Energy storage collaborative robust scheduling optimization method based on new energy field group joint sharing

Abstract

The invention provides an energy storage collaborative robust scheduling optimization method based on new energy field group joint sharing, and belongs to the technical field of energy storage collaborative robust scheduling optimization. The technical problem to be solved is to provide an energy storage collaborative robust scheduling optimization method based on new energy field group joint sharing. According to the technical scheme, a new energy field group joint sharing energy storage mode is adopted to carry out power transaction; weibull distribution and Beta distribution are adopted to describe probability density of natural wind speed and solar radiation intensity, and a predicted output scene set of wind and light in a day is generated by improving Latin hypercube sampling generation and K-means reduction; establishing a two-stage transaction framework rule of a day-ahead market stage and a real-time market stage; establishing transaction optimization models of a day-ahead market stage and a real-time market stage, and establishing a two-stage robust optimization model; the method is used for performing optimal scheduling on new energy and stored energy cooperative participation in the power market.

Description

Technical Field

[0001] The present invention provides an energy storage collaborative robust scheduling optimization method based on joint sharing of new energy field groups, belonging to the technical field of energy storage collaborative robust scheduling optimization. Background Art

[0002] With the continuous advancement of the use of renewable energy, the proportion of wind power and solar power generation has continued to increase in recent years, and their grid-connected power generation has participated in the electricity spot market. However, due to the strong uncertainty interference of wind and solar power, the grid connection of large-scale wind and solar power plants will pose a challenge to the security of the power grid. The introduction of regulatory resources such as energy storage can help new energy fields effectively resist uncertainty and reduce output fluctuations. Currently, most of the new energy supporting energy storage is independently built, with problems such as low resource utilization efficiency and small capacity. Against this background, the sharing economy theory has been introduced into the field of energy storage. Shared energy storage can utilize the complementarity of discharge and charging needs between different new energy units to fully improve resource utilization efficiency.

[0003] At present, the research on the cooperation between new energy and energy storage in the electricity market mainly focuses on the cooperation model between new energy and energy storage, uncertainty resolution methods, and cooperation benefit distribution. Some results have studied the participation of energy storage in the day-ahead electricity market or the real-time electricity market in the form of independence or aggregators, and some results have studied the trading mechanism of new energy and energy storage participating in multiple markets. However, it has failed to take into account the discussion of the multi-market interactive relationship between the day-ahead market, the real-time market and the green certificate market, and has failed to take into account the uncertainty of renewable energy output.

[0004] At the same time, the uncertainty of the output of distributed resources such as wind and solar power will cause deviations. How to effectively deal with this fluctuation factor is an important prerequisite for ensuring the safe and stable operation of the system. The processing methods commonly used in existing research are random optimization, scenario generation and robust optimization. Among them, random planning or scenario generation methods need to assume that the predicted power of renewable energy obeys a certain probability distribution and generate a large number of scenarios to fully characterize the uncertainty, but they all require deterministic probability curves to generate scenarios, which are difficult to obtain accurately. Robust optimization does not require pre-definition of probability distribution or generation of scenarios, but only focuses on the worst possible situation. However, making decisions under the worst situation rarely occurs in practice, and the high conservatism sacrifices a certain economic efficiency. Therefore, in order to overcome the problem of uncertainty in the output of distributed resources, it is necessary to combine scenario generation and robust optimization methods, use improved Latin hypercube sampling to generate uncertainty sets, and use two-stage robust optimization to obtain the optimal scheduling results on this basis to improve the conservatism of the robust method. Summary of the invention

[0005] In order to overcome the deficiencies in the prior art, the present invention aims to solve the technical problem of providing a method for optimizing energy storage coordination and robust scheduling based on joint sharing of new energy field groups.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups, including the following scheduling optimization steps:

[0007] Step 1: Use a new energy cluster joint shared energy storage model consisting of a new energy cluster entity and a shared energy storage entity to conduct electricity trading;

[0008] Step 2: Perform uncertainty processing, use Weibull distribution and Beta distribution to describe the probability density of natural wind speed and solar radiation intensity, and generate a set of wind and solar power forecast output scenarios within the day through improved Latin hypercube sampling and K-means reduction;

[0009] Step 3: Establish two-stage trading framework rules for the day-ahead market stage and the real-time market stage;

[0010] Step 4: Establish the transaction optimization model for the day-ahead market stage and the real-time market stage, and establish a two-stage robust optimization model, where:

[0011] In the transaction optimization model of the day-ahead market stage, based on the predicted power of WPP and PV, the day-ahead market bidding is carried out with the goal of maximizing the day-ahead market revenue;

[0012] In the transaction optimization model at the real-time market stage, the difference between the actual power generation of WPP and PV and the day-ahead bid power is considered, and a real-time adjustment plan is formulated with the goal of maximizing the real-time market profit;

[0013] Step 5: For the above two-stage robust optimization model, a column constraint generation algorithm is used to transform the above problem into a two-level optimization problem including a main problem and sub-problems, and the answer is solved by alternating the main problem and the sub-problems;

[0014] Step 6: Analyze the effect of robust scheduling optimization in various application scenarios.

[0015] In step 1, the new energy field group is specifically composed of multiple new energy fields, which are either wind farms or photovoltaic power stations.

[0016] New energy clusters can sell electricity in the spot market and green certificates in the green certificate market. It is defined that for every megawatt-hour of electricity produced by a new energy cluster, one green certificate will be obtained.

[0017] The shared energy storage entity includes physical energy storage PES and virtual energy storage VES.

[0018] The specific method for uncertainty processing in step 2 is:

[0019] Step 2.1: Conduct wind power uncertainty analysis:

[0020] The uncertainty of wind turbine output depends on the random characteristics of wind speed, which is described by Weibull distribution and expressed as:

[0021]

[0022] Where: v is the wind speed at any time, c is the scale parameter of the Weibull distribution, and k is the shape parameter, where k and c can be calculated based on the mathematical expectation and standard deviation of the wind speed sampling sequence samples;

[0023] Step 2.2: Perform optoelectronic uncertainty analysis:

[0024] The uncertainty of photovoltaic generator output depends on the random characteristics of solar radiation intensity, which is described by Beta distribution and expressed as:

[0025]

[0026] Where: r is the solar irradiance during period t, r max is the maximum solar irradiance during period t, α and β are the shape parameters of the Beta distribution, which are calculated from the expectation and variance of the solar radiation intensity;

[0027] Step 2.3: Use the improved Latin hypercube sampling method to simulate and generate uncertainty scenarios of actual wind and solar power output.

[0028] The specific method for simulating and generating the uncertainty scenario of the actual wind and solar power output in step 2.3 is:

[0029] Step 2.3.1: Sampling uncertainty scenarios:

[0030] Step 2.3.1.1: All historical wind and solar data are summarized by day and divided into 24 time periods with a span of hours, from 1h to 24h. Suppose any time period is i, i = 1, 2, ..., 24, then the wind speed parameter k of the i-th time period is i 、c i , and the solar radiation parameter α i , β i It can be calculated from the mean and standard deviation of wind speed and solar radiation samples;

[0031] Step 2.3.1.2: Based on the shape parameters of each period, the probability distribution functions of wind speed and solar radiation in each period are constructed according to Formula 1 and Formula 2, and the cumulative probability distribution model F is derived. k,i ;

[0032] Step 2.3.1.3: Accumulate the probability distribution model F for each period k,iLatin hypercube sampling is performed to generate multiple unrelated sample matrices. The sampling process is:

[0033] Assume that the sampling scale is N, and the cumulative probability distribution curve F derived in step 2.3.1.2 is k,i (X k,i ) The vertical axis is divided into N equally spaced, non-overlapping intervals. The range of the dependent variable of the cumulative probability distribution is set to [0,1], then the length of each interval is 1 / N. Then, completely random sampling is performed in each interval [(n-1) / N, n / N], and the vertical coordinate value drawn is defined as y kn,i , substitute the obtained ordinate value into the cumulative probability distribution function F k,i The inverse function of X k,i The nth sampling value of is expressed as:

[0034]

[0035] Step 2.3.2: Perform uncertainty scene reduction, specifically using the K-means algorithm to reduce the scene, the specific steps are:

[0036] Step 2.3.2.1: Set the number of iterations t = 1, the number of clusters to L, and randomly select the initial value J of each cluster center t (l), l = 1, 2, ..., L;

[0037] Step 2.3.2.2: Calculate each sample X k The Euclidean distance from L cluster centers, and classify the sample into the class where the cluster center with the smallest distance is located;

[0038] Step 2.3.2.3: Assume that the lth class has q samples, then the new cluster center J t+1 (l) is:

[0039]

[0040] Step 2.3.2.4: Repeat steps 2.3.2.2 and 2.3.2.3 until the position change of the centroid is less than the set threshold or the maximum number of iterations is reached;

[0041] Step 2.3.2.5: Output the final clustering results and the probability of each scene, where the probability of the lth scene is calculated as:

[0042]

[0043] In step 3, the trading framework rules for the day-ahead market phase are established as follows:

[0044] Each member of the new energy cluster provides the trading center with bidding power and quotation for the next 24 hours based on the predicted wind power and photovoltaic power generation. Each member submits a bid at a low electricity price to achieve the goal of winning the bid for all bidding power, and finally obtains power generation profits at the unified market clearing price.

[0045] The trading framework rules for the real-time market stage are as follows:

[0046] The members of the new energy farm group know the winning bid power and actual power generation, and make the following adjustments to the resulting deviations to reduce the deviation penalty costs:

[0047] In the case where the deviation between the actual output of some members and the bid output is positive and the deviation of some members is negative at the same time, the part that offsets the positive and negative deviations is directly connected to the grid through virtual energy storage, and the remaining deviation is smoothed by calling physical energy storage. If the shared energy storage is not enough to make up for the deviation, the new energy e-commerce company will be punished for the deviation. The positive deviation part will be sold in the real-time market, and the negative deviation part needs to be purchased from other power entities to smooth the deviation.

[0048] When there is sufficient electricity generated by renewable energy, the excess electricity can be purchased through the energy storage system;

[0049] When the power generation from renewable energy sources is insufficient, the stored electricity can be sold to earn a profit from the price difference.

[0050] The specific method of establishing the transaction optimization model in the day-ahead market stage in step 4 is:

[0051] New energy clusters and shared energy storage collaborate to participate in the day-ahead market to maximize their joint net benefits. Wind farms and photovoltaic power stations forecast the wind and solar output of the next day and report the electricity volume and price based on historical electricity price information:

[0052] Among them, the expression of the day-ahead market objective function is:

[0053]

[0054] Where: The benefits of the new energy cluster participating in the day-ahead market, are the daily average costs of the new energy cluster and shared energy storage, respectively, where RE = {WPP, PV};

[0055]

[0056] Where: T is 24; is the clearing electricity price in the day-ahead market at time t; N is the number of members in the new energy cluster; are the day-ahead market bid output and day-ahead output forecast of the new energy field i at time t; Δt is the time interval; The operating cost and depreciation cost of the new energy cluster respectively; i is the unit annual operation and maintenance cost of the new energy field i; RE,i is the unit annual investment cost of new energy field i; P i max is the investment power of new energy field i; r is the interest rate; γ RE,i is the depreciation period of the new energy field i;

[0057] The constraints of the day-ahead market objective function include:

[0058] The output constraint of new energy cluster members is expressed as:

[0059]

[0060] Virtual energy storage constraints:

[0061] Virtual energy storage does not need to consider the charging and discharging efficiency during operation, so the expression for the dynamic change of battery energy storage level in each period is:

[0062]

[0063] Where: and are the energies stored in the virtual energy storage at time t and time t+1 respectively; are the charging value and discharging value of the virtual energy storage at time t respectively;

[0064] The virtual energy storage is set to half of the maximum total power of the new energy cluster, and the expression is:

[0065]

[0066] Where: are the maximum values ​​of charging and discharging of virtual energy storage, respectively; It is the charging and discharging 0-1 state variable of the virtual energy storage;

[0067] The expression of the amount of electricity stored at time 0 of virtual energy storage is:

[0068]

[0069] Physical energy storage constraints:

[0070] Physical energy storage needs to consider the charging and discharging efficiency during operation. The expression for the dynamic change of battery energy storage level in each period is:

[0071]

[0072] Where: and are the energies stored in the physical energy storage at time t and time t+1 respectively; are the charging value and releasing value of the physical energy storage at time t respectively;

[0073] To truly reflect the operation strategy, the battery storage level at the beginning of each day should remain consistent, expressed as:

[0074]

[0075] In addition, physical energy storage needs to consider the following constraints:

[0076]

[0077]

[0078] Where: They are the maximum values ​​of physical energy storage charging and discharging respectively; Indicates the charge and discharge state of physical energy storage. When the value is 1, it indicates discharge, and when the value is 0, it indicates charge. and The minimum or maximum remaining capacity allowed, respectively;

[0079] Power balance constraints:

[0080] The expression that satisfies power balance is:

[0081]

[0082] The specific method of establishing the transaction optimization model in the real-time market stage in step 4 is:

[0083] The new energy cluster knows its own winning bid volume and winning bid price in the day-ahead market, and requires to reduce the deviation through shared energy storage and backup services to ensure the maximization of its own net profit:

[0084] The expression of the real-time market objective function is:

[0085]

[0086] Where: To enable new energy clusters to collaboratively share the benefits of energy storage participation in real-time markets; Proceeds from selling green certificates for new energy clusters; is the deviation penalty cost, the cost of purchasing ancillary services for the system from the electricity market;

[0087]

[0088] Where: is the clearing electricity price in the real-time market at time t; It is the additional electricity sold by the new energy site i in the real-time market stage; is the actual output of new energy field i at time t; green is the green certificate conversion factor, 1 megawatt of green electricity is converted into 1 green certificate; p green The price of green certificates;

[0089]

[0090] Where: is the amount of electricity purchased by new energy farm i in the real-time market stage; δ1 and δ2 are the penalty coefficients for positive and negative deviations in the real-time market respectively;

[0091] The constraints of the real-time market objective function include:

[0092] The power balance constraint is expressed as:

[0093]

[0094] The state constraints of electricity sales and purchase of auxiliary services are expressed as:

[0095]

[0096] Where: Indicates the purchase and sale status of the system. A value of 1 indicates the purchase of auxiliary services, and a value of 0 indicates the sale of electricity.

[0097] The specific method for establishing the two-stage robust optimization model in step 4 is:

[0098] Without considering the uncertainty of wind and solar power output, a deterministic optimization model for the participation of new energy clusters in the electricity market through collaborative shared energy storage is obtained. The compact form of the model is expressed as:

[0099]

[0100] In the formula, x and y are the decision variables of the first and second stages, respectively, and their expressions are:

[0101]

[0102] In order to make the model conform to the two-stage robust expression, the objective function of the first stage is negated and converted to the minimum to satisfy the expression:

[0103] a T x=-R ahead Formula 32;

[0104] In order to solve the uncertainty problem of wind and solar power, robust optimization is performed. In order to achieve the maximum value of the second-stage objective function, the max of the second stage and the max of the uncertainty set U are merged into one, so that the expression of the abstract constraint model of the two-stage robust optimization is:

[0105]

[0106] Where: a, b are the coefficient column vectors corresponding to the objective function expressions 5, 22 of the day-ahead and real-time stages, respectively; D, Q, K, R, S, M, L, Y are the coefficient matrices of the variables under the corresponding constraints; d, k, h are constant column vectors;

[0107] Where: The first and second rows of the constraints represent the inequality constraints in the model, where:

[0108] The first row includes formulas 12, 13, 18, 19, 20, 28, and 29;

[0109] The second row includes formula 10;

[0110] The third row represents the equality constraints in the model, including equations 11, 16, and 17;

[0111] The 4th row corresponds to equation 27;

[0112] The 5th row corresponds to equation 21;

[0113] Among them, the fluctuation range of each new energy cluster output is determined by the typical scenarios extracted. To reduce the output of the new energy field i at time t in the post-scenario j, the calculation formula for the maximum fluctuation range of the output of the new energy field i is:

[0114]

[0115] Where: is the fluctuation deviation at time t;

[0116] The maximum fluctuation value of the output of the new energy field group is obtained through the maximum fluctuation range calculation formula, and then the uncertainty of the output of the new energy field group is expressed as the uncertainty set U, which is expressed as:

[0117]

[0118] Where: Γ is the robustness coefficient, is the upper limit of the renewable energy output under consideration, P t RE0 To provide a lower limit for the new energy sources under consideration;

[0119] Since the output of the new energy field will face a greater deviation penalty cost when it takes the minimum value of the interval, the adjustment parameter L is introduced to adjust Equation 32 to the following form:

[0120]

[0121] Where: It is a binary variable. When it takes the value of 1, the uncertain variable of the corresponding period reaches the boundary of the interval.

[0122] The specific method for solving the answer of the two-stage robust optimization model in step 5 is:

[0123] By decomposing the expression 29 of the state constraints of electricity sales and purchase of auxiliary services, the expression of the main problem obtained at the kth iteration of the two-stage robust model is:

[0124]

[0125] Where: θ is the auxiliary variable for real-time market stage profit in the MP objective function; l is the current iteration number; y l is the solution of the subproblem after the lth iteration; is the value of the uncertain variable u in the worst scenario determined after the lth iteration;

[0126] The expression of the sub-problem is:

[0127]

[0128] After the above transformation, the two-stage robust model is decoupled into the main problem 33 and the sub-problem 34. Both problems are mixed integer linear programming problems, and then solved using the column constraint generation algorithm. The specific process is as follows:

[0129] Set the lower bound of the objective function to LB = -∞ and the upper bound to UB = +∞;

[0130] In the worst-case scenario forecast * Next, solve the main problem equation 33 and get the optimal solution The optimal solution of the main problem is taken as the new lower bound, that is

[0131] Solve the main problem Substitute into subproblem equation 34 and get the optimal solution of the subproblem Update upper bound

[0132] Assume that the algorithm convergence threshold is ε. If UB-LB≤ε, stop the calculation and return the optimal solution. Otherwise, let k=k+1 and go back to solve the main problem equation 33 again.

[0133] The beneficial effects of the present invention compared with the prior art are as follows: the present invention combines variable renewable energy sources such as wind and solar to form a new energy cluster to participate in the electricity spot market, designs a cooperation model for the new energy cluster and shared energy storage to jointly participate in the electricity market, and combines the advantages of scenario generation and robust optimization to propose an improved day-ahead-real-time two-stage robust optimization model of uncertainty interval. In the day-ahead stage, based on the predicted power of WT and PV, with the goal of maximizing the day-ahead market revenue, day-ahead market bidding is carried out. In the second stage, considering the difference between the actual power generation of WT and PV and the day-ahead winning power, a real-time adjustment plan is formulated with the goal of maximizing the real-time market revenue. The present invention is conducive to the consumption and effective utilization of renewable energy and the clean and low-carbon transformation of the overall energy structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0134] The present invention will be further described below in conjunction with the accompanying drawings:

[0135] Figure 1 This is a structural topology diagram of the new energy field group and shared energy storage participating in the electricity-green certificate market of the present invention;

[0136] Figure 2 It is a structural diagram of the two-stage transaction framework of the present invention;

[0137] Figure 3 A flowchart of the steps for solving the two-stage robust model of the present invention;

[0138] Figure 4 A statistical diagram of predicted / actual output of a wind farm group in an embodiment of the present invention;

[0139] Figure 5 A statistical diagram of predicted / actual output of a photovoltaic power station group in an embodiment of the present invention;

[0140] Figure 6 This is a schematic diagram of transaction results under different robustness coefficients of scenarios 2 to 4 in an embodiment of the present invention;

[0141] Figure 7 Schematic diagram of cluster scheduling in different scenarios in an embodiment of the present invention;

[0142] Figure 8 Schematic diagram comparing the energy storage operation conditions of scenarios 3 and 4 in an embodiment of the present invention. DETAILED DESCRIPTION

[0143] The present invention designs a cooperation model for new energy field groups and shared energy storage to jointly participate in the day-ahead market, real-time market and green certificate market, and combines the advantages of scenario generation and robust optimization to propose a day-ahead-real-time two-stage robust optimization model with improved uncertainty interval, which is conducive to the absorption and effective utilization of renewable energy, maximization of energy utilization and market benefits, and ultimately the clean and low-carbon transformation of the overall energy structure; for the characterization of wind and solar power output uncertainty, based on existing research, the present invention adopts Weibull distribution and Beta distribution to describe the probability density of natural wind speed and solar radiation intensity, and further reduces the generated wind and solar power forecast output scenario set within the day by improving Latin hypercube sampling generation and K-means algorithm to achieve robust scheduling optimization.

[0144] The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups provided by the present invention specifically includes the following optimization steps:

[0145] Step 1: Determine the new energy cluster and shared energy storage trading model.

[0146] The new energy cluster joint shared energy storage mode provided by the present invention overcomes the problem of non-sharing of energy storage between new energy clusters in the traditional independent energy storage mode, so that it is possible to use the energy storage of other new energy clusters to store excess electricity, and it is also possible to call virtual energy storage to offset the opposite energy storage demand, so as to maximize the overall energy utilization. Compared with the centralized dispatching mode, the NESs-SES (electricity-green certificate market) mode is a cooperative group spontaneously formed by new energy clusters and shared energy storage to trade electricity, without the need to transmit information to the centralized control center, which is conducive to reducing the burden of power grid dispatching and reducing the risk of information leakage. Figure 1 Shown is a schematic diagram of new energy clusters and shared energy storage participating in the electricity-green certificate market.

[0147] The joint shared energy storage model includes two main parts: new energy cluster and shared energy storage.

[0148] A new energy cluster is composed of multiple new energy clusters, which can be wind farms or photovoltaic power stations. A new energy cluster can sell electricity in the spot market and sell green certificates in the green certificate market, reflecting the dual economic and environmental characteristics of green electricity. For every megawatt-hour of electricity produced by a new energy cluster, it will receive a green certificate. Green certificates can be sold through the green certificate subscription platform. Enterprises that have not reached renewable energy quotas can choose to purchase green certificates to offset their quotas and reduce or avoid penalties.

[0149] Shared energy storage includes physical energy storage (PES) and virtual energy storage (VES). Among them, virtual energy storage can make full use of the complementary discharge behavior between new energy fields to offset part of the physical energy storage call demand, so that physical energy storage has a higher resource utilization rate. In actual scheduling, if there is a situation where the charging and discharging demands of a new energy field group coexist at a certain moment, the part of the charge and discharge quantity that offsets each other can be directly connected to the grid through virtual energy storage, without using physical energy storage and generating energy storage operation losses, and the remaining deviation is smoothed by calling physical energy storage.

[0150] Step 2: Deal with uncertainty.

[0151] The present invention adopts Weibull distribution and Beta distribution to describe the probability density of natural wind speed and solar radiation intensity, and generates a set of predicted wind and solar power output scenarios within a day by improving Latin hypercube sampling generation and K-means reduction. The specific method is as follows:

[0152] Step 2.1: Conduct wind power uncertainty analysis:

[0153] The uncertainty of wind turbine output depends on the random characteristics of wind speed, which is generally described by Weibull distribution, and its function expression is as follows:

[0154]

[0155] Where: v is the wind speed at any time; c is the scale parameter of the Weibull distribution, and k is the shape parameter. k and c can be calculated from the mathematical expectation and standard deviation of the wind speed sampling sequence.

[0156] Step 2.2: Perform optoelectronic uncertainty analysis:

[0157] The output uncertainty of photovoltaic generators depends on the random characteristics of solar radiation intensity, which can usually be described by Beta distribution, expressed as:

[0158]

[0159] Where: r is the solar irradiance during period t; r max is the maximum solar irradiance during period t; α and β are the shape parameters of the Beta distribution, which can be calculated from the expectation and variance of the solar radiation intensity.

[0160] Step 2.3: Generate uncertainty scenarios:

[0161] Although the traditional Latin hypercube sampling method can achieve full coverage of the sample interval and significantly reduce the aggregation phenomenon compared with the Monte Carlo simulation, it does not take into account the time series characteristics of the actual wind and solar power output in a day, resulting in all samples being obtained from the wind and solar power output model with the same shape parameters, which cannot reflect the time characteristics of wind and solar power. Based on this, the present invention proposes an improved Latin hypercube sampling method to simulate the uncertainty of the actual wind and solar power output. The specific method is:

[0162] Step 2.3.1: Sampling of uncertainty scenarios:

[0163] The present invention collects statistical data in different time periods based on the traditional Latin hypercube sampling method, and then constructs a cumulative probability distribution model with different shape parameters, so as to perform Latin hypercube sampling and obtain wind and solar output samples with daily characteristics. The specific method is as follows:

[0164] Step 2.3.1.1: All historical wind and solar data are summarized by day and divided into 24 time periods with a span of hours, from 1h to 24h. Let any time period be i, i = 1, 2, ..., 24. Then the wind speed parameter k in the i-th time period is i 、c i , and the solar radiation parameter α i , β i It can be calculated from the mean and standard deviation of wind speed and solar radiation samples.

[0165] Step 2.3.1.2: Based on the shape parameters of each period, the probability distribution functions of wind speed and solar radiation in each period are constructed according to formula (1) and formula (2), and the cumulative probability distribution model F is derived. k,i .

[0166] Step 2.3.1.3: Accumulate the probability distribution model F for each period k,i Latin hypercube sampling is performed to generate multiple unrelated sample matrices. The sampling process is:

[0167] Assume that the sampling scale is N, and the cumulative probability distribution curve F derived in step 2.3.1.2 is k,i (X k,i )The vertical axis is divided into N equally spaced, non-overlapping intervals. Since the range of the dependent variable of the cumulative probability distribution is [0,1], the length of each interval is 1 / N. Then, a completely random sampling is performed in each interval [(n-1) / N, n / N], and the vertical coordinate value drawn is set to y kn,i Substitute the obtained ordinate value into the cumulative probability distribution function F k,i The inverse function of X k,i The nth sampling value of is expressed as:

[0168]

[0169] Step 2.3.2: Uncertain scene reduction, using K-means algorithm to reduce the scene, the specific steps are:

[0170] Step 2.3.2.1: Set the number of iterations t = 1, the number of clusters to L, and randomly select the initial value J of each cluster center t (l), l=1,2,…,L.

[0171] Step 2.3.2.2: Calculate each sample X k The Euclidean distance from L cluster centers is calculated, and the sample is classified into the class where the cluster center with the smallest distance is located.

[0172] Step 2.3.2.3: Assume that the lth class has q samples, then the new cluster center J t+1 (l) is:

[0173]

[0174] Step 2.3.2.4: Repeat steps 2.3.2.2 and 2.3.2.3 until the change in the centroid position is less than the set threshold or the maximum number of iterations is reached.

[0175] Step 2.3.2.5: Output the final clustering results and the probability of each scene. The probability of the lth scene is:

[0176]

[0177] Step 3: Establish a two-stage trading framework model for the day-ahead and real-time electricity market.

[0178] like Figure 2 As shown, the present invention proposes a trading framework rule for NESs-SES alliance cooperation to participate in the day-ahead-real-time electricity market. In this framework, the transaction includes two stages: the day-ahead market and the real-time market.

[0179] Step 3.1: Day-ahead market stage: Each member of the new energy cluster provides the trading center with bidding power and quotation for the next 24 hours based on the predicted wind power and photovoltaic power generation. Since the marginal cost of the output of new energy units is low, each member can submit a bid at a lower electricity price to achieve the goal of winning the bid for all bidding power, and finally obtain power generation profits at the unified market clearing price.

[0180] Step 3.2: Real-time market stage: The members of the new energy field group know the winning bid and actual power generation, and can make certain adjustments to the resulting deviations to reduce the deviation penalty costs. Due to the uncertainty of wind and solar power, at the same time, the deviation between the actual output of some members and the winning bid will be positive, and the deviation of some members will be negative. At this time, the part where the positive and negative deviations offset each other can be directly connected to the grid through virtual energy storage, and the remaining deviation can be smoothed by calling physical energy storage. If the shared energy storage is not enough to make up for the deviation, the new energy e-commerce company will be punished for the deviation. The positive deviation part can be sold in the real-time market, and the negative deviation part needs to be purchased from other power entities to smooth out the deviation. In addition, when the new energy power generation is sufficient, the electricity price is low, and the excess electricity can be purchased through the energy storage system; when the new energy power generation is insufficient, the electricity price is high, and the stored electricity will be sold to obtain the price difference profit.

[0181] Step 4: Establish a two-stage day-ahead-real-time robust optimization model.

[0182] In the day-ahead stage, based on the predicted power of WPP and PV, day-ahead market bidding is carried out with the goal of maximizing the day-ahead market revenue. In the real-time stage, taking into account the difference between the actual power generation of WPP and PV and the day-ahead winning power, a real-time adjustment plan is formulated with the goal of maximizing the real-time market revenue.

[0183] Step 4.1: Establish a day-ahead market transaction optimization model.

[0184] New energy clusters and shared energy storage collaborate to participate in the day-ahead market to maximize their joint net benefits. Wind farms and photovoltaic power stations forecast the wind and solar output of the next day and report the electricity volume and price based on historical electricity price information. The objective function of the day-ahead market is:

[0185]

[0186] Where: The benefits of new energy clusters participating in the day-ahead market; They are the average daily costs of the new energy cluster and shared energy storage, respectively, where RE = {WPP, PV}.

[0187]

[0188] Where: T is 24; is the clearing electricity price in the day-ahead market at time t; N is the number of members in the new energy cluster; are the day-ahead market bid output and day-ahead output forecast of the new energy field i at time t; Δt is the time interval; The operating cost and depreciation cost of the new energy cluster respectively; i is the unit annual operation and maintenance cost of the new energy field i; RE,iis the unit annual investment cost of new energy field i; P i max is the investment power of new energy field i; r is the interest rate; γ RE,i is the depreciation period of new energy field i.

[0189] The calculation method of shared energy storage cost is similar to that of new energy clusters and will not be repeated here.

[0190] The day-ahead market objective function as a constraint condition of the transaction optimization model includes:

[0191] 1) Output constraints of new energy cluster members:

[0192]

[0193] 2) Virtual energy storage constraints:

[0194] Virtual energy storage does not need to consider the charging and discharging efficiency during operation, so the dynamic changes of battery energy storage levels in each period can be expressed as:

[0195]

[0196] Where: and are the energies stored in the virtual energy storage at time t and time t+1 respectively; are the charging value and discharging value of the virtual energy storage at time t respectively.

[0197] Since virtual energy storage does not rely on physical equipment, its maximum charging and discharging power can be infinite in theory. For the convenience of expression, it is set to half of the maximum total power of the new energy field group (because virtual energy storage is only called when the positive and negative outputs offset each other), and the expression is:

[0198]

[0199] Where: are the maximum values ​​of virtual energy storage charging and discharging respectively; U t VES It is the charging and discharging 0-1 state variable of the virtual energy storage.

[0200] In addition, since the charge and discharge amount of the virtual energy storage is equal at each moment, its storage capacity does not change within a time period. For the convenience of expression and calculation, the storage capacity of the virtual energy storage at time 0 is expressed as:

[0201]

[0202] 3) Physical energy storage constraints:

[0203] Physical energy storage needs to consider the charging and discharging efficiency during operation. The dynamic changes of battery energy storage levels in different time periods can be expressed as:

[0204]

[0205] Where: and are the energies stored in the physical energy storage at time t and time t+1 respectively; They are respectively the charging value and releasing value of the physical energy storage at time t.

[0206] In order to make the selected typical day more realistically reflect the annual operation strategy of the NESs-SES Alliance, the battery storage level at the beginning of each day should be consistent, which can be expressed as:

[0207]

[0208] In addition, physical energy storage needs to consider the following constraints:

[0209]

[0210] Where: They are the maximum values ​​of physical energy storage charging and discharging respectively; Indicates the charge and discharge state of physical energy storage. When the value is 1, it indicates discharge, and when the value is 0, it indicates charge. and The minimum / maximum remaining capacity allowed, respectively.

[0211] 4) Power balance constraint: satisfy the expression:

[0212]

[0213] Step 4.2: Build a real-time market transaction optimization model.

[0214] In the real-time market, the new energy cluster already knows its own winning bid volume and winning bid price in the market the day before. At this time, its main goal is to reduce deviations through shared energy storage, backup services, etc., to ensure the maximization of its own net profit.

[0215] The real-time market objective function is as follows:

[0216]

[0217] Where: To enable new energy clusters to collaboratively share the benefits of energy storage participation in real-time markets; Proceeds from selling green certificates for new energy clusters; is the deviation penalty cost, The cost of purchasing ancillary services for the system from the electricity market.

[0218]

[0219] Where: is the clearing electricity price in the real-time market at time t; It is the additional electricity sold by the new energy site i in the real-time market stage; is the actual output of the new energy field i at time t. ξ green is the green certificate conversion factor, 1 megawatt of green electricity is converted into 1 green certificate; p green The price of green certificate.

[0220]

[0221] Where: is the amount of electricity purchased by new energy farm i in the real-time market stage; δ1 and δ2 are the penalty coefficients for positive and negative deviations in the real-time market, respectively.

[0222] The real-time market objective function as a constraint condition of the transaction optimization model includes:

[0223] (1) Power balance constraint, expressed as:

[0224]

[0225] (2) The state constraints of electricity sales and purchase of auxiliary services are expressed as:

[0226]

[0227] Where: Indicates the purchase and sale status of the system. When the value is 1, it indicates the purchase of auxiliary services, and when the value is 0, it indicates the sale of electricity. In addition, the real-time market also needs to face the charging and discharging constraints of energy storage units, as shown in formulas (16) to (20).

[0228] Step 4.3: Establish a two-stage robust optimization model.

[0229] When the uncertainty of wind and solar power output is not considered, a deterministic optimization model for the participation of renewable energy clusters in the electricity market through collaborative shared energy storage can be obtained, which can be expressed in a compact form as follows:

[0230]

[0231] x and y are the decision variables of the first and second stages respectively. The specific expressions are:

[0232]

[0233] In order to make the model conform to the general expression of two-stage robustness, the objective function of the first stage is transformed into the minimum by taking the negative, that is:

[0234] a T x=-R ahead (32);

[0235] In order to solve the problem of wind and solar uncertainty, a robust optimization method is introduced. Since the objective function of the second stage is to obtain the maximum value, the max of the second stage and the max of the uncertainty set U are combined into one. The abstract constraint model of the two-stage robust optimization is as follows:

[0236]

[0237] Where: a, b are the coefficient column vectors corresponding to the objective functions (5) and (22) in the day-ahead and real-time stages, respectively; D, Q, K, R, S, M, L, Y are the coefficient matrices of the variables under the corresponding constraints; d, k, h are constant column vectors. In equation (33), the first and second rows of the constraints represent the inequality constraints in the model, where the first row includes equations (12)-(13), (18)-(20), (28)-(29); the second row includes equation (10); the third row represents the equality constraints in the model, including equations (11), (16)-(17); the fourth row corresponds to equation (27); and the fifth row corresponds to equation (21).

[0238] The fluctuation range of the output of each renewable energy cluster is determined by the typical scenarios extracted. To reduce the output of the new energy field i at time t in the post-scenario j, the maximum fluctuation range of the output of the new energy field i is:

[0239]

[0240] Where: is the fluctuation deviation at time t.

[0241] The maximum fluctuation value of the output of the renewable energy field group is obtained through the above formula, and then the uncertainty of the output of the renewable energy field group is expressed as the uncertainty set U:

[0242]

[0243] Where: Γ is the robustness coefficient, which is used to adjust the conservativeness of the uncertainty set. The larger the value, the more severe the output fluctuation of the new energy unit is, and the more robust the dispatch result is. is the upper limit of the renewable energy output under consideration, P t RE0 To provide lower limit for the new energy that is being considered.

[0244] When the output of the new energy field reaches the minimum value of the interval, it will face a greater deviation penalty cost, which is more in line with the definition of the "worst" scenario. Therefore, the adjustment parameter L is introduced to rewrite formula (32) into the following form:

[0245]

[0246] Where: It is a binary variable. When it takes the value of 1, the uncertain variable of the corresponding period reaches the boundary of the interval.

[0247] Step 5: Solve the two-stage model. The algorithm steps of the two-stage robust optimization model used in the present invention are as follows: Figure 3 As shown:

[0248] For the above two-stage robust optimization model, the present invention adopts a column constraint generation algorithm (C&CG) to transform the above problem into a two-level optimization problem including a main problem and sub-problems, and solves the answer by alternating and iterating the main problem and sub-problems.

[0249] Decomposing equation (29), the main problem form obtained at the kth iteration of the two-stage robust model is:

[0250]

[0251] Where: θ is the auxiliary variable for real-time market stage profit in the MP objective function; l is the current iteration number; y l is the solution of the subproblem after the lth iteration; It is the value of the uncertain variable u in the worst scenario determined after the lth iteration.

[0252] Since the objective function of the subproblem is to find the maximum value, the subproblem does not need to be converted into a dual problem. The resulting subproblem is in the form of:

[0253]

[0254] After the above transformation, the two-stage robust model is decoupled into the main problem (33) and the sub-problem (34), both of which are mixed integer linear programming problems. The C&CG algorithm is then used to solve the problem, and the process is as follows:

[0255] 1) Set the lower bound of the objective function to LB = -∞ and the upper bound UB = +∞.

[0256] 2) In the worst case scenario u * Next, solve the main problem equation (33) and get the optimal solution The optimal solution of the main problem is taken as the new lower bound, that is:

[0257] 3) Solve the main problem Substitute into the subproblem equation (34) and obtain the optimal solution of the subproblem Update upper bound

[0258] 4) Set the algorithm convergence threshold as ε. If UB-LB≤ε, stop the calculation and return the optimal solution. Otherwise, let k=k+1 and go to step 2).

[0259] Finally, in order to demonstrate the economic feasibility of new energy clusters and shared energy storage in collaboratively participating in the electricity market and the help of robust thinking in the participation of new energy clusters in the market, the following four scenarios are set up for analysis.

[0260] Scenario 1: Each renewable energy field participates in the electricity market independently, and the uncertainty of wind and solar power is not considered.

[0261] Scenario 2: Each renewable energy field participates in the electricity market independently and uses robust optimization methods to deal with wind and solar uncertainties.

[0262] Scenario 3: The renewable energy cluster builds its own energy storage to participate in the electricity market, that is, each renewable energy cluster configures energy storage for 5% of its installed capacity and 2 hours, and energy storage is not shared between different renewable energy clusters. Robust optimization methods are used to handle wind and solar uncertainties.

[0263] Scenario 4: New energy clusters and shared energy storage jointly participate in the electricity market, that is, the entire new energy cluster is equipped with energy storage at 5% of the total installed capacity and 2 hours, and energy storage is shared among different new energy clusters. Robust optimization methods are used to deal with wind and solar uncertainties.

[0264] This invention selects a new energy base as the background to verify the effectiveness of the proposed model and method. The base is equipped with 3 wind farms (W1, W2, W3) and 2 photovoltaic power stations (PV1, PV2), of which the installed capacity of the wind farms is 100MW, 150MW, and 200MW respectively, and the installed capacity of the photovoltaic power stations is 100MW and 200MW respectively. Each new energy field is equipped with shared energy storage at 5% of the installed capacity and 2 hours. The price of green certificates refers to the average monthly price of the green certificate subscription platform, which is 45 yuan per certificate. The other relevant parameters are shown in the following table:

[0265]

[0266] Table 1 Related parameters

[0267] Since there is currently no spot market in the region, the typical electricity price curve of a certain area’s spot market is used as the simulated electricity price, with the power shortage penalty coefficient set to 4 and the energy abandonment penalty coefficient set to 0.25.

[0268] Based on the wind and solar uncertain scene sampling method proposed in this invention, 500 wind and solar output scenes are generated by Latin hypercube sampling, and 10 typical wind and solar output scenes are obtained by scene reduction method. The scene with the highest probability is selected as the predicted wind and solar output for day-ahead transactions, and the scene with the second highest probability is selected as the actual wind and solar output. After obtaining the wind and solar uncertainty set, the predicted / actual output curve of the new energy field group can be drawn, such as Figure 4 and Figure 5As shown in the figure. The solid lines represent the actual output of each new energy field, and the shaded part corresponding to the solid line color is the uncertainty set of the predicted output of each new energy field when the robustness coefficient is 0.5. It can be seen that when the value of Γ is less than 1, the predicted output range will not be able to completely cover the actual output curve.

[0269] When calculating, we take the robustness coefficient = 0.5, L = 6, and calculate the transaction results under different scenarios. Since scenarios 2 to 4 use the robust optimization method, in order to determine the most suitable robustness coefficient for each scenario, a sensitivity analysis is performed on the robustness coefficient under scenarios 2 to 4. The results are as follows: Figure 6 As shown in the figure, the solid marked points on the broken line are the extreme points.

[0270] from Figure 6 It can be seen that with the increase of robustness coefficient, the negative deviation penalty of the new energy field group in each scenario gradually decreases, and the positive deviation penalty gradually increases. Among them, in scenario 2, when Γ≥1.0, the negative deviation penalty value tends to be stable at 274,900 yuan; in scenario 3, when Γ≥0.9, the negative deviation penalty value tends to be stable at 67,900 yuan; in scenario 4, when Γ≥0.4, the negative deviation penalty value tends to be stable at 11,300 yuan. At the same time, with the gradual increase of robustness coefficient, the benefits of the cluster in each scenario always show a trend of rising first and then falling, and reach a maximum value between [0.4,1]. Among them, scenario 2 achieves the maximum value at Γ=1, which is 2.6088 million yuan; scenario 3 achieves the maximum value at Γ=0.9, which is 2.7832 million yuan; scenario 4 achieves the maximum value at Γ=0.4, which is 2.8775 million yuan.

[0271] Comparing scenarios 2 to 4, we can see that with the addition of independent energy storage and shared energy storage, the cluster's ability to cope with uncertainty increases, causing the extreme value to shrink from 1 to 0.9 and then to 0.4, while the revenue increases twice, with increases of 6.69% and 3.39% respectively. Through comparisons within scenarios and between different scenarios, it can be seen that adding energy storage and using robust optimization methods can significantly improve the revenue of new energy clusters in the spot market, and for different scenarios, appropriate robust coefficients need to be selected to maximize the revenue.

[0272] It is believed that the benefits of scenarios 2 to 4 when they are optimal are the real benefits under the scenarios. Scenarios 1 to 4 are compared and analyzed. The transaction results and scheduling conditions are shown in Tables 2 and Figure 7 shown.

[0273]

[0274] Table 2 Revenue of new energy clusters under different scenarios (unit: 10,000 yuan)

[0275] According to Table 2, by comparing Scenario 1 and Scenario 2, the use of robust optimization can significantly reduce the negative deviation penalty of the new energy cluster, with a decrease of 54.23%. At the same time, the bid amount in Scenario 2 is more conservative, which reduces the revenue of day-ahead electricity sales. At the same time, the sales volume increases in the real-time stage, and the positive deviation penalty increases. Overall, the use of robust optimization increases the net income of the new energy cluster in Scenario 2 by 266,000 yuan, an increase of 11.35%. By comparing Scenario 2 and Scenario 3, although the addition of energy storage increases the investment cost and operating cost of the cluster by 5.60% and 12.87% respectively, the addition of energy storage to the new energy cluster can significantly reduce the negative deviation penalty. Overall, the income of Scenario 3 increases by 174,400 yuan compared with Scenario 2, an increase of 6.69%. By comparing Scenario 3 and Scenario 4, it can be seen that shared energy storage makes the cluster more resistant to risks, and the bid amount in the day-ahead market is greater than that in Scenario 3. The use of shared energy storage in Scenario 4 improves resource utilization efficiency. Therefore, the deviation penalty for Scenario 4 was reduced by RMB 96,800, and the net income growth increased by RMB 94,300, an increase of 3.28%.

[0276] And by Figure 7 It can be seen that by comparing Scenario 1 and Scenario 2, the use of robust optimization reduces the day-ahead bidding volume in some periods of Scenario 2, resulting in an increase in the real-time stage electricity sales. In addition, both Scenario 1 (7h, 9h et) and Scenario 2 (16h, 21het) have the situation that electricity purchase and sale occur simultaneously in some periods, which is because the new energy field groups fail to cooperate and cannot play a mutual assistance role. Comparing Scenario 2 and Scenario 3, it can be seen that the addition of energy storage makes it no longer necessary for the alliance to balance its own power generation deviation through the power grid in some periods, such as 20h; or significantly reduce the positive and negative deviation electricity, such as 13h and 14h. At the same time, it can be seen that in Scenario 3, there is a phenomenon that energy storage is both charged and discharged in the same period, such as 2h and 6h, resulting in the non-optimal utilization of resources. Comparing Scenario 3 and Scenario 4, it can be seen that the addition of shared energy storage increases the day-ahead bidding volume in most periods of Scenario 4. At the same time, due to the combined effect of virtual energy storage and physical energy storage, Scenario 4 does not have the phenomenon of coexistence of charging and discharging in the same period, and the real-time positive and negative deviation electricity is greatly reduced compared with Scenario 3.

[0277] To further verify the advantages of shared energy storage, the energy storage working status in scenarios 3 and 4 is analyzed in detail. Figure 8 As shown, from Figure 8 It can be seen that in scenario 3, in order to reduce its own deviation penalty, each new energy farm has a large number of opposite energy storage charging and discharging states, resulting in unnecessary waste of energy; although energy storage can pursue the price difference profit at some times, charging during the real-time low electricity price period (13h-14h) and discharging during the peak electricity price period (19h, 21h), but the different charging and discharging conditions between new energy farms result in the regulation function of energy storage not being maximized.

[0278] In scenario 4, on the virtual energy storage platform, the charging and discharging demands of the shared energy storage called in real-time scheduling by each member of the cluster offset each other at most times within a scheduling cycle. From the comparison of output power, it can be seen that the maximum power of the virtual shared energy storage can reach 52.7% of the maximum power of the physical shared energy storage; from the call time, the call time of the virtual shared energy storage is more than that of the physical shared energy storage. The use of virtual shared energy storage greatly reduces the operating losses and costs generated by shared energy storage; at the same time, the offset of the charging and discharging power of energy storage called between members reduces the capacity demand of shared energy storage to a certain extent. In the shared energy storage planning stage, the wind farm cluster can reduce investment costs and select energy storage with a smaller rated capacity. In summary, the addition of the energy storage proposed in the present invention and the use of robust optimization methods can significantly improve the profitability of new energy farms in the spot market, and shared energy storage can obtain higher profits than non-shared energy storage.

[0279] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing energy storage coordination and robust scheduling based on joint sharing of new energy farms, characterized by: The scheduling optimization steps include the following: Step 1: Use a new energy cluster joint shared energy storage model consisting of a new energy cluster entity and a shared energy storage entity to conduct electricity trading; Step 2: Perform uncertainty processing, use Weibull distribution and Beta distribution to describe the probability density of natural wind speed and solar radiation intensity, and generate a set of wind and solar power forecast output scenarios within the day through improved Latin hypercube sampling and K-means reduction; Step 3: Establish two-stage trading framework rules for the day-ahead market stage and the real-time market stage; Step 4: Establish the transaction optimization model for the day-ahead market stage and the real-time market stage, and establish a two-stage robust optimization model, where: In the transaction optimization model of the day-ahead market stage, based on the predicted power of WPP and PV, the day-ahead market bidding is carried out with the goal of maximizing the day-ahead market revenue; In the transaction optimization model at the real-time market stage, the difference between the actual power generation of WPP and PV and the day-ahead bid power is considered, and a real-time adjustment plan is formulated with the goal of maximizing the real-time market profit; Step 5: For the above two-stage robust optimization model, a column constraint generation algorithm is used to transform the above problem into a two-level optimization problem including a main problem and sub-problems, and the answer is solved by alternating the main problem and the sub-problems; Step 6: Analyze the effect of robust scheduling optimization in various application scenarios.

2. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 1 is characterized in that: In step 1, the new energy field group is specifically composed of multiple new energy fields, which are either wind farms or photovoltaic power stations. New energy clusters can sell electricity in the spot market and green certificates in the green certificate market. It is defined that for every megawatt-hour of electricity produced by a new energy cluster, one green certificate will be obtained. The shared energy storage entity includes physical energy storage PES and virtual energy storage VES.

3. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 2 is characterized in that: The specific method for uncertainty processing in step 2 is: Step 2.1: Conduct wind power uncertainty analysis: The uncertainty of wind turbine output depends on the random characteristics of wind speed, which is described by Weibull distribution and expressed as: Where: v is the wind speed at any time, c is the scale parameter of the Weibull distribution, and k is the shape parameter, where k and c can be calculated based on the mathematical expectation and standard deviation of the wind speed sampling sequence samples; Step 2.2: Perform optoelectronic uncertainty analysis: The output uncertainty of photovoltaic generators depends on the random characteristics of solar radiation intensity, which is described by Beta distribution and expressed as: Where: r is the solar irradiance during period t, r max is the maximum solar irradiance in period t, α and β are the shape parameters of the Beta distribution, which are calculated from the expectation and variance of the solar radiation intensity; Step 2.3: Use the improved Latin hypercube sampling method to simulate and generate uncertainty scenarios of actual wind and solar power output.

4. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 3 is characterized in that: The specific method for simulating and generating the uncertainty scenario of the actual wind and solar power output in step 2.3 is: Step 2.3.1: Sampling uncertainty scenarios: Step 2.3.1.1: All historical wind and solar data are summarized by day and divided into 24 time periods with a span of hours, from 1h to 24h. Suppose any time period is i, i = 1, 2, ..., 24, then the wind speed parameter k of the i-th time period is i 、c i , and the solar radiation parameter α i , β i It can be calculated from the mean and standard deviation of wind speed and solar radiation samples; Step 2.3.1.2: Based on the shape parameters of each period, the probability distribution functions of wind speed and solar radiation in each period are constructed according to formula (1) and formula (2), and the cumulative probability distribution model F is derived. k,i ; Step 2.3.1.3: Accumulate the probability distribution model F for each period k,i Latin hypercube sampling is performed to generate multiple unrelated sample matrices. The sampling process is: Assume that the sampling scale is N, and the cumulative probability distribution curve F derived in step 2.3.1.2 is k,i (X k,i ) The vertical axis is divided into N equally spaced, non-overlapping intervals. The range of the dependent variable of the cumulative probability distribution is set to [0,1], then the length of each interval is 1 / N. Then, completely random sampling is performed in each interval [(n-1) / N, n / N], and the vertical coordinate value drawn is defined as y kn,i , substitute the obtained ordinate value into the cumulative probability distribution function F k,i The inverse function of X k,i The nth sampling value of is expressed as: Step 2.3.2: Perform uncertainty scene reduction, specifically using the K-means algorithm to reduce the scene, the specific steps are: Step 2.3.2.1: Set the number of iterations t = 1, the number of clusters to L, and randomly select the initial value J of each cluster center t (l), l = 1, 2, ..., L; Step 2.3.2.2: Calculate each sample X k The Euclidean distance from L cluster centers, and classify the sample into the class where the cluster center with the smallest distance is located; Step 2.3.2.3: Assume that the lth class has q samples, then the new cluster center J t+1 (l) is: Step 2.3.2.4: Repeat steps 2.3.2.2 and 2.3.2.3 until the position change of the centroid is less than the set threshold or the maximum number of iterations is reached; Step 2.3.2.5: Output the final clustering results and the probability of each scene, where the probability of the lth scene is calculated as:

5. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 4 is characterized in that: In step 3, the trading framework rules for the day-ahead market phase are established as follows: Each member of the new energy cluster provides the trading center with bidding power and quotation for the next 24 hours based on the predicted wind power and photovoltaic power generation. Each member submits a bid at a low electricity price to achieve the goal of winning the bid for all bidding power, and finally obtains power generation profits at the unified market clearing price. The trading framework rules for the real-time market stage are as follows: The members of the new energy farm group know the winning bid power and actual power generation, and make the following adjustments to the resulting deviations to reduce the deviation penalty costs: In the case where the deviation between the actual output of some members and the bid output is positive and the deviation of some members is negative at the same time, the part that offsets the positive and negative deviations is directly connected to the grid through virtual energy storage, and the remaining deviation is smoothed by calling physical energy storage. If the shared energy storage is not enough to make up for the deviation, the new energy e-commerce company will be punished for the deviation. The positive deviation part will be sold in the real-time market, and the negative deviation part needs to be purchased from other power entities to smooth the deviation. When there is sufficient electricity generated by renewable energy, the excess electricity can be purchased through the energy storage system; When the power generation from renewable energy sources is insufficient, the stored electricity can be sold to earn a profit from the price difference.

6. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 5 is characterized in that: The specific method of establishing the transaction optimization model in the day-ahead market stage in step 4 is: New energy clusters and shared energy storage collaborate to participate in the day-ahead market to maximize their joint net benefits. Wind farms and photovoltaic power stations forecast the wind and solar output of the next day and report the electricity volume and price based on historical electricity price information: Among them, the expression of the day-ahead market objective function is: Where: The benefits of the new energy cluster participating in the day-ahead market, are the daily average costs of the new energy cluster and shared energy storage, respectively, where RE = {WPP, PV}; Where: T is 24; is the clearing electricity price in the day-ahead market at time t; N is the number of members in the new energy cluster; are the day-ahead market winning output and day-ahead output forecast of the new energy field i at time t; Δt is the time interval; The operating cost and depreciation cost of the new energy cluster respectively; i is the unit annual operation and maintenance cost of the new energy field i; RE,i is the unit annual investment cost of new energy field i; P i max is the investment power of new energy field i; r is the interest rate; γ RE,i is the depreciation period of the new energy field i; The constraints of the day-ahead market objective function include: The output constraint of new energy cluster members is expressed as: Virtual energy storage constraints: Virtual energy storage does not need to consider the charging and discharging efficiency during operation, so the expression for the dynamic change of battery energy storage level in each period is: Where: and are the energies stored in the virtual energy storage at time t and time t+1 respectively; are the charging value and discharging value of the virtual energy storage at time t respectively; The virtual energy storage is set to half of the maximum total power of the new energy cluster, and the expression is: Where: are the maximum values ​​of charging and discharging of virtual energy storage, respectively; It is the charging and discharging 0-1 state variable of the virtual energy storage; The expression of the amount of electricity stored at time 0 of virtual energy storage is: Physical energy storage constraints: Physical energy storage needs to consider the charging and discharging efficiency during operation. The expression for the dynamic change of battery energy storage level in each period is: Where: and are the energies stored in the physical energy storage at time t and time t+1 respectively; are the charging value and releasing value of the physical energy storage at time t respectively; To truly reflect the operation strategy, the battery storage level at the beginning of each day should remain consistent, expressed as: In addition, physical energy storage needs to consider the following constraints: Where: They are the maximum values ​​of physical energy storage charging and discharging respectively; Indicates the charge and discharge state of physical energy storage. When the value is 1, it indicates discharge, and when the value is 0, it indicates charge. and The minimum or maximum remaining capacity allowed, respectively; Power balance constraints: The expression that satisfies power balance is:

7. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 6 is characterized in that: The specific method of establishing the transaction optimization model in the real-time market stage in step 4 is: The new energy cluster knows its own winning bid volume and winning bid price in the day-ahead market, and requires to reduce the deviation through shared energy storage and backup services to ensure the maximization of its own net profit: The expression of the real-time market objective function is: Where: To enable new energy clusters to collaboratively share the benefits of energy storage participation in real-time markets; Proceeds from selling green certificates for new energy clusters; is the deviation penalty cost, the cost of purchasing ancillary services for the system from the electricity market; Where: is the clearing electricity price in the real-time market at time t; It is the additional electricity sold by the new energy site i in the real-time market stage; is the actual output of new energy field i at time t; green is the green certificate conversion factor, 1 megawatt of green electricity is converted into 1 green certificate; p green is the selling price of green certificates; Where: is the amount of electricity purchased by new energy farm i in the real-time market stage; δ1 and δ2 are the penalty coefficients for positive and negative deviations in the real-time market respectively; The constraints of the real-time market objective function include: The power balance constraint is expressed as: The state constraints of electricity sales and purchase of auxiliary services are expressed as: Where: Indicates the purchase and sale status of the system. A value of 1 indicates the purchase of auxiliary services, and a value of 0 indicates the sale of electricity.

8. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 7 is characterized in that: The specific method for establishing the two-stage robust optimization model in step 4 is: Without considering the uncertainty of wind and solar power output, a deterministic optimization model for the participation of new energy clusters in the electricity market through collaborative shared energy storage is obtained. The compact form of the model is expressed as: In the formula, x and y are the decision variables of the first and second stages, respectively, and their expressions are: In order to make the model conform to the two-stage robust expression, the objective function of the first stage is negated and converted to the minimum to satisfy the expression: a T x=-R ahead (32); In order to solve the uncertainty problem of wind and solar power, robust optimization is performed. In order to achieve the maximum value of the second-stage objective function, the max of the second stage and the max of the uncertainty set U are merged into one, so that the expression of the abstract constraint model of the two-stage robust optimization is: Where: a, b are the coefficient column vectors corresponding to the objective function expressions (5) and (22) in the day-ahead and real-time stages, respectively; D, Q, K, R, S, M, L, Y are the coefficient matrices of the variables under the corresponding constraints; d, k, h are constant column vectors; Where: The first and second rows of the constraints represent the inequality constraints in the model, where: The first row includes equations (12), (13), (18), (19), (20), (28), and (29); The second row includes equation (10); The third row represents the equality constraints in the model, including equations (11), (16), and (17); The 4th row corresponds to equation (27); The 5th row corresponds to equation (21); Among them, the fluctuation range of each new energy cluster output is determined by the typical scenarios extracted. To reduce the output of the new energy field i at time t in the post-scenario j, the calculation formula for the maximum fluctuation range of the output of the new energy field i is: Where: is the fluctuation deviation at time t; The maximum fluctuation value of the output of the new energy field group is obtained through the maximum fluctuation range calculation formula, and then the uncertainty of the output of the new energy field group is expressed as the uncertainty set U, which is expressed as: Where: Γ is the robustness coefficient, is the upper limit of the renewable energy output under consideration, P t RE0 To provide a lower limit for the new energy sources under consideration; Since the output of the new energy field will face a greater deviation penalty cost when it takes the minimum value of the interval, the adjustment parameter L is introduced to adjust formula (32) into the following form: Where: It is a binary variable. When it takes the value of 1, the uncertain variable of the corresponding period reaches the boundary of the interval.

9. The energy storage coordinated robust scheduling optimization method based on joint sharing of new energy field groups according to claim 8 is characterized in that: The specific method for solving the answer of the two-stage robust optimization model in step 5 is: By decomposing the expression (29) of the state constraints of electricity sales and purchase of ancillary services, the expression of the main problem obtained at the kth iteration of the two-stage robust model is: Where: θ is the auxiliary variable for real-time market stage profit in the MP objective function; l is the current iteration number; y l is the solution of the subproblem after the lth iteration; is the value of the uncertain variable u in the worst scenario determined after the lth iteration; The expression of the sub-problem is: After the above transformation, the two-stage robust model is decoupled into the main problem (33) and the sub-problem (34). Both problems are mixed integer linear programming problems, which are then solved using the column constraint generation algorithm. The specific process is as follows: Set the lower bound of the objective function to LB = -∞ and the upper bound to UB = +∞; In the worst-case scenario forecast * Next, solve the main problem equation (33) and get the optimal solution The optimal solution of the main problem is taken as the new lower bound, that is Solve the main problem Substitute into the subproblem equation (34) and obtain the optimal solution of the subproblem Update upper bound Assume that the algorithm convergence threshold is ε. If UB-LB≤ε, stop the calculation and return the optimal solution. Otherwise, let k=k+1 and go back to solve the main problem (33).

Images