAB platform to solve the above mixed-integer linear programming model.

[0103] The specific process of solving the bidding-operation optimization model is as follows: First, continuous variables in the model, including charging and discharging power, need to be included. and Energy storage energy status and Deviation in charge, etc., and integer variables including state of charge and discharge. Backup arrangement status These should be clearly defined, and their upper and lower limits should be set; Secondly, the probability of combining electricity price scenarios with photovoltaic power output scenarios. , As a weighting factor, it is introduced into the objective function to reflect the expected returns under different scenarios; Next, the Cplex solver is used to call the branch and bound algorithm to solve the mixed-integer linear programming model. This algorithm first solves the relaxed linear programming subproblem to obtain an initial solution; then, it enumerates the integer variables through branch operations and prunes branches that do not meet the optimal conditions through bound operations, gradually approaching the optimal solution. During the solution process, parameters such as solution accuracy and maximum solution time are set according to actual needs to ensure solution efficiency and reliability.

[0104] After the solution is completed, Cplex will output the optimal decision variable values ​​for each time period under various electricity price scenarios. This yields the optimal bidding volume for each time period under various electricity price scenarios. standby capacity , and the power regulation results of real-time stage energy storage , .

[0105] Simulation results demonstrate that this method, while ensuring model solvability, can effectively coordinate the bidding behavior of photovoltaic-storage systems in the energy and backup markets, significantly improve the overall system revenue, and reduce the risk of real-time deviation.

[0106] Figure 5 The chart shows the bidding results for the day-ahead phase, with sub-figures (a) and (b) representing the day-ahead electricity market bidding results and the standby market bidding results, respectively. During the day-ahead phase, energy storage systems perform charging operations when day-ahead electricity market prices are low, reducing the amount of electricity the photovoltaic-storage system bids in the day-ahead electricity market. Conversely, when electricity prices are high, the energy storage system performs discharging operations to increase the amount of electricity the photovoltaic-storage system bids in that market, ultimately improving profitability.

[0107] Figure 6 This plan outlines the day-ahead charge / discharge schedule and reserve capacity arrangement for energy storage. When configuring energy storage, it provides increased reserve capacity during periods of higher reserve prices to reduce the reserve capacity purchased by the photovoltaic-storage system in the reserve market for the corresponding periods; and it provides decreased reserve capacity during periods of lower reserve prices to meet the time-series capacity constraints of energy storage throughout the day.

[0108] Figure 7The real-time operational results of energy storage include the real-time adjusted output of energy storage in sub-figure (a) and the total real-time output of energy storage in sub-figure (b). During the real-time phase, the energy storage system uses the deviation between the actual output of photovoltaic (PV) power and the day-ahead output forecast as the basis for judging its adjustment behavior. Under the constraint of day-ahead reserve arrangements, the energy storage system reduces the deviation penalty of the PV-storage system in the real-time market by implementing real-time power regulation operations. Furthermore, the total output of energy storage during the real-time phase equals the sum of the day-ahead charge / discharge plan of energy storage and the real-time adjusted output of energy storage.

[0109] The above analysis verifies that this method can effectively coordinate the bidding behavior of photovoltaic and energy storage systems in the energy and reserve markets, reduce real-time deviation penalties, and thus significantly improve the total revenue of photovoltaic and energy storage systems participating in electricity market bidding.

Claims

1. A joint bidding method for the photovoltaic-storage system energy-sustainability market that takes into account the uncertainties of photovoltaic power generation and electricity price, characterized in that... Includes the following steps: Step 1: Determine the bidding strategy for photovoltaic and energy storage systems in the day-ahead electricity market and the standby market, including the photovoltaic power output forecast, energy storage charging and discharging plan and standby capacity arrangement, and formulate an energy storage power adjustment scheme that can cope with the uncertainty of photovoltaic power output in the real-time stage; Step 2: Typical time-series photovoltaic output scenarios based on historical data are generated using Latin hypercube sampling and scenario reduction methods, and day-ahead electricity price scenarios are formed using multi-dimensional scenario generation methods. Step 3: Construct a day-ahead and real-time two-stage bidding-operation optimization model with the goal of maximizing total revenue. This model comprehensively considers electricity market revenue, standby market expenditure, and real-time deviation settlement, and also includes energy storage operation constraints and real-time adjustment constraints. Step 4: Solve the bidding-operation optimization model constructed in Step 3 to obtain the optimal bidding amount, reserve capacity, and real-time energy storage power adjustment results for each time period under various electricity price scenarios.

2. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 1, is characterized in that: In step 1, the day-ahead forecast value of photovoltaic power generation at time t is... The lower boundary of the uncertainty range of photovoltaic power generation The difference is defined as the maximum day-ahead forecast deviation for photovoltaic power generation, i.e. The energy price, the increase in reserve electricity price, and the decrease in reserve electricity price at time t are respectively expressed as follows: , , ; In the day-ahead energy-reserve market on the distribution side, photovoltaic power generators report day-ahead forecast information, including the day-ahead forecast value and the uncertainty range of photovoltaic power generation; the day-ahead forecast value falls within the envelope of the uncertainty range of photovoltaic power generation. ; Then the revenue from photovoltaic power generation at time t can be expressed as: Photovoltaic operators need to purchase reserve capacity in the distribution-side reserve market equal to the maximum daily forecast deviation of photovoltaic power generation. The reserve expenditure of photovoltaic power generators is... .

3. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 2, is characterized in that: In step 1, when bidding separately for photovoltaic power, the bid volume in the day-ahead electricity market is the day-ahead forecast value of photovoltaic power generation. The bid volume in the standby market represents the largest forecast deviation. ; When photovoltaics and energy storage participate in the market together, energy storage arranges day-ahead charging plans. and discharge plan Compared with the day-ahead forecast of photovoltaic power generation Together, they constitute the day-ahead electrical energy bidding information for the photovoltaic-storage system; In addition, energy storage arrangements have increased reserve capacity. and reduce reserve capacity Maximum prediction deviation Together, these constitute the backup purchase bidding information for the photovoltaic energy storage system.

4. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 3, is characterized in that: In step 1, during the real-time phase, the actual output of the photovoltaic-storage system... There is a deviation between the current day's electricity energy bid and the actual bid amount; this deviation will be denoted as... ; When the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system will be charged at a price lower than the day-ahead electricity price. Electricity sold; When the actual output of the photovoltaic and energy storage system is lower than the day-ahead electricity bid amount, i.e. At that time, the photovoltaic-storage system is priced higher than the day-ahead electricity price. The price will be charged, and punishment will be imposed. Constrained by the reserve capacity determined in the day-ahead market for energy storage, the real-time charging power of energy storage is adjusted. and real-time discharge power This reduces losses caused by deviations in the real-time phase of the photovoltaic-storage system; When energy storage is not adjusted in real time If the actual output of the photovoltaic storage system The bid amount exceeded the day-ahead electricity volume, while energy storage capacity was reduced during the day-ahead phase. ;like The real-time charging power of energy storage ;like The real-time charging power of energy storage ; If energy storage does not have a plan to reduce reserve capacity. In this case, energy storage will not be adjusted in real time; If the actual output of the photovoltaic storage system The amount was lower than the day-ahead electricity bid, while energy storage capacity was increased during the day-ahead phase. and The real-time discharge power of the stored energy ; like The real-time discharge power of the stored energy ; If energy storage does not plan to increase reserve capacity. In this case, energy storage will not be adjusted in real time.

5. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 4, is characterized in that: In step 2, the uncertainties faced by the bidding-operation optimization problem of the photovoltaic-storage system include the uncertainty of photovoltaic power generation and the uncertainty of electricity price; the uncertainty of photovoltaic power generation is described by uncertainty interval in the day-ahead stage and characterized by multi-scenario technology in the real-time stage. In the day-ahead phase, when the day-ahead forecast value of photovoltaic power generation is obtained... Then, based on the day-ahead forecast deviation of photovoltaic power generation... Historical data is used to obtain the uncertainty interval at a given confidence level.

6. The joint bidding method for the photovoltaic-storage system energy-sustainable market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 5, is characterized in that: First, collect the day-ahead forecasts for the same time period t over the past N days. Actual output of photovoltaic and energy storage systems Calculate the daily forecast deviation. And remove outliers; Secondly, the prediction bias before the fitting date probability distribution: Photovoltaic prediction biases typically follow a normal distribution; the distribution parameters can be estimated using historical bias data. ①. Calculate the day-ahead forecast deviation Expectations : （1）； ②. Calculate the day-ahead forecast deviation Standard deviation : （2）； The KS test is used to verify whether the deviation conforms to a normal distribution. If the test passes, the normal distribution is adopted. Otherwise, a t-distribution or a mixed distribution can be used; Next, based on the preset confidence level α, the corresponding quantiles are obtained from the quantile table to determine the interval boundaries: The range of quantiles for the deviation is determined as follows: ; and then combined with the previous day's forecast value The uncertainty range of photovoltaic output is derived as follows: ; Finally, adjust the interval: Because photovoltaic power output has a physical upper limit, namely "installed capacity" The interval derived needs to be truncated and corrected to include the lower limit "0". （3）。 7. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 6, is characterized in that: To model the uncertainties of photovoltaic (PV) power generation in the real-time phase, we first generate typical PV output scenarios for each time period based on the Latin hypercube sampling (LHS) method and scenario reduction technique. Then, we combine the single-time-period scenarios into typical time-series PV output scenarios through an iterative reduction and fusion method. The specific steps are as follows: First, random sampling of photovoltaic power output samples in the real-time stage is performed using Latin hypercube sampling (LHS). Secondly, a scenario reduction method was used to reduce the photovoltaic output samples obtained from each time period to identify several representative scenarios; specifically including: b1: Scene preprocessing; b2: Initialize parameter settings; b3: Inter-scene distance calculation; b4: Eliminate redundant scenes; Finally, the scenarios from each time period are stitched together to form a final set of typical photovoltaic output scenarios for each time period. The specific steps are: repeat the operation until R... t The number of scenes is To obtain representative scenarios .

8. The joint bidding method for the photovoltaic-storage system energy-reserve market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 7, is characterized in that: To address the uncertainty modeling of electricity prices, the day-ahead electricity price, the increase in reserve electricity price, and the decrease in reserve electricity price are all uncertain. Based on historical electricity price data, a multi-dimensional typical scenario generation method is used to generate day-ahead electricity price scenarios. The specific steps are as follows: c1: First, clean the historical electricity price data, remove outliers, and extract multidimensional features; c2: Using the weighted K-means clustering algorithm, the standardized electricity price feature matrix is ​​processed. Compression is performed to reduce the number of scenarios, and the differences between scenarios are characterized by weighted Euclidean distance to ensure that the clustering results conform to the weight of the impact of electricity price on photovoltaic and energy storage bidding. c3: After clustering converges, the final number of typical scenarios is determined through error verification, and the probability of occurrence of each scenario is calculated to form a set of electricity price scenarios that can be directly used for bidding optimization; Finally, t typical electricity price scenarios are generated, each scenario containing the specific values ​​of 3 types of electricity prices and their probability of occurrence.

9. The joint bidding method for the photovoltaic-storage system energy-standby market, taking into account the uncertainties of photovoltaic power generation and electricity price, as described in claim 8, is characterized in that: In step 3, the photovoltaic-storage system participates in day-ahead market bidding with the goal of maximizing total revenue and performs real-time phase optimization operation. The objective function of the day-ahead and real-time two-stage bidding-operation optimization model is expressed as: （14）； In equation (14): , These represent the probabilities for day-ahead electricity price scenarios and real-time photovoltaic power scenarios, respectively. , These are the corresponding scene numbers; Set as time interval ; , , These are the number of day-ahead electricity price scenarios, the number of day-ahead bidding periods, and the number of real-time photovoltaic power generation output scenarios, respectively. and These represent the electricity price and the increased reserve electricity price under the day-ahead electricity price scenario at time t; , These represent the ratios of the settlement price for the portion of electricity delivered by the photovoltaic-storage system that is greater than or less than the day-ahead tendered electricity, respectively, to the day-ahead electricity price. , ; and These are the portions of the electricity generated by the photovoltaic and energy storage system that are greater than and less than the day-ahead electricity bid. , Let represent the day-ahead discharge power plan and day-ahead charging power plan of the photovoltaic-storage system at time t and the ω-th day-ahead electricity price scenario, respectively. This represents the lower boundary value of the uncertainty interval of photovoltaic power generation output at time t; Equation (14) includes: Item 1 This indicates the revenue obtained by the photovoltaic and energy storage system from day-ahead electricity market bidding; Item 2 This indicates the expenditure on purchasing spare parts in the spare parts market for photovoltaic energy storage systems; Item 3 This indicates the revenue generated from the excess power generated when the actual output of the photovoltaic-storage system exceeds the day-ahead electricity bid during the real-time phase. Item 4 This indicates the penalty imposed on the under-generated electricity when the actual output of the photovoltaic-storage system is less than the day-ahead electricity bid during the real-time phase.

10. The joint bidding method for photovoltaic-storage system energy-standby market that takes into account the uncertainty of photovoltaic power generation and the uncertainty of electricity prices, as described in claim 9, is characterized in that: In step 4, the specific process of solving the bidding-operation optimization model is as follows: First, continuous variables in the model, including charging and discharging power, need to be included. and Energy storage energy status and Deviation in charge, etc., and integer variables including state of charge and discharge. Backup arrangement status These should be clearly defined, and their upper and lower limits should be set; Secondly, the probability of combining electricity price scenarios with photovoltaic power output scenarios. , As a weighting factor, it is introduced into the objective function to reflect the expected returns under different scenarios; Next, the mixed-integer linear programming model is solved by calling the branch and bound algorithm through the Cplex solver; After the solution is completed, Cplex will output the optimal decision variable values ​​for each time period under various electricity price scenarios; Obtain the optimal bidding volume for each time period under various electricity price scenarios. standby capacity , and the power regulation results of real-time stage energy storage , .

Images