Uncertain scene-oriented micro-grid and shared energy storage collaborative robust optimization method

Abstract

The invention provides an uncertain scene-oriented micro-grid and shared energy storage collaborative robust optimization method. The method comprises the following steps of S1, constructing a micro-grid group and shared energy storage collaborative scheduling optimization operation framework; s2, constructing a wind and light uncertain scene set by adopting a scene generation technology; s3, constructing a load uncertainty scene set based on data-driven K-means clustering; s4, the activation probability of the load disturbance boundary is stably estimated based on the Wasserstein distance; s5, constructing a microgrid group and shared energy storage two-stage robust scheduling optimization model; step S6: carrying out dual transformation and Camp; solving the two-stage robust optimization model through a CG algorithm; step S7, designing an improved Shapley value method income allocation mechanism based on the network topology sensitive model; through load scene construction of data driving and clustering analysis, wind and light scene generation and clustering reduction technologies and in combination with collaborative optimization scheduling of the micro-grid group and shared energy storage, the renewable energy utilization efficiency can be improved, the intraday operation cost can be reduced, and collaborative development of energy storage and new energy can be assisted.

Description

Technical Field

[0001] This invention relates to the field of microgrid cluster and shared energy storage collaborative optimization technology, and in particular to a robust optimization method for microgrid and shared energy storage collaborative optimization in uncertain scenarios. Background Technology

[0002] With the continuous increase in electricity demand and the transformation of the energy structure, electrochemical energy storage has been widely used in scenarios involving new energy sources, the grid, and users due to its high energy density and bidirectional power regulation characteristics. In microgrids or multi-energy systems at the park level, energy storage systems are widely used for local new energy consumption, peak shaving and valley filling, and load regulation. However, at present, most microgrids still adopt a "decentralized self-built, self-consumption-oriented" energy storage configuration approach, investing in and constructing energy storage equipment separately. Although this model has a certain degree of independence and flexibility, it also has problems such as redundant construction, high operation and maintenance costs, and low capacity utilization.

[0003] Shared energy storage, as an important means to improve the utilization efficiency of energy storage resources, has received widespread attention in recent years in multi-energy microgrids, park-level energy systems and distributed energy scenarios. It has broad development prospects in the power system and can be divided into two types according to its investment and operation model: user co-construction and sharing and operator investment and sharing.

[0004] Current research largely focuses on investment and operation strategies or profit game modeling for shared energy storage. Among existing studies, some have proposed planning, investment, operation, and pricing models under the cloud energy storage business model; others have established a master-slave game model for shared energy storage, solving for optimal energy storage revenue through scenario generation and Nash bargaining; still others have established a three-sided energy storage optimization operation model involving new energy sources, the power grid, and users, proposing a new capacity sharing mechanism.

[0005] Renewable energy sources and load fluctuations within microgrids exhibit uncertainty and uncontrollability. When source-grid-load coordination occurs, it brings a series of impacts to distribution network protection configuration and power quality. To mitigate the potential reduction in renewable energy utilization and unsatisfactory investment returns for self-built energy storage in microgrid clusters due to uncertainties on both the source and load sides, a shared energy storage collaborative optimization model is introduced. However, how to integrate the uncertainties of both renewable energy output and user-end load in shared energy storage dispatch, construct a robust model, and fully leverage the collaborative advantages of source-grid-load-storage remains a pressing issue. Summary of the Invention

[0006] This invention proposes a robust optimization method for the coordinated operation of microgrids and shared energy storage in uncertain scenarios. This method uses data-driven and cluster analysis to construct load scenarios and generate wind and solar scenarios, as well as cluster reduction technology. Combined with the coordinated optimization and scheduling of microgrid clusters and shared energy storage, it can improve the utilization efficiency of renewable energy, reduce daily operating costs, and promote the coordinated development of energy storage and new energy.

[0007] The present invention adopts the following technical solution.

[0008] A robust optimization method for coordinating microgrids and shared energy storage in uncertain scenarios includes the following steps:

[0009] Step S1: Construct a scheduling optimization operation framework for the coordinated operation of microgrid clusters and shared energy storage;

[0010] Step S2: Construct a set of uncertain landscape scenes using scene generation technology;

[0011] Step S3: Construct a set of load uncertainty scenarios based on data-driven K-means clustering;

[0012] Step S4: Robustly estimate the activation probability of the load disturbance boundary based on Wasserstein distance;

[0013] Step S5: Construct a two-stage robust scheduling optimization model for microgrid clusters and shared energy storage;

[0014] Step S6: Solve the two-stage robust optimization model using dual transformation and the C&CG algorithm;

[0015] Step S7: Design of a revenue sharing mechanism for the improved Shapley value method based on the network topology sensitivity model.

[0016] In step S1, for the uncertain wind and solar power scenario (i.e., the uncertainty of wind turbine power generation and photovoltaic power generation), an operational framework is constructed that considers the shared energy storage operator's power capacity in the microgrid group's source-grid-load system. This includes: source-grid-load models within each area of ​​the microgrid group (MGs) and a shared energy storage operator (SESS) model. The SESS access point is located at the point of common connection (PCC). Each microgrid in the microgrid group can flexibly share power with the SESS through the PCC. The overall operational framework is described below. Figure 2 As shown;

[0017] Each MGs park contains distributed photovoltaic (PV), wind turbine (PW), controllable distributed power sources, and loads. The introduction of SESS (Support for Energy Storage and Renewable Energy) aims to promote local consumption of renewable energy and peak shaving and valley filling while minimizing the daily operating costs of MGs. First, MGs prioritize the consumption of renewable energy generation and balance load demand through controllable distributed power sources. Second, MGs reduce unnecessary load demand by responding to load demand within the MGs and incentivize MGs to utilize the power capacity of SESS to respond to peak shaving and valley filling. Finally, MGs purchase electricity from the upper-level grid to meet the remaining unbalanced load.

[0018] The operational goal of microgrid groups (MGs) is to aggregate the demand of each subgrid to achieve large-scale demand response while ensuring the reliability of users' energy consumption. At the same time, by combining controllable distributed power sources within the MGs and sharing energy storage capacity through leasing third parties, it can coordinate the response to renewable energy output and load disturbances, thereby minimizing the daily operating cost of the system.

[0019] SESS's operational objective is to support peak shaving and valley filling of loads within MGs. This is mainly achieved by constructing centralized shared energy storage power stations at PCC points in various MGs parks. SESS provides leasing services to MGs, and the charging and discharging strategies of SESS are formulated based on the leasing needs of MGs. Arbitrage is achieved by utilizing low storage capacity and high discharge capacity based on time-of-use pricing.

[0020] The specific model of the scheduling optimization operation framework is expressed by the following formula:

[0021]

[0022] Formula 1 is the shared energy storage operator model. The discharge and charge power of SESS are both non-negative integers, λ. ssD For discharge identification, λ ssC This is a charging indicator; NT represents the upper limit of the discharge / charge power that SESS can provide; NT represents the number of cells in the microgrid group; T is taken as 24 hours; and η is the charge / discharge efficiency. and To impose upper and lower limits on the amount of content in SESS during different time periods under the discharge / charging states during the sharing process. This means that SESS can share the initial power. Δt is the daily scheduling step size, which takes the value of 1 hour. t = 1 and t' represent the time interval from the first scheduling time point to any subsequent time point. Quantifying the cost of SESS charge / discharge degradation k is the battery degradation coefficient. ss (t) represents the charging and discharging cost of SESS, and k prof This represents the expected revenue for SESS.

[0023]

[0024] Formula 2 is the controllable distributed power source model, P g (t) represents the sum of the outputs of the gas turbine and the heat pump at time t, P i,rq η is the rated output power of the gas turbine / heat pump within the park. rq For gas turbine efficiency, η rb For heat pump efficiency, c gs The gas purchase price is (yuan / kW); P G.max P G.min These are the upper and lower limits of the output of controllable distributed power sources within a microgrid group;

[0025]

[0026] Formula 3 is the demand response model for microgrid groups, P DR (t) represents the demand response load power within the microgrid group, S DR The total scheduling capacity for demand response load from the first point in time to a certain point in time; The minimum and maximum power demand of the demand response load for a certain period of time.

[0027]

[0028] Formula 4 is the interaction model between the microgrid group and the upper-level distribution network. Let t be the power purchased by the microgrid group from the distribution network at time t; P represents the power sold by the microgrid group to the distribution network at time t. L (t) represents the active load pre-connected to the microgrid group at time t; P DPS (t) represents the output power of distributed renewable energy sources within the microgrid group at time t; P MG.max The upper limit of the power that can be transmitted by the incoming line of the upper-level distribution network; λ PUR , λ SEL This indicates the purchase and sale of electricity.

[0029]

[0030] Formula 5 represents the overall objective function and the objective functions of each subsystem for the collaborative scheduling optimization of microgrid clusters and shared energy storage operators. C ss (t) represents the total cost of leasing a SESS to a microgrid cluster; C g (t) represents the total cost of a controllable distributed power source, a g C represents the linear first-order coefficient of a controllable distributed power source, where b is a known constant; DR (t) represents the demand response cost, k DR C represents the demand response cost coefficient. gtax (t) represents the transmission and distribution costs incurred by SESS using the existing power grid; C INR (t) represents the cost of interacting with the distribution network.

[0031] The implementation method of step S2 is as follows:

[0032] Assuming that the PV power prediction error follows a normal distribution, while the wind turbine PW power error conforms to a Beta distribution, the PV and PW power prediction output models are as follows:

[0033]

[0034] In equation (6), Daytime power correction value during period t under scenario s; This is the predicted value for the previous day; ξ j R represents the prediction error. j,t is a predicted random number that follows a normal distribution; r is a correction factor; the subscript j represents any photovoltaic power output unit; μ and σ are the expectation and variance values ​​of the prediction model; α and β are the shape and scale parameters of the Beta distribution, respectively;

[0035] Based on the above prediction model, typical scenes are generated using Latin hypercube sampling (LHS) technology. K-means clustering is then used to reduce the number of typical power output scenes. Finally, the upper and lower limits of the prediction deviation of the wind and light scene are calculated using the clustered scenes.

[0036] The implementation method of step S3 is as follows: the load is determined by the elbow rule based on historical data, the L2 norm between each sample is calculated by K-means iteration to obtain each stable cluster center, and the volatility evaluation index of the samples in each cluster is calculated. The upper and lower limits of the load deviation in each cluster scenario are output by the proportion of volatility.

[0037]

[0038] In Formula 7, x i For the i-th sample, μ k(i) The mean vector represents the cluster centers, K is the number of cluster centers, and C is the mean vector. k Represents the k-th cluster, δ i Let d be the sign deviation function. i For L2 norm, SVP i Used to calculate the proportion of sample fluctuation, i.e. the fluctuation boundary of the load; where sign(·) is the sign function, used to preserve the deviation direction between the sample and the cluster center.

[0039] The implementation method of step S4 is as follows: An adjustment factor is generated using an activation probability estimation method driven by Wasserstein distance and historical data to objectively reconstruct the number of fluctuations under typical load scenarios, which is then used to construct an activation probability model for uncertain factors with data-driven interpretability. The Wasserstein distance constraint modeling formula is as follows:

[0040]

[0041] l(x) = {x > τ} (t)} Formula 9;

[0042]

[0043] In the above formula, τ (t) Representative of samples in the scene-based clustering results Samples at the right-tailed activation threshold calculated using IQR (Interquartile Range) are used with Quantile. 0.75 (·) represents the indicator function after activation; l(x) is the corresponding indicator function after activation; Let represent the supremum estimate of the expectation of the indicator function, where For the empirical probability distribution of the sample, The most unfavorable probability distribution of the sample is given, where W1 represents the first-order Wasserstein distance, θ is the radius of the Wasserstein sphere used to control the size of the uncertain set, and U... θ Represents a spherically uncertain set, containing all sets with the same empirical distribution. The distribution with a distance not exceeding θ; Equation (11) represents the linearization transformation of (8)-(10) under the condition of Lipschitz continuity, transforming the discrete typical discretized samples into a linear programming form for solution, where p i d represents the probability weight of the i-th sample in the most unfavorable probability distribution. i For each sample x i Wasserstein distance to the center of the empirical distribution.

[0044] The implementation method of step S5 is as follows:

[0045] In the first stage, a three-level objective function model of min-max-min is constructed, as shown below;

[0046]

[0047] In Formula 12, x represents the decision variable of the main problem, y represents the decision variable of the subproblem, and u represents the uncertain scenario belonging to the uncertain scenario set U.

[0048] In the second stage, MGs initializes the main problem based on the load and new energy output forecast values, and solves the main problem. The objective function is the minimum daily operating cost of MGs. The lower bound LB of the main problem is calculated. The decision variable x obtained from the upper layer is then passed to the lower layer.

[0049] In the third stage, firstly, based on the upper and lower limits of the deviation range of the uncertain set of wind and solar loads in step S2, deviation interval constraints are imposed on the inner layer of unfavorable conditions. Secondly, an uncertainty adjustment factor for the inner layer of wind and solar loads is set. Thirdly, a load fluctuation adjustment factor is generated using the activation probability estimation method driven by historical data in step S3. Fourthly, based on this, the inner layer problem of max-min is transformed into a max single-layer problem through duality. Then, the objective function value of the inner layer under the most unfavorable conditions is calculated based on the aforementioned steps, and the calculated LB and uncertain scenario variable u of the inner layer are returned to the main problem. The main problem then solves for MB and decision variable x again. The fourth step is repeated until UB-LB converges to the preset residual value requirement, at which point the iteration exits.

[0050] The implementation method of step S6 is as follows: the two-stage robust inner max-min problem is transformed into a single-layer max problem through dual transformation and KKT conditions, and the C&CG algorithm is used to solve the two-stage model.

[0051] The implementation method of step S7 is as follows: Based on the two-stage robust optimization model created in step S5 with the lowest daily operating cost of MGs and shared energy storage cooperation as the objective function, an improved Shapley value revenue allocation method based on network topology sensitivity is proposed to more fairly reflect the revenue allocation problem caused by the different transmission distances of different members of the microgrid group from the point of common connection (PCC) in the grid; the specific model is as follows:

[0052]

[0053] In formula 13, MG in microgrid cluster i The marginal contribution of a member, v(s) represents the characteristic function of the revenue of a microgrid group member, and v(s\i) represents the marginal contribution of a member without considering a certain MG. i The benefits of the alliance; n is the MG in the microgrid i The total number, v(n) is the sum of the alliance's profits, x i For the profit sharing among all alliance members, v(i) represents the profit shared by MG, which is the only member in the alliance. i The alliance benefit at that time; s is the MG in the alliance currently being calculated. i The number of members, w(|s|) represents the weight coefficient under this condition when the number of members in the alliance is s;

[0054] In Formula 14, U is the nominal value of the voltage at the point of common coupling, and P... MGi For each MG i The average active load of the members, R MGi For each MG i The impedance parameters within a member have nominal values, R Li For the line impedance parameters of the corresponding member and the PCC point, P' MGiFor the adjusted MG i The active load of the member, R' MGi For the adjusted MG i The impedance parameter within a member has a nominal value, λ. i This is the loss sensitivity factor.

[0055] This invention relates to a robust optimization method for the coordinated operation of microgrids and shared energy storage in uncertain scenarios, comprising the following steps: Step S1: Constructing a scheduling optimization operation framework for the coordinated operation of microgrid clusters and shared energy storage; Step S2: Constructing a set of uncertain wind and solar scenarios using scenario generation technology; Step S3: Constructing a set of uncertain load scenarios based on data-driven K-means clustering; Step S4: Robustly estimating the activation probability of load disturbance boundaries based on Wasserstein distance; Step S5: Constructing a two-stage robust scheduling optimization model for microgrid clusters and shared energy storage; Step S6: Solving the two-stage robust optimization model through dual transformation and C&CG algorithm; Step S7: Designing a revenue sharing mechanism based on the improved Shapley value method using a network topology sensitivity model. This method, through data-driven and cluster analysis-based load scenario construction and wind and solar scenario generation and cluster reduction techniques, combined with the coordinated optimization scheduling of microgrid clusters and shared energy storage, improves the utilization efficiency of renewable energy, reduces daily operating costs, and promotes the coordinated development of energy storage and new energy.

[0056] Compared with existing technologies, the present invention has the following beneficial effects: It provides a robust optimization method for the coordinated operation of microgrids and shared energy storage in uncertain scenarios. This method constructs load scenarios and generates wind and solar scenarios through data-driven and cluster analysis, and uses cluster reduction technology. It combines Wasserstein distance robust estimation of load boundary activation probability to reduce the conservatism of the two-stage robustness in the artificial setting of scenario uncertainty factors. Based on the two-stage robust structure, it optimizes the coordinated scheduling of microgrid groups and shared energy storage, improves the utilization efficiency of renewable energy, reduces the daily operating cost of microgrid groups, and promotes the coordinated development of energy storage and new energy. Attached Figure Description

[0057] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0058] Appendix Figure 1 This is a schematic diagram of the method implementation process according to an embodiment of the present invention;

[0059] Appendix Figure 2 This is a schematic diagram of the collaborative system framework of the Shared Energy Storage Station (SESS) in an embodiment of the present invention;

[0060] Appendix Figure 3 This is a schematic diagram of the process for solving the two-stage robust model according to an embodiment of the present invention;

[0061] Appendix Figure 4This is a schematic diagram of the photovoltaic power output prediction curve and error upper and lower limits of the microgrid group in an embodiment of the present invention;

[0062] Appendix Figure 5 This is a schematic diagram of the wind turbine output prediction curve and error upper and lower limits of the microgrid group in this embodiment of the invention;

[0063] Appendix Figure 6 This is a schematic diagram of the load curve and upper and lower limits of error for a typical scenario 1 of a microgrid group in an embodiment of the present invention;

[0064] Appendix Figure 7 This is a schematic diagram of the load curve and upper and lower limits of error for a typical scenario 2 of a microgrid group in an embodiment of the present invention;

[0065] Appendix Figure 8 This is a schematic diagram of the power balance of a microgrid group and SESS in a typical scenario 1 of this invention.

[0066] Appendix Figure 9 This is a schematic diagram of the power grid group and SESS power balance in typical scenario 2 of the embodiments of the present invention;

[0067] Appendix Figure 10 These are schematic diagrams of operating cost surfaces under two typical load scenarios according to embodiments of the present invention. Detailed Implementation

[0068] As shown in the figure, a robust optimization method for microgrids and shared energy storage in uncertain scenarios includes the following steps:

[0069] Step S1: Construct a scheduling optimization operation framework for the coordinated operation of microgrid clusters and shared energy storage;

[0070] Step S2: Construct a set of uncertain landscape scenes using scene generation technology;

[0071] Step S3: Construct a set of load uncertainty scenarios based on data-driven K-means clustering;

[0072] Step S4: Robustly estimate the activation probability of the load disturbance boundary based on Wasserstein distance;

[0073] Step S5: Construct a two-stage robust scheduling optimization model for microgrid clusters and shared energy storage;

[0074] Step S6: Solve the two-stage robust optimization model using dual transformation and the C&CG algorithm;

[0075] Step S7: Design of a revenue sharing mechanism for the improved Shapley value method based on the network topology sensitivity model.

[0076] In step S1, for the uncertain wind and solar power scenario (i.e., the uncertainty of wind turbine power generation and photovoltaic power generation), an operational framework is constructed that considers the shared energy storage operator's power capacity in the microgrid group's source-grid-load system. This includes: source-grid-load models within each area of ​​the microgrid group (MGs) and a shared energy storage operator (SESS) model. The SESS access point is located at the point of common connection (PCC). Each microgrid in the microgrid group can flexibly share power with the SESS through the PCC. The overall operational framework is described below. Figure 2 As shown;

[0077] Each MGs park contains distributed photovoltaic (PV), wind turbine (PW), controllable distributed power sources, and loads. The introduction of SESS (Support for Energy Storage and Renewable Energy) aims to promote local consumption of renewable energy and peak shaving and valley filling while minimizing the daily operating costs of MGs. First, MGs prioritize the consumption of renewable energy generation and balance load demand through controllable distributed power sources. Second, MGs reduce unnecessary load demand by responding to load demand within the MGs and incentivize MGs to utilize the power capacity of SESS to respond to peak shaving and valley filling. Finally, MGs purchase electricity from the upper-level grid to meet the remaining unbalanced load.

[0078] The operational goal of microgrid groups (MGs) is to aggregate the demand of each subgrid to achieve large-scale demand response while ensuring the reliability of users' energy consumption. At the same time, by combining controllable distributed power sources within the MGs and sharing energy storage capacity through leasing third parties, it can coordinate the response to renewable energy output and load disturbances, thereby minimizing the daily operating cost of the system.

[0079] SESS's operational objective is to support peak shaving and valley filling of loads within MGs. This is mainly achieved by constructing centralized shared energy storage power stations at PCC points in various MGs parks. SESS provides leasing services to MGs, and the charging and discharging strategies of SESS are formulated based on the leasing needs of MGs. Arbitrage is achieved by utilizing low storage capacity and high discharge capacity based on time-of-use pricing.

[0080] The specific model of the scheduling optimization operation framework is expressed by the following formula:

[0081]

[0082] Formula 1 is the shared energy storage operator model. The discharge and charge power of SESS are both non-negative integers, λ. ssD For discharge identification, λ ssC This is a charging indicator; NT represents the upper limit of the discharge / charge power that SESS can provide; NT represents the number of cells in the microgrid group; T is taken as 24 hours; and η is the charge / discharge efficiency. and To impose upper and lower limits on the amount of content in SESS during different time periods under the discharge / charging states during the sharing process. This means that SESS can share the initial power. Δt is the daily scheduling step size, which takes the value of 1 hour. t = 1 and t' represent the time interval from the first scheduling time point to any subsequent time point. Quantifying the cost of SESS charge / discharge degradation k is the battery degradation coefficient. ss (t) represents the charging and discharging cost of SESS, and k prof This represents the expected revenue for SESS.

[0083]

[0084] Formula 2 is the controllable distributed power source model, P g (t) represents the sum of the outputs of the gas turbine and the heat pump at time t, P i,rq η is the rated output power of the gas turbine / heat pump within the park. rq For gas turbine efficiency, η rb For heat pump efficiency, c gs The gas purchase price is (yuan / kW); P G.max P G.min These are the upper and lower limits of the output of controllable distributed power sources within a microgrid group;

[0085]

[0086] Formula 3 is the demand response model for microgrid groups, P DR (t) represents the demand response load power within the microgrid group, S DR The total scheduling capacity for demand response load from the first point in time to a certain point in time; The minimum and maximum power demand of the demand response load for a certain period of time.

[0087]

[0088] Formula 4 is the interaction model between the microgrid group and the upper-level distribution network. Let t be the power purchased by the microgrid group from the distribution network at time t; P represents the power sold by the microgrid group to the distribution network at time t. L (t) represents the active load pre-connected to the microgrid group at time t; P DPS (t) represents the output power of distributed renewable energy sources within the microgrid group at time t; P MG.max The upper limit of the power that can be transmitted by the incoming line of the upper-level distribution network; λ PUR , λ SEL This indicates the purchase and sale of electricity.

[0089]

[0090] Formula 5 represents the overall objective function and the objective functions of each subsystem for the collaborative scheduling optimization of microgrid clusters and shared energy storage operators. C ss (t) represents the total cost of leasing a SESS to a microgrid cluster; C g (t) represents the total cost of a controllable distributed power source, a g C represents the linear first-order coefficient of a controllable distributed power source, where b is a known constant; DR (t) represents the demand response cost, k DR C represents the demand response cost coefficient. gtax (t) represents the transmission and distribution costs incurred by SESS using the existing power grid; C INR (t) represents the cost of interacting with the distribution network.

[0091] The implementation method of step S2 is as follows:

[0092] Assuming that the PV power prediction error follows a normal distribution, while the wind turbine PW power error conforms to a Beta distribution, the PV and PW power prediction output models are as follows:

[0093]

[0094] In equation (6), Daytime power correction value during period t under scenario s; This is the predicted value for the previous day; ξ j R represents the prediction error. j,t is a predicted random number that follows a normal distribution; r is a correction factor; the subscript j represents any photovoltaic power output unit; μ and σ are the expectation and variance values ​​of the prediction model; α and β are the shape and scale parameters of the Beta distribution, respectively;

[0095] Based on the above prediction model, typical scenes are generated using Latin hypercube sampling (LHS) technology. K-means clustering is then used to reduce the number of typical power output scenes. Finally, the upper and lower limits of the prediction deviation of the wind and light scene are calculated using the clustered scenes.

[0096] The implementation method of step S3 is as follows: the load is determined by the elbow rule based on historical data, the L2 norm between each sample is calculated by K-means iteration to obtain each stable cluster center, and the volatility evaluation index of the samples in each cluster is calculated. The upper and lower limits of the load deviation in each cluster scenario are output by the proportion of volatility.

[0097]

[0098] In Formula 7, x i For the i-th sample, μ k(i) The mean vector represents the cluster centers, K is the number of cluster centers, and C is the mean vector. k Represents the k-th cluster, δi Let d be the sign deviation function. i For L2 norm, SVP i Used to calculate the proportion of sample fluctuation, i.e. the fluctuation boundary of the load; where sign(·) is the sign function, used to preserve the deviation direction between the sample and the cluster center.

[0099] The implementation method of step S4 is as follows: An adjustment factor is generated using an activation probability estimation method driven by Wasserstein distance and historical data to objectively reconstruct the number of fluctuations under typical load scenarios, which is then used to construct an activation probability model for uncertain factors with data-driven interpretability. The Wasserstein distance constraint modeling formula is as follows:

[0100]

[0101] l(x) = {x > τ} (t)} Formula 9;

[0102]

[0103] In the above formula, τ (t) Representative of samples in the scene-based clustering results Samples at the right-tailed activation threshold calculated using IQR (Interquartile Range) are used with Quantile. 0.75 (·) represents the indicator function after activation; l(x) is the corresponding indicator function after activation; Let represent the supremum estimate of the expectation of the indicator function, where For the empirical probability distribution of the sample, The most unfavorable probability distribution of the sample is given, where W1 represents the first-order Wasserstein distance, θ is the radius of the Wasserstein sphere used to control the size of the uncertain set, and U... θ Represents a spherically uncertain set, containing all sets with the same empirical distribution. The distribution with a distance not exceeding θ; Equation (11) represents the linearization transformation of (8)-(10) under the condition of Lipschitz continuity, transforming the discrete typical discretized samples into a linear programming form for solution, where p i d represents the probability weight of the i-th sample in the most unfavorable probability distribution. i For each sample x i Wasserstein distance to the center of the empirical distribution.

[0104] The implementation method of step S5 is as follows:

[0105] In the first stage, a three-level objective function model of min-max-min is constructed, as shown below;

[0106]

[0107] In Formula 12, x represents the decision variable of the main problem, y represents the decision variable of the subproblem, and u represents the uncertain scenario belonging to the uncertain scenario set U.

[0108] In the second stage, MGs initializes the main problem based on the load and new energy output forecast values, and solves the main problem. The objective function is the minimum daily operating cost of MGs. The lower bound LB of the main problem is calculated. The decision variable x obtained from the upper layer is then passed to the lower layer.

[0109] In the third stage, firstly, based on the upper and lower limits of the deviation range of the uncertain set of wind and solar loads in step S2, deviation interval constraints are imposed on the inner layer of unfavorable conditions. Secondly, an uncertainty adjustment factor for the inner layer of wind and solar loads is set. Thirdly, a load fluctuation adjustment factor is generated using the activation probability estimation method driven by historical data in step S3. Fourthly, based on this, the inner layer problem of max-min is transformed into a max single-layer problem through duality. Then, the objective function value of the inner layer under the most unfavorable conditions is calculated based on the aforementioned steps, and the calculated LB and uncertain scenario variable u of the inner layer are returned to the main problem. The main problem then solves for MB and decision variable x again. The fourth step is repeated until UB-LB converges to the preset residual value requirement, at which point the iteration exits.

[0110] The implementation method of step S6 is as follows: the two-stage robust inner max-min problem is transformed into a single-layer max problem through dual transformation and KKT conditions, and the C&CG algorithm is used to solve the two-stage model.

[0111] The implementation method of step S7 is as follows: Based on the two-stage robust optimization model created in step S5 with the lowest daily operating cost of MGs and shared energy storage cooperation as the objective function, an improved Shapley value revenue allocation method based on network topology sensitivity is proposed to more fairly reflect the revenue allocation problem caused by the different transmission distances of different members of the microgrid group from the point of common connection (PCC) in the grid; the specific model is as follows:

[0112]

[0113] In formula 13, MG in microgrid cluster i The marginal contribution of a member, v(s) represents the characteristic function of the revenue of a microgrid group member, and v(s\i) represents the marginal contribution of a member without considering a certain MG. i The benefits of the alliance; n is the MG in the microgrid i The total number, v(n) is the sum of the alliance's profits, x i For the profit sharing among all alliance members, v(i) represents the profit shared by MG, which is the only member in the alliance. iThe alliance benefit at that time; s is the MG in the alliance currently being calculated. i The number of members, w(|s|) represents the weight coefficient under this condition when the number of members in the alliance is s;

[0114] In Formula 14, U is the nominal value of the voltage at the point of common coupling, and P... MGi For each MG i The average active load of the members, R MGi For each MG i The impedance parameters within a member have nominal values, R Li For the line impedance parameters of the corresponding member and the PCC point, P' MGi For the adjusted MG i The active load of the member, R' MGi For the adjusted MG i The impedance parameter within a member has a nominal value, λ. i This is the loss sensitivity factor.

[0115] By updating the revenue weights of different members using Formula 14, the actual revenue ratio of each alliance member changes. Members closer to the PCC point have lower line losses and need to share more of the losses of members with longer path lengths and higher line impedances in the alliance proportionally. Members farther from the PCC point and with higher line impedances can also increase their revenue sharing weight proportionally based on their original marginal contribution.

[0116] As can be seen from the calculation results in Table 3, because MG1 and MG5 are far from the PCC point, have small cable cross-sections and high line impedance, and their average power levels account for a low proportion in the alliance, the line loss costs that should have been borne by MG1 and MG5 are appropriately reduced by MG2, MG3, and MG4 members by reducing a certain percentage of their own revenue, thereby increasing the revenue of MG1 and MG5, stabilizing their participation in the alliance, and aiming to maximize the total revenue of the alliance.

[0117] Table 2. Average power percentage of members and line specifications and length from PCC to metering point.

[0118]

[0119] Table 3. Revenue sharing weights after improved Shapley value calculation

[0120]

[0121] It can be seen that differentiated revenue distribution can more objectively and fairly reflect the line losses caused by the physical and electrical characteristics of the existing grid structure in microgrid clusters, through the loss sensitivity factor λ. iAdjusting the revenue sharing among different microgrid clusters can enhance the enthusiasm of each member to participate in alliance cooperation. At the same time, it can play a positive role in improving the utilization efficiency of renewable energy and reducing the daily operating costs of microgrid clusters in uncertain source-load scenarios, thus helping to promote the coordinated development of energy storage and new energy.

[0122] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A robust optimization method for the coordinated operation of microgrids and shared energy storage in uncertain scenarios, characterized in that: Includes the following steps: Step S1: Construct a scheduling optimization operation framework for the coordinated operation of microgrid clusters and shared energy storage; Step S2: Construct a set of uncertain landscape scenes using scene generation technology; Step S3: Construct a set of load uncertainty scenarios based on data-driven K-means clustering; Step S4: Robustly estimate the activation probability of the load disturbance boundary based on Wasserstein distance; Step S5: Construct a two-stage robust scheduling optimization model for microgrid clusters and shared energy storage; Step S6: Solve the two-stage robust optimization model using dual transformation and the C&CG algorithm; Step S7: Design of a revenue sharing mechanism for the improved Shapley value method based on the network topology sensitivity model.

2. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 1, characterized in that: In step S1, the uncertain wind and solar power scenario refers to the scenario where wind turbine power generation and photovoltaic power generation are uncertain. An operation framework considering the power capacity of the shared energy storage operator in the microgrid group source-grid-load system is constructed. This includes: the source-grid-load model in each park of the microgrid group MGs and the shared energy storage operator SESS model. The SESS access point is located at the common connection point PCC. Each microgrid in the microgrid group shares with the SESS through the PCC. Each MGs park contains distributed photovoltaic (PV), wind turbine (PW), controllable distributed power sources, and loads. The introduction of SESS (Support for Energy Storage and Renewable Energy) aims to promote local consumption of renewable energy and peak shaving / valley filling while minimizing the daily operating costs of MGs. First, MGs prioritize consuming renewable energy generation and balance load demand through controllable distributed power sources. Second, MGs reduce unnecessary load demand through internal demand response and incentivize MGs to utilize the power capacity of SESS for peak shaving and valley filling. Finally, MGs purchase electricity from the upper-level grid to meet remaining unbalanced loads.

3. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 2, characterized in that: The operational goal of microgrid groups (MGs) is to aggregate the demand of each subgrid to achieve large-scale demand response while ensuring the reliability of users' energy consumption. At the same time, by combining controllable distributed power sources within the MGs and sharing energy storage capacity through leasing third parties, it can coordinate the response to renewable energy output and load disturbances, thereby minimizing the daily operating cost of the system. SESS's operational objective is to support peak shaving and valley filling of loads within MGs. By constructing centralized shared energy storage power stations at PCC points in various MGs parks, SESS provides leasing services to MGs. SESS's charging and discharging strategies are formulated based on the leasing needs of MGs. According to time-of-use pricing, energy storage capacity is used to store electricity at low levels and release it at high levels for arbitrage.

4. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 3, characterized in that: The specific model of the scheduling optimization operation framework is expressed by the following formula: Formula 1 is the shared energy storage operator model. The discharge and charge power of SESS are both non-negative integers, λ. ssD For discharge identification, λ ssC This is a charging indicator; The maximum discharge / charge power that SESS can provide; NT represents the number of zones in the microgrid cluster, T is taken as 24 hours, and η is the charging and discharging efficiency; and To impose upper and lower limits on the amount of content in SESS during different time periods under the discharge / charging states during the sharing process. This means that SESS can share the initial power. Δt is the daily scheduling step size, which takes the value of 1 hour. t = 1 and t' represent the time interval from the first scheduling time point to any subsequent time point. Quantifying the cost of SESS charge / discharge degradation k is the battery degradation coefficient. ss (t) represents the charging and discharging cost of SESS, and k prof SESS's expected revenue; Formula 2 is the controllable distributed power source model, P g (t) represents the sum of the outputs of the gas turbine and the heat pump at time t, P i,rq η is the rated output power of the gas turbine / heat pump within the park. rq For gas turbine efficiency, η rb For heat pump efficiency, c gs The gas purchase price is (yuan / kW); P G.max P G.min These are the upper and lower limits of the output of controllable distributed power sources within a microgrid group; Formula 3 is the demand response model for microgrid groups, P DR (t) represents the demand response load power within the microgrid group, S DR The total scheduling capacity for demand response load from the first point in time to a certain point in time; The minimum and maximum power demand of the demand response load for a certain period of time. Formula 4 is the interaction model between the microgrid group and the upper-level distribution network. Let t be the power purchased by the microgrid group from the distribution network at time t; P represents the power sold by the microgrid group to the distribution network at time t. L (t) represents the active load pre-connected to the microgrid group at time t; P DPS (t) represents the output power of distributed renewable energy sources within the microgrid group at time t; P MG.max This represents the upper limit of the power that can be transmitted by the incoming line of the upper-level distribution network; λ PUR , λ SEL Indicates the purchase and sale of electricity; Formula 5 represents the overall objective function and the objective functions of each subsystem for the collaborative scheduling optimization of microgrid clusters and shared energy storage operators. C ss (t) represents the total cost of leasing a SESS to a microgrid cluster; C g (t) represents the total cost of a controllable distributed power source, a g C represents the linear first-order coefficient of a controllable distributed power source, where b is a known constant; DR (t) represents the demand response cost, k DR This is the demand response cost coefficient; C gtax (t) represents the transmission and distribution costs incurred by SESS using the existing power grid; C INR (t) represents the cost of interacting with the distribution network.

5. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 3, characterized in that: The method for implementing step S2 is as follows: Assuming that the PV power prediction error follows a normal distribution, the wind turbine PW power error conforms to a Beta distribution. The PV and PW power prediction models are as follows: In formula 6, Daytime power correction value during period t under scenario s; This is the predicted value for the previous day; ξ j R represents the prediction error. j,t is a predicted random number that follows a normal distribution; r is a correction factor; the subscript j represents any photovoltaic power output unit; μ and σ are the expectation and variance values ​​of the prediction model; α and β are the shape and scale parameters of the Beta distribution, respectively; Based on the above prediction model, typical scenes are generated by Latin hypercube sampling. K-means clustering is used to reduce the typical power output scene based on the sampling. Finally, the upper and lower limits of the prediction deviation of the landscape scene are calculated based on the clustered scenes.

6. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 3, characterized in that: The implementation method of step S3 is as follows: the load is determined by the elbow rule based on historical data, the L2 norm between each sample is calculated by K-means iteration to obtain each stable cluster center, and the volatility evaluation index of the samples in each cluster is calculated. The upper and lower limits of the load deviation in each cluster scenario are output by the proportion of volatility. In Formula 7, x i For the i-th sample, μ k(i) The mean vector represents the cluster centers, K is the number of cluster centers, and C is the mean vector. k Represents the k-th cluster, δ i Let d be the sign deviation function. i For L2 norm, SVP i Used to calculate the proportion of sample fluctuation, i.e. the fluctuation boundary of the load; where sign(·) is the sign function, used to preserve the deviation direction between the sample and the cluster center.

7. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 3, characterized in that: The implementation method of step S4 is as follows: An adjustment factor is generated using an activation probability estimation method driven by Wasserstein distance and historical data to reconstruct the number of fluctuations under typical load scenarios, which is used to construct an activation probability model for uncertain factors with data-driven interpretability; the Wasserstein distance constraint modeling formula is as follows: In the above formula, τ (t) Representative of samples in the scene-based clustering results Samples at the right-tailed activation threshold calculated using IQR (Interquartile Range) are used with Quantile. 0.75 (·)express; This corresponds to the indicator function after activation; Let represent the supremum estimate of the expectation of the indicator function, where For the empirical probability distribution of the sample, The most unfavorable probability distribution of the sample is given, where W1 represents the first-order Wasserstein distance, θ is the radius of the Wasserstein sphere used to control the size of the uncertain set, and U... θ Represents a spherically uncertain set, containing all sets with the same empirical distribution. The distribution with a distance not exceeding θ; Equation 11 represents the linearization transformation of Equations 8 to 10 under the condition of Lipschitz continuity, transforming the discrete typical discretized samples into a linear programming form for solution, where p i d represents the probability weight of the i-th sample in the most unfavorable probability distribution. i For each sample x i Wasserstein distance to the center of the empirical distribution.

8. The robust optimization method for microgrids and shared energy storage in uncertain scenarios according to claim 3, characterized in that: The implementation method of step S5 is as follows: In the first stage, a three-level objective function model of min-max-min is constructed, as shown below; In Formula 12, x represents the decision variable of the main problem, y represents the decision variable of the subproblem, and u represents the uncertain scenario belonging to the uncertain scenario set U; In the second stage, MGs initializes the main problem based on the load and new energy output forecast values, and solves the main problem. The objective function is the minimum daily operating cost of MGs. The lower bound LB of the main problem is calculated. The decision variable x obtained from the upper layer is then passed to the lower layer. In the third stage, firstly, based on the upper and lower limits of the deviation range of the uncertain set of wind and solar loads in step S2, deviation interval constraints are imposed on the inner layer of unfavorable conditions. Secondly, an inner layer wind and solar uncertainty adjustment factor is set. Thirdly, a load fluctuation adjustment factor is generated using the activation probability estimation method driven by historical data in step S3. Fourthly, after transforming the inner layer problem (max-min problem) into a max single-layer problem through duality, the objective function value of the inner layer under the most unfavorable conditions is calculated, and the calculated LB and uncertain scenario variable u of the inner layer are returned to the main problem. The main problem then solves for MB and decision variable x again. Step four is repeated until UB-LB converges to the preset residual value requirement, at which point the iteration exits.

9. A robust optimization method for microgrids and shared energy storage in uncertain scenarios as described in claim 3, characterized in that: The implementation method of step S6 is as follows: the two-stage robust inner-layer max-min problem is transformed into a single-layer max problem through dual transformation and KKT conditions, and the C&CG algorithm is used to solve the two-stage model.

10. A robust optimization method for microgrids and shared energy storage in uncertain scenarios as described in claim 3, characterized in that: The implementation method of step S7 is as follows: Based on the two-stage robust optimization model created in step S5 with the lowest daily operating cost of MGs and shared energy storage cooperation as the objective function, an improved Shapley value revenue allocation method based on network topology sensitivity is proposed to solve the revenue allocation problem caused by the different transmission distances of different members of the microgrid group from the point of common connection (PCC) in the grid; the specific model is as follows: In formula 13, MG in microgrid cluster i The marginal contribution of a member, v(s) represents the characteristic function of the revenue of a microgrid group member, and v(s\i) represents the marginal contribution of a member without considering a certain MG. i The benefits of the alliance; n is the MG in the microgrid i The total number, v(n) is the sum of the alliance's profits, x i For the profit sharing among all alliance members, v(i) represents the profit shared by MG, which is the only member in the alliance. i Alliance revenue at that time; s represents the MG in the current calculation alliance. i The number of members, w(|s|) represents the weight coefficient under this condition when the number of members in the alliance is s; In Formula 14, U is the nominal value of the voltage at the point of common coupling, and P... MGi For each MG i The average active load of the members, R MGi For each MG i The impedance parameters within a member have nominal values, R. Li For the line impedance parameters of the corresponding member and the PCC point, P' MGi For the adjusted MG i The active load of the member, R' MGi For the adjusted MG i The impedance parameter within a member has a nominal value, λ. i This is the loss sensitivity factor.

Images