Shared energy storage microgrid dispatching method based on power access quota and marginal exchange

Abstract

The application discloses a shared energy storage micro-grid scheduling method based on power access quota and marginal exchange, and relates to the technical field of micro-grid energy management. The method collects wind, light, load, electricity price, emission factor and energy storage state data of multiple micro-grids, constructs a wind, light and load experience uncertainty set based on historical complete day samples; the sum of charging power and discharging power of shared energy storage of each micro-grid is used to represent interface concurrent occupation, as a first stage power access quota, and a second stage tracing scheduling model containing charging and discharging mutual exclusion restriction is established under unified shared energy storage state of charge; the marginal information of the tracing model is used for budget keeping exchange, a more unfavorable source and load scene is generated, and the power quota is corrected and fed back. The application does not allocate a fixed physical battery capacity to a single micro-grid, converts shared energy storage access right into executable and balanceable interface power limit, improves new energy consumption and reduces tail risk of extreme scenes.

Description

Technical Field

[0001] This invention relates to the field of microgrid energy management technology, and in particular to a shared energy storage microgrid scheduling method based on power access quotas and marginal exchange. Background Technology

[0002] In scenarios where a high proportion of renewable energy is integrated into the distribution system and the integrated energy system of industrial parks, the output of wind and solar power is significantly affected by meteorological conditions, and the load of industrial parks is also characterized by time-varying, random, and different energy-consuming entities. If only deterministic forecasting is used for day-ahead dispatching, problems such as renewable energy curtailment, insufficient energy storage reserves, grid interaction power exceeding limits, or increased dispatching costs may easily occur in actual operation.

[0003] Shared energy storage pools energy storage capacity, converters, and interface capabilities, providing shared flexibility resources for multiple microgrids. However, when multiple entities simultaneously access shared energy storage, key issues in engineering applications include how to define the interface rights available to each microgrid, how to avoid over- or under-reservation, and how to arrange executable scheduling plans under uncertain source-load conditions.

[0004] Existing research on shared energy storage typically employs fixed capacity allocation, energy leasing, cloud service rights, or deterministic collaborative scheduling methods. Fixed capacity leasing allocates energy capacity or physical capacity slices, power reservation generally only describes reserved power, and point-to-point quotas are mostly determined based on energy contribution or transaction volume. None of these methods explicitly characterize the concurrent access rights of microgrids to shared energy storage converters and interfaces during the same time period, nor do they link this access right with the unified state of charge pool and the subsequent recursive scheduling after scenario implementation. On the other hand, simply constructing extreme scenarios based on electricity price levels can easily overlook periods of interface shortages, limited power balance, or significant penalties for no power supply, resulting in insufficient stress testing for robust scheduling. Summary of the Invention

[0005] The purpose of this invention is to provide a shared energy storage microgrid scheduling method based on power access quotas and marginal exchange, in order to solve the problems in the prior art, such as the difficulty in converting shared energy storage access rights into executable interface limits, the unclear concurrent access rights of each microgrid under a unified state of charge pool, the difficulty in identifying scenarios with tight system margins by simple price sorting, and the insufficient physical executability of scheduling results.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] Dispatch methods for shared energy storage microgrids based on power access quotas and marginal exchange include: Collect and preprocess data on available wind power, available photovoltaic power, load demand, electricity price, emission factors, and energy storage status of multi-microgrids; Based on historical full-day data, an empirical uncertainty set of wind, solar and load is constructed, and a budget parameter set by hour is introduced to constrain the activation degree of adverse deviations within the same hour. The concurrent power occupancy of each microgrid's interface to the shared energy storage pool is used as the first-stage power access quota. The occupancy is equal to the sum of the shared energy storage charging power and discharging power of the microgrid in any scheduling period, and the shared energy storage pool maintains a unified state of charge based on the total charging and discharging of all microgrids. After the uncertainty of wind and solar load is realized, a second-stage tracing scheduling model is established to satisfy power balance, energy storage evolution, charge state boundary, interface occupancy and charge-discharge mutual exclusion constraints. Under a given power access quota, the marginal information of the tracing model is extracted. While keeping the budget constant, the activated and inactive deviations are swapped to generate extreme scenarios. The quota decision and scenario cut plane are iteratively updated until the stopping condition is met. Output shared energy storage power quotas, collaborative scheduling plans and operating costs, renewable energy consumption and risk indicators.

[0008] Furthermore, the shared energy storage pool does not allocate a fixed physical battery capacity or independent energy inventory to a single microgrid; the second-stage pursuit scheduling model uses a closed-loop pursuit variable family as decision variables, which includes grid power purchase and sale, renewable energy utilization power and abandoned power, local energy storage charging and discharging power, shared energy storage charging and discharging power, energy storage energy status, inter-microgrid exchange power, and power not supplied, wherein energy storage energy is a state variable, and power not supplied is fixed at zero in embodiments where load shedding is not allowed.

[0009] Furthermore, the power access quota satisfies the upper and lower bound constraints of the quota, and the sum of the shared energy storage charging power and discharging power of any microgrid in any scheduling period does not exceed its power access quota; the total charging power and total discharging power of the shared energy storage pool do not exceed the limit of the pool-level converter, and the unified state of charge evolves according to the charging and discharging power and corresponding efficiency of all microgrids; the tracing scheduling model also satisfies the terminal energy regression constraint and the inter-microgrid exchange power constraint.

[0010] Furthermore, the empirical uncertainty set is obtained through the following steps: screening historical complete days that are valid at each scheduling point after data quality checks; grouping by hour and variable category and performing quantile truncation; performing resampling with replacement on a complete day basis to retain intra-day temporal correlation and intervariate correlation; forming a nominal trajectory and empirical envelope based on the median and upper and lower quantiles of the resampled trajectory, and taking the larger of the distances from the nominal value to the upper and lower boundaries as the symmetrical maximum deviation.

[0011] Furthermore, the budget parameter is a Bertsimas-Sim type uncertain budget set on an hourly basis, used to limit the normalized activation total of at least one adverse deviation among load demand increase, photovoltaic available power decrease and wind power available power decrease within the same hour; when the budget parameter is zero, it degenerates into deterministic scheduling, and when the budget parameter increases, the robustness and conservatism are improved.

[0012] Furthermore, the objective function of the second-stage pursuit scheduling model includes at least one of the following: grid transaction cost, shared energy storage quota leasing cost, energy storage charging and discharging degradation cost, electricity purchase carbon emission cost, microgrid inter-exchange cost, and non-power supply penalty term; wherein the electricity purchase carbon emission cost is determined by summing the products of carbon cost coefficient, grid average operating emission factor during the time period, electricity purchase power, and scheduling duration.

[0013] Furthermore, the extraction of marginal information from the tracing model includes: solving the second-stage tracing linear programming or its continuous relaxation model under a given power access quota, extracting the dual multipliers of shared energy storage interface constraints, shared energy storage pool-level power constraints, microgrid power balance constraints, grid power purchase constraints, or non-power supply penalty-related constraints, and rearranging the degree of adverse effects of uncertain variables accordingly.

[0014] Furthermore, the exchange of activated and inactive deviations includes: under the premise of keeping the same hourly uncertainty budget unchanged, exchanging activated and inactive deviation variables according to the rearrangement results of marginal information, and resolving the tracing model for the exchange candidates; accepting the exchange and forming a new extreme scenario only when the tracing cost or risk indicator increases after the exchange.

[0015] Furthermore, the iterative quota update decision and scenario cut plane are performed under the column and constraint generation framework, which includes a constrained master problem and scenario subproblems. The constrained master problem determines the power access quota, and the scenario subproblems generate unfavorable wind and solar loads and corresponding tracing scheduling under a given quota through marginal information extraction and budget-maintaining exchange, and add the scenario and its cut plane to the constrained master problem. The scenario subproblems of the entire scheduling cycle can be implemented heuristically, and the output is a robust scheduling scheme based on the generated set of unfavorable scenarios.

[0016] Furthermore, the method also includes out-of-sample stress testing of deterministic quota, robust quota, and fixed quota strategies; the deterministic quota is the power access quota obtained when the nominal trajectory or budget parameter is zero, the robust quota is the power access quota obtained under a set uncertain budget, and the fixed quota strategy includes proportional quota or quota allocated according to load ratio; the evaluation indicators include average cost, quantile cost, conditional risk value, renewable energy curtailment, unsupplied power, and the occurrence rate or duration of load shedding events.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: First, this invention defines the shared energy storage access right as an interface concurrent power quota, which limits the interface occupancy at the same time period by the sum of charging power and discharging power, and does not split the fixed physical capacity under a unified state of charge pool. This transforms the shared energy storage access right into an executable, settlementable, and issueable interface power quota, overcoming the shortcomings of fixed capacity leasing and point-to-point energy quotas, which do not explicitly depict concurrent occupancy rights.

[0018] Secondly, this invention describes wind and solar load fluctuations without relying on precise probability distributions by resampling historical complete days and budget uncertainty sets, taking into account intraday correlation, intervariate correlation and robust conservatism, and avoiding the problems of increased curtailment and insufficient reserves in deterministic prediction scheduling under source-load deviations.

[0019] Furthermore, this invention separates quota reservation from post-scenario scheduling, which avoids excessive reservation of shared energy storage interfaces and retains the ability to adaptively adjust to wind and solar load deviations during operation. Combined with charging and discharging mutual exclusion restrictions and executability correction, it ensures that the issued scheduling instructions are physically executable.

[0020] Then, this invention utilizes the marginal information of the pursuit model to identify interface occupancy, power balance, and power outage bottlenecks, and generates more destructive extreme scenarios through budget-maintaining switching, overcoming the inadequacy of simple price sorting in identifying periods of tight system margin; quota decision-making and extreme scenario generation thus form an iterative linkage, significantly improving the sufficiency of robust scheduling stress testing.

[0021] Finally, this invention can output power quotas, scheduling curves, risk indicators, and visualization results, and can be embedded into multi-entity collaborative energy management software modules to provide decision support for the integrated operation of wind, solar, and energy storage at the park level. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall process of the method of the present invention; Figure 2 This is a schematic diagram of the shared energy storage microgrid power quota collaborative architecture of the present invention; Figure 3 This is a schematic diagram of the solution process for edge exchange and column and constraint generation in this invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments are used to explain the present invention, not to limit the scope of protection. Those skilled in the art can make equivalent substitutions for the construction method of the uncertainty set, the form of the optimization model, and the solution algorithm without departing from the technical concept defined in the claims.

[0024] Terminology Explanation Shared energy storage: An energy storage operation model that pools energy storage capacity, converters, and interface capabilities, allowing multiple microgrids to share the resources without splitting physical capacity slices.

[0025] Power access quota: The maximum concurrent usage limit of the shared energy storage charging and discharging interface of a microgrid during any scheduling period. It is equal to the maximum sum of the charging power and discharging power of the shared energy storage of the microgrid, which is different from the description of energy capacity allocation or reserved power.

[0026] State of charge (SOC): The ratio of the current energy of the energy storage to its rated energy. Shared energy storage pools maintain a uniform SOC based on the total charging and discharging power of the entire microgrid.

[0027] Empirical uncertainty set: a description of the range of values ​​of uncertain variables constructed from historical complete daily samples through quantile truncation, resampling with replacement, and quantile envelope.

[0028] Budget uncertainty set: The set of all deviation dimensions that are simultaneously on the unfavorable frontier, constrained by budget parameters. The larger the budget parameter, the higher the robustness and conservatism.

[0029] Pursuit scheduling: The second-stage scheduling that solves for specific scenarios after the realization of uncertainties, including pursuit variables such as power purchase and sale, energy storage charging and discharging, microgrid mutual assistance, and power outage.

[0030] Marginal information: Dual multipliers of constraints in the optimization model, reflecting the impact of unit changes in the right-hand side of the constraint on the objective function, and used to identify bottleneck constraints.

[0031] Budget swapping: A mechanism that swaps activated and inactive bias variables to construct more unfavorable scenarios without changing the total amount of uncertain budget for the same hour.

[0032] Column and Constraint Generation: A two-stage robust optimization solution framework that alternately solves the constrained master problem and scene subproblems and incorporates the scene cutting plane into the master problem.

[0033] Conditional Value at Risk (VaR): The expected loss at a given confidence level in the tail of the loss distribution, used to evaluate the severity of risk in extreme scenarios.

[0034] Out-of-sample stress test scenarios refer to stress test scenarios that are not included in the set of cut planes of the restricted principal problem, are independently sampled from the same set of uncertain budgets, and are used to evaluate the performance of deterministic quota, robust quota, or fixed quota strategies in new scenarios.

[0035] like Figure 1As shown, the method of this invention includes six stages: data access and preprocessing, construction of an empirical uncertainty set, first-stage power access quota decision, second-stage tracing and collaborative scheduling, generation and iterative solution of marginal exchange scenarios, and output of scheduling scheme. Each stage is executed sequentially and the quota decision is corrected through scenario iteration feedback. Figure 2 As shown, each microgrid includes photovoltaic, wind power, load, and local energy storage, and is connected to the shared energy storage pool through a shared energy storage access interface; the shared energy storage coordinator is responsible for allocating the power access quota of each microgrid and maintaining a unified state of charge; each microgrid is connected to the main grid via a power purchase and sale interface, and a certain capacity of power exchange is allowed between microgrids.

[0036] 1. System Objects and Data Access Let the multi-micronet set be The set of scheduling periods is Each microgrid includes photovoltaic (PV), wind power, loads, local energy storage, and a power purchase and sale interface with the distribution network. Each microgrid connects to a shared energy storage pool through a shared energy storage access interface. The shared energy storage pool has a unified state of charge. The microgrid does not own fixed physical battery cells, but instead uses the shared energy storage's charging and discharging interface through power access quotas.

[0037] The system collects or reads the following data for each microgrid at each time period: available photovoltaic power, available wind power, load demand, electricity purchase price, grid connection price or electricity sales discount, average grid operation emission factor for each time period, local energy storage capacity and power constraints, shared energy storage pool capacity and power constraints, inter-microgrid exchange limits, and operating cost parameters. Data preprocessing includes missing value repair, outlier truncation, time zone or sampling interval standardization, aggregation by hour or scheduling interval, and screening of historical complete daily samples.

[0038] A complete day refers to a natural day in which wind power, solar power, and load data are valid throughout the entire scheduling period after timeline alignment and data quality checks; natural days that cannot be reliably repaired or have continuous long-term missing data are excluded.

[0039] 2. Construction of Empirical Uncertainty Sets for Wind-Solar-Herbs For microgrids During the period Given the available photovoltaic power, available wind power, and load demand, construct an uncertain variable vector. The uncertain variables in the entire time domain are obtained by concatenating all microgrids, all time periods, and all source-load variables.

[0040] Historical full-day samples were grouped by time period and variable category. Each group was truncated using empirical 1% and 99th percentiles to mitigate the impact of anomalous observations. Resampling with replacement was then performed on a full-day basis to preserve intraday temporal correlations and the correlation between wind, solar, and load. The nominal trajectory was obtained based on the 50th percentile of the resampled trajectory, and the empirical lower and upper boundaries were obtained based on the 5th and 95th percentiles. The deviation radius was taken as the larger of the distance from the nominal value to the lower boundary and the distance from the nominal value to the upper boundary.

[0041] Introducing budget parameters Construct a set of budget uncertainty, where The constraint form satisfied by the normalized deviation variable. Let be the normalized bias in the k-th dimension. The budget uncertainty set restricts the total activation of normalized adverse biases to not exceeding the budget parameter. Degenerates into deterministic scheduling; increases Allowing more time periods or variables to operate on unfavorable boundaries can improve robustness but may sacrifice economics. For wind-solar-storage microgrids, typical unfavorable directions are load increase, decrease in available photovoltaic power, and decrease in available wind power.

[0042] In a preferred embodiment, the budget parameter is set hourly. For three microgrids, each hour includes three types of variables: photovoltaic, wind power, and load, totaling nine uncertain dimensions; the hourly budget parameter can be set to... This indicates that the total number of normalized adverse bias activations does not exceed 3 within the same hour.

[0043] 3. First-phase power access quota decision Before the uncertainty materializes, the power access quota for each microgrid is determined by the shared energy storage coordinator. . microgrid The upper limit of interface concurrent usage is represented by the sum of the shared energy storage charging power and discharging power at any given time period, rather than the microgrid. The fixed energy capacity it possesses.

[0044] Quotas should meet If necessary, it also meets the following requirements. This is to prevent microgrids from over-reserving the limited shared interfaces of energy storage converters. During the period The shared energy storage charging and discharging power should meet the following requirements. Meanwhile, the total charging and discharging power of the shared energy storage pool does not exceed the capacity of its converter.

[0045] Power quotas can be divided into two modes: guaranteed and non-guaranteed. The guaranteed mode is suitable for scenarios where all microgrid reserved interfaces are required to fulfill their obligations simultaneously, and the sum of all microgrid quotas does not exceed the available inverter capacity of the shared energy storage pool or its reserved ratio. The non-guaranteed mode is suitable for scenarios where the probability of multiple microgrids operating at full power simultaneously is low and the goal is to improve the utilization rate of pooled resources. It allows the sum of quotas to exceed the pool-level capacity, but the actual total charging power and total discharging power at any given time are still constrained by the pool-level inverter capacity, and the coordinator allocates available interfaces according to priority, scenario cost, or deviation settlement rules.

[0046] Quota cost can be expressed as ,in The power quota rental price or converted cost, This refers to the number of days for scheduling. A larger number... This enhances the microgrid's ability to adapt to uncertainty, but increases quota leasing costs; smaller... It reduces the reserved cost, but may limit the response capability of shared energy storage in adverse scenarios.

[0047] The quota leasing cost and the second-stage operating cost both adopt the same optimization time domain. The unit of power quota leasing price is RMB per kilowatt-day, which can be determined by the quotation of the shared energy storage operator, or it can be calculated based on the energy storage investment cost, operation and maintenance cost, life depreciation and reasonable return; in this embodiment, RMB 0.10 per kilowatt-day is used.

[0048] 4. Second-stage tracking and coordinated scheduling Once the uncertainty of wind, solar, and load conditions is actually realized or given by the sub-problems of the scenario, the second stage is... , Together with the curtailed power of photovoltaic power, curtailed power of wind power, and unsupplied power, they constitute a closed-loop family of variables; in embodiments where load shedding is not permitted, the unsupplied power is fixed at 0.

[0049] Specifically, the closed-loop tracing variable family includes: grid power purchases. Power sold by the power grid Photovoltaic power utilization Wind power utilization Solar power curtailment, wind power curtailment, and local energy storage charging power. Local energy storage discharge power Shared energy storage charging power Shared energy storage discharge power Local energy storage Shared energy storage pool Inter-microgrid switching power And the power not supplied; where the energy storage energy is a state variable, and the power not supplied is a relaxation variable in the reliability diagnosis or allowable load shedding embodiment.

[0050] The constraints for tracing dispatch include: constraints on the decomposition of available renewable energy power, constraints on microgrid power balance, upper and lower limits on grid power purchase and sale, constraints on local energy storage charging and discharging power, charging and discharging mutual exclusion and state of charge, constraints on the evolution of the state of charge of shared energy storage, constraints on mutual exclusion of charging and discharging of shared energy storage, constraints on the return of terminal energy, constraints on the exchange power between microgrids, and constraints on the shared energy storage interface not exceeding the first phase quota.

[0051] The target objectives for dispatching include grid power purchase costs, electricity sales revenue, energy storage charging and discharging degradation costs, shared energy storage quota costs, carbon emission costs of power purchases, inter-microgrid exchange costs, and penalties for power outages when necessary. Carbon emission costs of power purchases are determined by summing the products of the carbon cost coefficient, the average grid operating emission factor for the time period, the purchased power, and the dispatch duration. Shared energy storage quota costs are determined by the quota leasing price, power access quota, and the number of days for optimized time-domain coverage. Through this objective function, while meeting equipment safety constraints, the aim is to comprehensively reduce operating costs, decrease wind and solar curtailment, suppress peak power purchases, and mitigate the severity of tail risks in extreme scenarios.

[0052] In the basic embodiment, mutual exclusion of energy storage charging and discharging is achieved through 0-1 directional state variables: the discharge power is zero when the state is in the charging direction, and the charging power is zero when the state is in the discharging direction. The tracing model used for marginal information extraction can adopt a continuously relaxed form of this mutual exclusion constraint to improve solution efficiency, but the final dispatch instructions should be corrected by the mutual exclusion constraint model or executability.

[0053] 5. Marginal exchange and column constraint generation solution process To improve the ability to identify adverse scenarios, this method combines the column and constraint generation framework with marginal information extraction and budget preservation exchange. The constrained master problem solves the first-stage power access quota on the generated set of adverse scenarios; the scenario subproblem, under a given quota, uses the dual multipliers or marginal costs of the pursuit model to identify bottleneck constraints such as interface occupancy, power balance, and no power supply, and finds wind and solar load realizations that increase pursuit costs or risk indicators, and adds the scenario and the corresponding cut plane to the constrained master problem.

[0054] The scenario subproblem first constructs a price-weighted unfavorable vertex, such as prioritizing the activation of load increase and renewable energy decrease biases during periods of high electricity prices or tight system margins; then, it solves a pursuit linear programming problem or its continuous relaxation model with a fixed quota to extract dual multipliers and marginal information of key constraints such as power balance, shared energy storage interface occupancy, shared energy storage pool-level power, or no power supply; finally, it maintains budget... Under the condition that remains unchanged, enumerate or filter several candidates for exchanging activated and inactive biases. If the retrieval cost increases after the exchange, then accept the exchange.

[0055] like Figure 3 As shown, the solution process for edge exchange and column and constraint generation starts from the initialization of the nominal scenario, goes through the solution of the constrained main problem, the solution of the scenario subproblems, the calculation of relative differences and the judgment of stopping conditions. When the stopping conditions are not met, the scenario and the cutting plane are added to the main problem and iterated. After the stopping conditions are met, the robust quota and scheduling scheme is output.

[0056] The iteration stopping condition can be set as follows: the rate of change of the target in adjacent iterations is lower than a threshold; the relative difference between the constrained master problem value and the current scenario value is lower than a threshold; no new extreme scenarios are generated for a certain number of consecutive iterations; or the maximum number of iterations is reached. In engineering implementation, the relative tolerance can be set to... to The maximum number of iterations can be set according to the computing power scale; in the preferred embodiment, a relative tolerance can be adopted. The maximum number of iterations is 6, with the specific value depending on the program settings. Since the entire scheduling cycle can be generated heuristically, the system output is presented as a robust scheduling scheme based on the generated set of unfavorable scenarios, not as an unproven globally robust optimal solution.

[0057] 6. Real-time running interface for the current day During the day-ahead phase, the system determines the power access quota and day-ahead plan for each microgrid based on the forecast baseline, historical experience uncertainty set, and the availability status of shared energy storage. During the intraday phase, the system continuously adjusts and tracks scheduling based on updated wind and solar load forecasts, shared energy storage status of charge, and microgrid operating status. During the real-time phase, the system executes limit correction, peak shaving and valley filling, power curtailment reduction, or load protection strategies when necessary, based on deviation alarms, interface occupancy rates, and status of charge safety zones.

[0058] The software implementation may include a data management module, an uncertainty set construction module, a power quota optimization module, a tracing scheduling module, an extreme scenario generation module, and a visualization output module. These modules are used to execute the data access, set construction, quota optimization, tracing scheduling, extreme scenario generation, and result output steps in the aforementioned methods, respectively. The extreme scenario generation module is responsible for marginal information extraction and budget-preservation exchange, while the visualization output module integrates out-of-sample stress testing functionality. Output results include risk indicators such as microgrid quotas, shared energy storage charging and discharging plans, local energy storage charging and discharging plans, grid power purchase and sale plans, inter-microgrid exchange plans, renewable energy consumption, peak power purchase, carbon emissions, cost decomposition, and conditional value at risk.

[0059] Example 1 The following describes the implementation process of this invention using a three-microgrid park-level integrated wind, solar, and energy storage scenario. This scenario includes three park microgrids, a shared energy storage pool, and a main grid interface. Each microgrid can be configured with photovoltaic, wind power, load, and local energy storage, and allows for a certain capacity of inter-microgrid power exchange. The scheduling interval can be 1 hour, and the scheduling cycle can be 24 hours; in the extended embodiment, a 168-hour scheduling cycle can be used.

[0060] In this embodiment, wind power, photovoltaic, and load data from January to March 2025 are first read, 90 complete days are selected, and 2160 hourly observation points are formed. For each hourly variable of the same type, empirical 1% and 99% quantiles are truncated. Resampling with replacement is performed on a fixed random seed of 2025, using complete days as units, generating 512 complete day trajectories for wind, photovoltaic, and load. Empirical upper and lower boundaries are formed using the 5% and 95% quantiles, and a symmetrical deviation radius is taken. Then, budget parameters are set hourly. Control the degree of activation of adverse deviations within the same hour.

[0061]

[0062] Secondly, establish the first-stage quota model. Quota variables. microgrid The concurrent usage limit for shared energy storage charging and discharging interfaces. In the guaranteed mode, the sum of the quotas of all microgrids is limited by the pool-level limit; in the non-guaranteed mode, statistical over-selling is allowed, but in actual retrieval scheduling, the sum of the shared energy storage charging and discharging power of each microgrid must not exceed its quota, and the actual total charging power and total discharging power of all microgrids must not exceed the pool-level converter limit.

[0063] Next, a second-stage tracing model is established. For any wind-solar-load scenario, the basic implementation solves a mixed-integer linear scheduling model that satisfies power balance, renewable energy utilization, local energy storage state of charge, shared energy storage state of charge, terminal energy regression, power purchase and sale ceiling, microgrid mutual assistance ceiling, shared energy storage interface occupancy, and charge / discharge mutual exclusion constraints. The implementation for marginal information extraction can solve its continuously relaxed model. If grid connection safety, N-1 safety, or voltage constraints are required in engineering, linearized power flow, voltage upper and lower bounds, line power limits, and reserve margin constraints can be added in the second stage.

[0064] Finally, perform marginal swaps and column and constraint generation iterations. Initialize one or more adverse scenarios, solve the constrained master problem to obtain power access quotas; solve scenario subproblems with fixed quotas, and generate new adverse scenarios through price-weighted vertices, marginal cost sorting, and budget-maintaining swaps; add the scenario and its corresponding cut plane to the master problem; repeat the iteration until the stopping condition is met. After completion, output robust quotas and corresponding scheduling schemes, expressed as nominal trajectories or... Based on the determined quota, out-of-sample stress tests are used to evaluate indicators such as average cost, P95 cost, conditional risk value, abandoned power, power outages, and the rate of load shedding events.

[0065] In a budget parameter containing 9 uncertain dimensions In the simplified hourly example, 252 budget polyhedron vertices were precisely enumerated. The unfavorable cost obtained by simple price sorting was 7463.268 yuan, while the precisely enumerated result was 34,734,217.838 yuan. After using marginal cost screening and budget preservation swap, the result was 34,734,217.609 yuan, with a relative difference of approximately [missing information]. It should be noted that this example only addresses a single-hour scenario subproblem, and the listed costs are the recourse costs for that hour under a given quota; the 500 out-of-sample stress test results presented later correspond to the total cost over the entire scheduling cycle, and since they are not based on the same statistical caliber, they are not directly compared in terms of order of magnitude. This result demonstrates that the marginal exchange mechanism can identify joint stress scenarios missed by simple price ranking; the full 24-hour implementation is still considered a heuristic method for scenario generation.

[0066] In the current case study, optimizing shared energy storage access reduces system operating costs and renewable energy curtailment. In the baseline scenario, compared to without shared energy storage, the total cost after optimizing shared energy storage power quotas decreased from RMB 8.535 million to RMB 7.841 million, a reduction of approximately 8.13%; the curtailed electricity decreased from 2.186 GWh to 1.790 GWh, a reduction of approximately 18.1%. In the high-renewable energy scenario, shared energy storage access reduced curtailed electricity from 5.299 GWh to 4.626 GWh, a reduction of approximately 12.69%.

[0067] In 500 out-of-sample stress test scenarios, using nominal trajectory or Based on the deterministic power quota obtained below, the robust quota reduces the conditional risk value penalty cost from RMB 3.437 million to RMB 0.14723 million, the maximum power outage from 137.1095 kWh to 0.01096 kWh, and the average power outage from 1.65 kWh to... Kilowatt-hours. The failure rate for both strategies was 1.2%, indicating that the robust quota in this embodiment primarily reduces the severity of extreme events, rather than eliminating all out-of-sample failure events. The above results support the effectiveness of the examples in this embodiment and do not constitute an absolute performance guarantee for all scenarios.

[0068] Preferably, the uncertainty set can be replaced by the empirical quantile envelope, prediction error interval, scenario clustering interval, chance constraint, or sparse bar set, as long as the budget parameter or risk parameter is still retained to control the conservatism.

[0069] Power access quotas can be expanded into a combined power and energy quota, or combined with fairness factors, historical usage weights, priority factors, and capacity availability confidence bands.

[0070] The second-stage basic model employs a mixed-integer linear programming model with charge-discharge constraints. In embodiments used for rapid screening or marginal information extraction, its continuous relaxation model can be used, combined with positive energy storage degradation costs, efficiency losses, and charge-discharge netting post-processing to reduce the probability of simultaneous charge-discharge solutions. Start-stop constraints, network voltage constraints, and reserve constraints can be used as further extensions.

[0071] The solution algorithm can be implemented using column and constraint generation, Benders decomposition, alternating direction multiplier method coordination, rolling model predictive control, or a combination thereof, where marginal cost screening and budget preservation exchange can be used as extreme scenario generation submodules.

[0072] The application scope can be expanded from three-park microgrids to two or more microgrids, virtual power plants, source-grid-load-storage parks, distribution network areas, or multi-entity systems containing charging stations and data center flexible loads.

[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A shared energy storage microgrid dispatching method based on power access quotas and marginal exchange, characterized in that, include: Collect and preprocess data on available wind power, available photovoltaic power, load demand, electricity price, emission factors, and energy storage status of multi-microgrids; Based on historical full-day data, an empirical uncertainty set of wind, solar and load is constructed, and a budget parameter set by hour is introduced to constrain the activation degree of adverse deviations within the same hour. The concurrent power occupancy of each microgrid's interface to the shared energy storage pool is used as the first-stage power access quota. The occupancy is equal to the sum of the shared energy storage charging power and discharging power of the microgrid in any scheduling period, and the shared energy storage pool maintains a unified state of charge based on the total charging and discharging of all microgrids. After the uncertainty of wind and solar load is realized, a second-stage tracing scheduling model is established to satisfy power balance, energy storage evolution, charge state boundary, interface occupancy and charge-discharge mutual exclusion constraints. Under a given power access quota, the marginal information of the tracing model is extracted. While keeping the budget constant, the activated and inactive deviations are swapped to generate extreme scenarios. The quota decision and scenario cut plane are iteratively updated until the stopping condition is met. Output shared energy storage power quotas, collaborative scheduling plans and operating costs, renewable energy consumption and risk indicators.

2. The method according to claim 1, characterized in that, The shared energy storage pool does not allocate a fixed physical battery capacity or independent energy inventory to a single microgrid; the second-stage tracking scheduling model uses a closed family of tracking variables as decision variables, which includes grid power purchase and sale, renewable energy utilization power and abandoned power, local energy storage charging and discharging power, shared energy storage charging and discharging power, energy storage energy status, inter-microgrid exchange power, and power not supplied, wherein energy storage energy is a state variable, and power not supplied is fixed at zero in embodiments where load shedding is not allowed.

3. The method according to claim 1, characterized in that, The power access quota satisfies the upper and lower bound constraints, and the sum of the shared energy storage charging power and discharging power of any microgrid in any scheduling period does not exceed its power access quota; the total charging power and total discharging power of the shared energy storage pool do not exceed the pool-level converter limit, and the unified state of charge evolves according to the charging and discharging power and corresponding efficiency of all microgrids; the tracing scheduling model also satisfies the terminal energy regression constraint and the inter-microgrid exchange power constraint.

4. The method according to claim 1, characterized in that, The empirical uncertainty set is obtained through the following steps: filtering historical complete days that are valid at each scheduling point after data quality checks; grouping by hour and variable category and performing quantile truncation; Resampling with replacement is performed on a full-day basis to preserve intra-day temporal correlation and intervariate correlation; a nominal trajectory and empirical envelope are formed based on the median and upper and lower quantiles of the resampled trajectory, and the larger of the distances from the nominal value to the upper and lower boundaries is taken as the symmetrical maximum bias.

5. The method according to claim 1, characterized in that, The budget parameter is a Bertsimas-Sim type uncertainty budget set on an hourly basis, used to limit the normalized activation total of at least one adverse deviation among load demand increase, photovoltaic available power decrease and wind power available power decrease within the same hour; when the budget parameter is zero, it degenerates into deterministic scheduling, and when the budget parameter is increased, the robustness and conservatism are improved.

6. The method according to claim 1, characterized in that, The objective function of the second-stage tracking dispatch model includes at least one of the following: grid transaction cost, shared energy storage quota leasing cost, energy storage charging and discharging degradation cost, electricity purchase carbon emission cost, microgrid inter-exchange cost, and non-power supply penalty term; wherein the electricity purchase carbon emission cost is determined by summing the products of carbon cost coefficient, grid average operating emission factor during the time period, electricity purchase power, and dispatch duration.

7. The method according to claim 1, characterized in that, The extraction of marginal information from the tracing model includes: solving the second-stage tracing linear programming or its continuous relaxation model under a given power access quota, extracting the dual multipliers of shared energy storage interface constraints, shared energy storage pool-level power constraints, microgrid power balance constraints, grid power purchase constraints, or non-power supply penalty-related constraints, and rearranging the degree of adverse effects of uncertain variables accordingly.

8. The method according to claim 1, characterized in that, The exchange of activated and inactivated deviations includes: under the premise of keeping the same hourly uncertainty budget unchanged, exchanging activated and inactivated deviation variables according to the rearrangement results of marginal information, and resolving the tracing model for the exchange candidates; the exchange is accepted only when the tracing cost or risk indicator increases after the exchange, thus forming a new extreme scenario.

9. The method according to claim 1, characterized in that, The iterative quota update decision and scenario cut plane are performed under the column and constraint generation framework, which includes a constrained master problem and scenario subproblems. The constrained master problem determines the power access quota, and the scenario subproblems generate unfavorable wind and solar loads and corresponding tracing scheduling under a given quota through marginal information extraction and budget-maintaining exchange, and add the scenario and its cut plane to the constrained master problem. The scenario subproblems of the entire scheduling cycle can be implemented heuristically, and the output is a robust scheduling scheme based on the generated set of unfavorable scenarios.

10. The method according to claim 1, characterized in that, The method further includes out-of-sample stress testing of deterministic quota, robust quota, and fixed quota strategies; the deterministic quota is the power access quota obtained when the nominal trajectory or budget parameter is zero, the robust quota is the power access quota obtained under a set uncertain budget, and the fixed quota strategy includes proportional quota or quota allocated according to load ratio; the evaluation indicators include average cost, quantile cost, conditional risk value, renewable energy curtailment, unsupplied power, and the occurrence rate or duration of load shedding events.

Images