Clean energy microgrid energy storage capacity robust optimization method and system fusing prediction uncertainty

Abstract

The invention provides a clean energy micro-grid energy storage capacity robust optimization method and system fusing prediction uncertainty, and relates to the technical field of clean energy micro-grid planning and operation control. According to the method, a basic data set is constructed by collecting historical operation data and prediction data, a new energy power generation uncertainty model and a load uncertainty model are constructed based on prediction error distribution, and an uncertainty scene set is formed; establishing an energy storage capacity robust optimization model containing the uncertainty scene set, and solving an energy storage capacity optimal configuration scheme under the most unfavorable scene by adopting robust optimization by taking the minimum weighted sum of the capacity cost and the operation risk as an objective function; a data-driven energy management strategy is constructed based on the scheme, model parameters are updated through actual operation data, a planning-operation-feedback robust optimization closed loop is realized, and the prediction deviation resistance and operation reliability of energy storage configuration can be improved.

Description

Technical Field

[0001] This invention relates to the field of clean energy microgrid planning and operation control technology, and in particular to a robust optimization method and system for clean energy microgrid energy storage capacity that incorporates predictive uncertainties. Background Technology

[0002] Currently, my country's integrated energy microgrid construction has entered a stage of rapid development, mostly implemented in small-scale areas such as industrial parks and communities in the form of grid-connected systems. These systems integrate distributed renewable energy generation with new energy storage devices to build stable and efficient local energy supply systems. Regarding energy storage capacity planning and optimization, typical practices currently employ methods such as differential compensation, fluctuation mitigation analysis, and economic characteristic optimization. These traditional methods usually allocate energy storage scale based on empirical forecasts of future load and power generation to meet supply-demand balance and economic requirements under normal operating conditions.

[0003] However, the above approach faces key challenges in microgrids with a high proportion of clean energy integration. Due to the significant intermittency and volatility of renewable energy output such as solar and wind power, and the random variations in load demand, configuring energy storage capacity based on a single forecast often fails to accurately match demand in actual operation, potentially leading to power imbalances, voltage and frequency fluctuations, or even power outages within the microgrid. Furthermore, traditional deterministic optimization is highly sensitive to forecast errors, while stochastic optimization methods rely on precise probability distributions and lack robustness to extreme adverse conditions. This makes it difficult for existing energy storage planning to balance economic efficiency and reliability, and it is significantly inadequate in addressing the uncertainty of renewable energy output forecasts. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a robust optimization method and system for energy storage capacity in clean energy microgrids that integrates prediction uncertainty. By constructing an uncertainty model of new energy power generation and load prediction and embedding it into the robust optimization framework of energy storage capacity, the resistance of energy storage planning to prediction bias and extreme disturbances is improved.

[0005] To achieve the above objectives, the present invention provides the following solution: A robust optimization method for energy storage capacity in clean energy microgrids that incorporates predictive uncertainties includes: S1. Collect historical operation data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset; S2. Analyze the prediction error distribution based on the historical operating data and the prediction data, construct a new energy power generation uncertainty model and a load uncertainty model based on the prediction error distribution, and generate a set of uncertainty scenarios. S3. Establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. Incorporate the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity. Use the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. S4. The robust optimization solution method is used to solve the robust optimization model of the energy storage capacity. The objective function is used as the solution basis. By iteratively searching for the most unfavorable scenario and adjusting the energy storage capacity decision, the robust optimization solution for the prediction deviation is achieved, and the optimal configuration scheme of the energy storage capacity is obtained. S5. Configure the energy storage device capacity according to the optimal configuration scheme, and build a data-driven energy management strategy based on the basic dataset. Adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. S6. Collect actual operating data, update the parameters of the new energy power generation uncertainty model, the load uncertainty model and the energy storage capacity robust optimization model, and repeat steps S2 to S5 to improve the robustness of energy storage capacity configuration.

[0006] Preferably, the historical operating data includes the measured power of renewable energy generation, the measured power of load, the power of external grid interaction, and the state of charge of energy storage in multiple historical scheduling periods of the clean energy microgrid; the forecast data includes the forecast power of renewable energy generation and the forecast power of load in each scheduling period.

[0007] Preferably, in step S2, the analysis of the prediction error distribution includes: The difference between the predicted power of new energy generation and the corresponding measured power of new energy generation, as well as the difference between the predicted power of load and the corresponding measured power of load, are calculated to obtain the time series of error in new energy generation prediction and the time series of error in load prediction. Statistical analysis is performed on the time series of the load prediction error to obtain the empirical probability distribution of the prediction error in each scheduling period, and the empirical probability distribution is used as the prediction error distribution.

[0008] Preferably, in step S2, the uncertainty model of new energy power generation and the uncertainty model of load are constructed based on the prediction error distribution. Specifically, the upper and lower bounds of the prediction error of new energy power generation and the prediction error of load are determined according to a given confidence level in each scheduling period. The upper and lower bounds are superimposed with the prediction power of new energy power generation and the prediction power of load to obtain the power range of new energy power generation and the power range of load. The uncertainty scenario set is composed of the power trajectory set formed by the combination of the power range of new energy power generation and the power range of load in each scheduling period.

[0009] Preferably, in step S3, the operational risk is obtained by calculating the power fluctuation amplitude of grid interaction, the amount of wind and solar power curtailment, and the amount of unmet load demand under each scenario within the set of uncertain scenarios, and then performing a weighted summation. The objective function is a linear weighted sum of the annualized cost of capacity investment corresponding to the rated capacity and rated power of the energy storage device and the operational risk.

[0010] Preferably, in step S3, the microgrid power balance constraint is that, under any scheduling period and any uncertainty scenario, the sum of the output of distributed power sources, the charging and discharging power of energy storage, and the interaction power of the external power grid equals the load power; the energy storage charging and discharging constraint is that the charging and discharging power of energy storage is mutually exclusive in any scheduling period and the energy storage state of charge changes within a preset upper and lower limit; the equipment operation constraint is that the output of distributed power sources and the tie-line power operate within the rated capacity range.

[0011] Preferably, in step S4, the robust optimization solution method adopts a column and constraint generation algorithm that alternates between the main problem and sub-problems; the main problem optimizes the energy storage capacity decision under the current finite uncertainty scenario and outputs candidate energy storage capacity schemes; the sub-problems search for the most unfavorable scenario with the strictest constraints on the candidate energy storage capacity schemes within the value range corresponding to the uncertainty scenario set, and add the most unfavorable scenario to the main problem until there are no new most unfavorable scenarios.

[0012] Preferably, in step S5, the data-driven energy management strategy adopts a rolling time-domain optimization method. At the beginning of each rolling optimization period, the optimal configuration scheme of the basic dataset and the energy storage capacity is used to construct a short-term operation optimization model, solve for the energy storage charging and discharging power command and external grid interaction power reference value for each scheduling period in the current rolling optimization period, and correct the energy storage charging and discharging power command according to the real-time prediction deviation during actual operation.

[0013] Preferably, in step S6, the actual operating data collected within the preset update cycle is incorporated into the basic dataset. When the scale or statistical characteristics of the actual operating data meet the preset update conditions, steps S2 to S4 are re-executed to obtain the updated optimal configuration scheme of energy storage capacity, and the updated optimal configuration scheme of energy storage capacity is adopted in the subsequent rolling time-domain optimization. The clean energy microgrid includes at least one new energy power generation unit, at least one set of energy storage devices, and at least one type of dispatchable load.

[0014] A robust optimization system for clean energy microgrid energy storage capacity that integrates predictive uncertainty includes: The data acquisition and construction module is used to collect historical operating data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset. The uncertainty modeling module is used to analyze the prediction error distribution based on the historical operating data and the prediction data, and to construct a new energy power generation uncertainty model and a load uncertainty model based on the prediction error distribution, with a set of uncertainty scenarios. The robust optimization modeling module is used to establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. It incorporates the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity, and takes the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. The robust optimization solution module is used to solve the robust optimization model of the energy storage capacity using a robust optimization solution method. Based on the objective function, it iteratively searches for the most unfavorable scenario and adjusts the energy storage capacity decision to achieve robust optimization solution for prediction deviation and obtain the optimal configuration scheme of energy storage capacity. The energy management and robust control module is used to configure the energy storage device capacity according to the optimal configuration scheme, and to build a data-driven energy management strategy based on the basic dataset, and to adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. The model update and adaptive optimization module is used to collect actual operating data, update the parameters of the new energy power generation uncertainty model, the load uncertainty model and the energy storage capacity robust optimization model, and cyclically execute the uncertainty modeling module to the energy management and robust control module to improve the robustness of energy storage capacity configuration.

[0015] The present invention discloses the following technical effects: This invention overcomes the limitations of existing methods such as differential compensation, fluctuation mitigation analysis, and traditional economic optimization, which rely solely on a single forecast curve for energy storage configuration, by explicitly introducing uncertainty models for new energy power generation and load forecasting during the energy storage capacity planning stage. This allows energy storage solutions to be optimized for a representative set of uncertainty scenarios, rather than just for "average operating conditions" or a few typical operating conditions. By analyzing the forecast error distribution based on historical operating data and forecast data in step S2, and constructing uncertainty models for new energy power generation and load, and then embedding this set of uncertainty scenarios into the robust optimization model for energy storage capacity in step S3, this invention can "see" the system performance under different deviation scenarios in advance during the capacity planning stage, thereby improving the adaptability of the planning scheme to forecast errors from the source.

[0016] This invention introduces a unified evaluation index—the weighted sum of energy storage capacity cost and operational risk—into the objective function. This integrates capacity investment with operational safety, absorption capacity, grid interaction stability, and other operational risk factors, addressing the tendency of existing methods to focus solely on economics or safety margins. Through operational risk modeling and weighting in step S3, this invention can provide a cost-risk trade-off curve under different weight settings. This allows planners to select a suitable balance between economics and reliability based on the engineering scenario, thereby enabling adjustable and interpretable energy storage capacity configuration decisions in microgrids with a high proportion of clean energy integration.

[0017] This invention employs a robust optimization method in step S4, iteratively searching for the most unfavorable scenario within the uncertainty scenario set and adjusting the energy storage capacity decision. Essentially, it achieves targeted defense against "extreme deviation scenarios," overcoming the shortcomings of traditional stochastic optimization, which relies excessively on precise probability distributions and lacks robustness to tail-end extreme cases. Unlike approaches that only test the feasibility of solutions in a limited number of typical scenarios, this invention, through "most unfavorable scenario-driven" iterative solutions, gradually incorporates scenarios with the greatest impact on system security into the optimization process. This ensures that the final optimal energy storage capacity configuration remains feasible throughout the entire uncertainty scenario set, thereby significantly reducing the risks of power imbalance, voltage frequency fluctuations, and even power outages.

[0018] This invention achieves close coupling between the planning and operation control phases through step S5, which involves configuring the energy storage device capacity according to the optimal configuration scheme and constructing a data-driven energy management strategy based on the basic dataset, adjusting the energy storage charging and discharging in real time according to the prediction deviation. This eliminates the separation between energy storage capacity planning and daily energy management. Based on the same uncertainty cognition framework and data-driven model, this invention not only provides static capacity configuration results but also transforms them into robust energy management strategies that can be executed during operation. This enables energy storage to dynamically respond to the latest prediction deviations during actual operation, further enhancing the microgrid's real-time buffering capacity against wind and solar power output fluctuations and load disturbances, and improving the overall resilience of the integrated operation of power generation, grid, load, and storage.

[0019] This invention constructs a closed-loop adaptive mechanism through step S6, which involves "collecting actual operating data—updating the uncertainty model and optimizing model parameters—cyclically executing S2 to S5." This transforms robust optimization of energy storage capacity from a "one-off planning" process into a dynamic process that continuously evolves with changes in microgrid operating characteristics, prediction accuracy, and load structure. As operating data accumulates, the uncertainty models for new energy generation and load become more closely aligned with actual operating conditions. The corresponding uncertainty scenario set and robust optimization results also better reflect the long-term operating characteristics of the microgrid, thereby gradually optimizing energy storage capacity configuration and energy management strategies. This achieves a comprehensive improvement in economy, reliability, and adaptability under conditions of high-proportion clean energy integration. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] The purpose of this invention is to provide a robust optimization method and system for energy storage capacity of clean energy microgrids that integrates predictive uncertainty, establishes a planning-operation-feedback closed-loop mechanism, and enables the energy storage capacity configuration and energy management strategy to be dynamically and adaptively updated according to the actual operating characteristics of the microgrid, thereby significantly enhancing the economy, reliability and operational resilience of the microgrid.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1As shown, this invention provides a robust optimization method for energy storage capacity in clean energy microgrids that incorporates predictive uncertainties, comprising: S1. Collect historical operation data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset; S2. Analyze the prediction error distribution based on historical operating data and forecast data, and construct a new energy power generation uncertainty model and a load uncertainty model based on the prediction error distribution, with a set of uncertainty scenarios. S3. Establish a robust optimization model for energy storage capacity that includes a set of uncertain scenarios. Incorporate the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity, and use the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. S4. The robust optimization solution method is adopted to solve the robust optimization model of energy storage capacity. The objective function is used as the solution basis. By iteratively searching for the most unfavorable scenario and adjusting the energy storage capacity decision, the robust optimization solution for prediction deviation is achieved, and the optimal configuration scheme of energy storage capacity is obtained. S5. Configure the energy storage device capacity according to the optimal configuration scheme, and build a data-driven energy management strategy based on the basic dataset. Adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. S6. Collect actual operating data, update the parameters of the new energy power generation uncertainty model, load uncertainty model and energy storage capacity robust optimization model, and repeat steps S2 to S5 to improve the robustness of energy storage capacity configuration.

[0026] Specifically, step S1 in this embodiment includes: In this embodiment, multi-source operational information of the clean energy microgrid is first collected uniformly to construct a foundational dataset for subsequent uncertainty modeling and robust optimization. The data collection uses "historical scheduling periods" as the time-resolution unit. A historical scheduling period refers to a discrete time interval divided according to fixed time intervals during the microgrid's past operation, for example, set to 15 minutes per period. Over a continuous 30-day collection period, 2880 time interval data sequences can be obtained. The clean energy microgrid consists of photovoltaic power generation units, wind power generation units, energy storage devices, load units, and tie lines connecting to the external power grid. This embodiment records the operating power of the above units within each historical scheduling period to ensure that the foundational dataset covers the operational characteristics across all time periods.

[0027] The historical operational data collected in this embodiment includes measured power of renewable energy generation, measured load power, external grid interaction power, and energy storage state of charge. Measured power of renewable energy generation refers to the actual output power (in kW) of photovoltaic and wind power generation units during each historical scheduling period; for example, the measured power of photovoltaic power can reach 220kW during midday on a sunny day. Measured load power refers to the real-time total power demand of industrial and residential loads within the microgrid; for example, it can reach 180kW during peak hours. External grid interaction power is defined as the power flow between the microgrid and the main grid (positive values ​​indicate purchased electricity, negative values ​​indicate transmitted electricity), with a typical range of -150kW to 150kW. Energy storage state of charge refers to the energy level of the energy storage device at each time period, expressed as a percentage; for example, the state of charge of the energy storage device can be as low as 25% at night. The above data are arranged according to the same historical scheduling period to form a multi-dimensional time series, constituting the historical operational data in this embodiment.

[0028] This embodiment further constructs a forecast dataset containing new energy power generation forecast data and load forecast data. The new energy power generation forecast data includes photovoltaic power forecast and wind power forecast, which are future power estimates based on external variables such as meteorological irradiance and wind speed. For example, 48 power values ​​are predicted for each 30-minute period within the next 24 hours, with a typical forecast range of 0–250 kW. The load forecast data is generated based on historical load patterns and factors such as holiday types, temperature, and humidity conditions. For example, in the forecast for the next 24 hours, the typical range of predicted load power is 90 kW–200 kW. This embodiment arranges the above forecast values ​​sequentially according to the future scheduling period dimension to form a forecast data sequence, and aligns it with the historical operating data on the time axis to construct a continuous data set covering historical and future periods, thus forming a basic dataset for subsequent uncertainty analysis and robust optimization processing.

[0029] Optionally, step S2 in this embodiment includes: After completing the construction of the basic dataset described in step S1, for each historical scheduling period, the predicted power of new energy generation is compared with the measured power of new energy generation, and the predicted power of load is compared with the measured power of load, hourly, to obtain the time series of new energy generation prediction errors and the time series of load prediction errors. Let the set of historical scheduling periods be denoted as... For any given time period The predicted power generation of new energy sources is denoted as The measured power of new energy power generation is recorded as The predicted load power is denoted as The measured power of the load is recorded as Then, the prediction error of new energy power generation is defined. and load forecasting error as follows:

[0030] in, A set of historical scheduling periods; The number of historical scheduling periods, for example, in this embodiment, is taken as... Corresponding to 30d and 15min resolution; For the first The predicted power generation of new energy sources for each historical dispatch period, in kW; For the first Measured power of new energy generation during each historical dispatch period, in kW; For the first The predicted load power for each historical dispatch period, in kW; For the first The measured load power for each historical dispatch period, in kW; For the first The prediction error of new energy power generation during a historical dispatch period; For the first Load forecasting error for each historical scheduling period.

[0031] In this embodiment, the prediction error samples from all historical scheduling periods are used to construct a time series of prediction errors for new energy power generation. With load forecast error time series The empirical distribution method was used to perform statistical analysis. By ranking the error samples, empirical quantiles at a given confidence level were defined, for example, at a confidence level of... The lower and upper quantiles of the prediction error for new energy power generation are as follows:

[0032] The lower and upper quantiles of the load forecasting error are as follows:

[0033] in, For the time series of prediction errors for new energy power generation; This is a time series of load forecasting errors; The quantile operator corresponding to the empirical distribution function of the prediction error of new energy power generation; The quantile operator corresponding to the empirical distribution function of load forecasting error; To determine the confidence level for the prediction error, this embodiment preferably selects... For example, take ; and The prediction error of new energy power generation at the confidence level is respectively Lower and upper quantiles; and The load forecasting error at the confidence level is respectively The lower and upper quantiles.

[0034] In this embodiment, based on the prediction error distribution and quantile information, a new energy power generation uncertainty model is constructed, representing the predicted power of new energy power generation in the form of interval power during future scheduling periods. Let the set of future scheduling periods be denoted as... For any future scheduling period The predicted power generation of new energy sources is denoted as Then the range of new energy power generation Represented as:

[0035] in, For the set of future scheduling periods; This represents the number of future scheduling time periods, for example, in this embodiment, it is taken as the number of time periods with a resolution of 24 hours or 30 minutes in the future. ; For the first The predicted power generation of new energy sources for each future scheduling period, in kW; To be at confidence level Next The lower bound of renewable energy power generation capacity for a future scheduling period; To be at confidence level Next The upper limit of new energy power generation capacity for a future scheduling period; and The above-defined quantiles represent the prediction errors for new energy power generation.

[0036] Similarly, this embodiment constructs a load uncertainty model for the predicted load power, representing the load power in interval form over future scheduling periods to characterize the uncertain fluctuations in load demand. For any future scheduling period... The predicted load power is denoted as Then the load power range Represented as:

[0037] in, Let be the predicted load power for the t-th future scheduling period, in kW; This represents the lower bound of the load power for the t-th future scheduling period at confidence level α. This represents the upper bound of the load power in the t-th future scheduling period at confidence level α. and The load forecasting error quantiles (i.e., empirical quantiles) defined above can, for example, fall within the error range of -40kW to 40kW, and are used to reflect the degree of deviation in load forecasting under different operating conditions.

[0038] Based on the established ranges for new energy power generation and load power, this embodiment combines the ranges of each future scheduling period into a set of uncertain scenarios to describe different possible power trajectories. For ease of description, the set of scenarios is denoted as... For any scenario With any future scheduling period Define the scenario value of new energy power generation. With load power scenario value satisfy:

[0039] Thus, the scene is obtained. Corresponding power trajectory The power trajectory set of all scenarios constitutes the uncertainty scenario set, which is used in the subsequent robust optimization model of energy storage capacity to characterize the joint fluctuation behavior of new energy power generation and load under different combinations of deviations. A collection of indexes for uncertain scenarios; The number of scenes can be selected in this embodiment based on the interval discretization precision. or Equal to finite values; For the scene Next The renewable energy power generation capacity for a future scheduling period; For the scene Next Load power for a future scheduling period; For the scene The power trajectory sequence is used to reflect the confidence level. Possible operating trajectories under the combined effects of new energy power generation and load forecasting errors.

[0040] As an example, in this embodiment, to intuitively demonstrate the structure of the renewable energy power generation range and load power range in uncertainty modeling within future scheduling periods, Table 1 presents the upper and lower bounds of power for each future scheduling period in an interval format. By combining these intervals hourly, a set of power trajectories describing different operational possibilities can be formed, constituting an uncertainty scenario set. Each row corresponds to a future scheduling period, and each column represents the renewable energy power generation range and the load power range, respectively. Combining the interval sequences of all periods generates multiple scenario trajectories for subsequent robust optimization model solving.

[0041] Table 1 Power Upper and Lower Bounds

[0042] Specifically, step S3 in this embodiment includes: This embodiment, after obtaining the set of uncertain scenarios, introduces decision variables related to energy storage capacity and operation control to establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. Let the set of future scheduling periods be denoted as... Uncertainty scenarios are set as The rated capacity of the energy storage device is denoted as The rated charge and discharge power of the energy storage device is denoted as In the scene and future scheduling periods The energy storage charging power and discharging power are denoted as follows: and Therefore, the robust optimization objective function of this embodiment... Represented as:

[0043] in, The annualized cost of investment per unit energy storage capacity is expressed in yuan / kWh. The annualized cost of investment per unit energy storage capacity is expressed in yuan / kW. This refers to the rated energy capacity of the energy storage device, expressed in kWh. The rated charge and discharge power of the energy storage device, expressed in kW; This is the trade-off coefficient between capacity investment and operational risk, and is a real number greater than 0. For example, in this embodiment, it can be taken as... or ; For operational risk indicators.

[0044] In this embodiment, operational risks The microgrid's operational performance is characterized by comprehensively considering grid interaction power fluctuations, wind and solar power curtailment, and unmet load demands within a set of uncertain scenarios. (Scenario description follows.) The probability weights are The power exchange between the power grids is The amount of wind and solar power curtailed is The load did not meet the electricity demand. Then define scenario risk. and overall operational risks as follows:

[0045] in, For the scene The probability or weight of occurrence, satisfying and For the scene Next The interaction power between the power grid and the external power grid during a future scheduling period, in kW, and its absolute value. Used to measure the fluctuation range of power exchange in the power grid; For the scene Next The amount of wind and solar power curtailed during a future scheduling period, in kWh; For the scene Next The unmet electricity demand during the future scheduling period is expressed in kWh. and These are penalty coefficients for wind and solar power curtailment and unmet load demand, respectively, used to unify different physical quantities onto the same risk measurement scale. In this embodiment, we can take... Within the range of 1 to 10 The range is 10 to 100 to reflect the higher penalty weight for system operation when the load is not met.

[0046] This embodiment introduces a microgrid power balance constraint into the robust optimization model for energy storage capacity, requiring that, under any future scheduling period and any uncertainty scenario, the sum of distributed power generation output, energy storage charging and discharging power, and external grid interaction power equals the load power. (Scenario description follows.) and future scheduling periods The output of other schedulable distributed power sources is The power generation capacity of new energy sources is The load power is Then the power balance constraint can be written as:

[0047] in, For the scene Next The renewable energy generation capacity for each future scheduling period comes from the set of uncertain scenarios constructed in step S2; For the scene Next The total output of other dispatchable distributed power sources in a future scheduling period, in kW, is typically... Within the range; These are the energy storage discharge power and charging power, respectively, in kW. Power exchanged between power grids, measured in kW; The load power is measured in kW and is also sampled from the load power range in the uncertainty scenario set described in step S2.

[0048] This embodiment further sets energy storage charge / discharge constraints and state of charge constraints in the robust optimization model to ensure that the energy storage device operates within physically permissible limits. (Scenario description follows) and future scheduling periods The energy state of the lower energy storage is The minimum energy storage is Charging efficiency is The discharge efficiency is The scheduling period length is (Unit: h), then the energy storage state update and boundary constraints are:

[0049] And mutual exclusion constraints are used to ensure that energy storage does not charge and discharge simultaneously at any given time:

[0050] in, For the scene Next The energy storage status for a future scheduling period, in kWh; The lower limit of energy storage capacity can be determined by multiplying the rated capacity by the minimum state-of-charge factor, for example, taking... The rated energy capacity of the energy storage is related to the objective function. same; and These represent the energy storage charging efficiency and the discharging efficiency, respectively. In this embodiment, we can take... , For example, take the length of the scheduling period. h corresponds to 30 minutes; For 0-1 type mutually exclusive variables, when Allowed energy storage during time period When charging and discharging power is 0, Discharge is allowed while charging power is 0.

[0051] To ensure the safe operation of equipment within the microgrid within its rated capacity, this embodiment incorporates operational constraints on distributed generation output and tie-line power into the robust optimization model for energy storage capacity. For the scenario... and future scheduling periods The power output of other dispatchable distributed power sources interacting with the grid satisfies the following:

[0052] in, This represents the upper limit of the rated capacity of other schedulable distributed power sources, in kW. For example, in this embodiment, it can be taken as... kW; 'max' represents the maximum allowable grid interconnection power amplitude of the tie line, in kW. For example, it can be taken as... 150kW, used to limit the power exchange between clean energy microgrids and the upper-level grid to no more than the line transmission capacity; and Each symbol is consistent with the one in the power balance constraint mentioned above, thereby ensuring that the power flow of the entire robust optimization model is coordinated in each scenario and time period and meets the operating limits of the equipment.

[0053] Optionally, step S4 in this embodiment includes: In this embodiment, after obtaining the robust optimization model for energy storage capacity described in step S3, the model is solved using a column and constraint generation algorithm. The set of uncertain scenarios is divided into progressively expanding finite scenario subsets, and the robust solution for the entire scenario is approximated iteratively. Let the complete set of uncertain scenarios be denoted as... The iteration number index is , No. The subset of scenarios that have been incorporated into the main problem in the next iteration is: , No. The decision variable vector corresponding to the next iteration is denoted as... It includes the rated capacity of the energy storage device. Rated power of energy storage device and in scene subsets The operating variables, such as the energy storage charging and discharging power, external grid interaction power, and output of other dispatchable distributed power sources in each future scheduling period, constitute a solution framework that iteratively combines the main problem and sub-problems. In this embodiment, in the first... In the next iteration, the main problem is constructed, considering the current finite subset of the scenario. The energy storage capacity decision is optimized, and the objective function is the weighted sum of capacity investment cost and operational risk for this subset:

[0054] in, For the first The objective function value of the main problem in the next iteration; The decision variable vector for the main problem includes components such as the rated capacity of energy storage. Rated power of energy storage and in scene subset and future scheduling time set The operating power variable defined above; The annualized cost of investment per unit of energy storage capacity; The annualized cost of investment per unit of energy storage capacity; This is a trade-off coefficient between capacity investment and operational risk; For the scene The weights or probabilities of occurrence satisfy For the scene The operational risk indicators are calculated within the scenario subset according to the definition of operational risk in step S3. The main problem simultaneously satisfies the power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints given in step S3 in all scenarios. and established on.

[0055] In this embodiment, after solving for the first... Candidate solutions to the main problem in the next iteration Then, construct sub-problems within the complete scenario set. The worst-case scenario corresponding to the candidate solution is searched to measure the robustness boundary of the current solution. For this purpose, a definition is defined on the candidate solution... Risk function for the next scenario And determine the worst-case scenario index by maximizing the following problem. :

[0056] in, For the first In the next iteration, in the complete set of uncertainty scenarios Inner relative to candidate solutions Index of scenarios with the highest operational risk; In candidate solutions The following scenario is calculated according to the method defined in step S3. The operational risk value, through all of By making comparisons, the scenario with the most stringent constraints and the most unfavorable impact on current energy storage capacity decisions can be identified.

[0057] In this embodiment, the most unfavorable scenario index is obtained. Next, the scene subset is updated if the scene is not yet included in the current scene subset. If it is in the middle, then add it to form a new subset. And based on this, proceed to the next iteration to solve the main problem; if the scenario is already included If the improvement in the operational risk of the selected scenario or its corresponding scenario is less than a preset convergence threshold, then the decision column and constraint generation process converges. The above scenario subset update process can be represented as:

[0058] in, For the first A subset of the scenes in the next iteration; For the first A subset of the scenes in the next iteration; The worst-case scenario index is defined in the previous formula, and it is used in a series of iterations. and When no further changes occur, it is considered that the robust optimization solution method has obtained the optimal configuration scheme of energy storage capacity that satisfies robustness across all scenarios.

[0059] Further, step S5 of this embodiment includes: After obtaining the optimal energy storage capacity configuration in step S4, a data-driven energy management strategy is constructed using a rolling time-domain optimization approach. This strategy is used to generate energy storage charging and discharging power commands and external grid interaction power reference values ​​for each scheduling period. Let the set of scheduling periods within the rolling optimization period be denoted as […]. The decision variables during the current rolling optimization period include energy storage charging power. Discharge power External power grid interaction power and other dispatchable distributed power output Therefore, the objective function of the short-term optimization model in this embodiment can be expressed as:

[0060] in, The objective function for operating costs within a rolling optimization period; This refers to the set of scheduling periods included in the current rolling optimization period. To optimize the number of scheduling periods within a rolling time period, for example, in this embodiment, we can take... Corresponding to 24 one-hour time periods; The price of electricity purchased from the external power grid is expressed in yuan / kWh. The settlement price for selling electricity to the external power grid, expressed in yuan / kWh, typically meets the following requirements. This is a penalty coefficient for power fluctuations in the power grid, used to suppress large power fluctuations. In this embodiment, it can be taken as... Within the range of 0.1 to 1; For the first The power exchanged between the dispatch period and the external power grid, in kW. A positive value indicates that the power is purchased from the power grid, and a negative value indicates that the power is sent to the power grid. , , The physical meaning of the variable with the same name in step S3 above is consistent, but the scene index is no longer distinguished. Instead, the optimization is performed on the deterministic prediction data within the rolling optimization period.

[0061] In this embodiment, the power balance and energy storage constraints in the rolling time-domain optimization model are consistent with the constraints in step S3, but a single-scenario predicted power trajectory is used in the rolling optimization scenario. Let the first... The predicted power generation of new energy sources for each scheduling period is The predicted load power is The energy storage state is Then the power balance constraint and the energy storage state update constraint are respectively:

[0062]

[0063] in, For the first The predicted power generation of new energy sources for each scheduling period, in kW; For the first The predicted load power for each scheduling period, in kW; For the first Energy storage status during each scheduling period, in kWh; For example, take the length of the scheduling period. 1 hour; and These refer to energy storage charging efficiency and discharging efficiency, respectively. For example, in this embodiment, we can take... , and These are the lower and upper limits of energy storage state of being, respectively. The rated energy storage capacity obtained in step S3, Can be taken as The rated charge and discharge power of the energy storage obtained in step S3; , , , During each scheduling period, the equipment operation constraints given in step S3 (including upper and lower bound constraints on other schedulable power sources and tie line power) must also be met to ensure consistency between short-term operation optimization results and robust planning capacity.

[0064] In this embodiment, to achieve real-time response to prediction deviations, a power correction mechanism based on prediction deviations is further introduced after obtaining the energy storage charging and discharging power command through rolling optimization. Let the... The measured power of new energy generation during each scheduling period is The measured power of the load is Then the net power prediction deviation is defined. for:

[0065] Based on this, the charging and discharging power command is corrected according to the direction of deviation to obtain the real-time charging and discharging power. and :

[0066] in, For the first The net power prediction deviation for each scheduling period is as follows: a positive value indicates that the actual net load is higher than the prediction and that it is necessary to increase discharge or purchase electricity; a negative value indicates that the actual net load is lower than the prediction and that it is possible to increase charging or reduce electricity purchase. For the first Measured power generation of new energy sources during each scheduling period; For the first Measured load power during each scheduling period; and These are the real-time command values ​​of energy storage discharge power and charging power, respectively, after correction based on prediction deviations. This is the deviation compensation coefficient, used to allocate the net power deviation portion to energy storage for adjustment. In this embodiment, it can be taken as... The value should be kept within the range of 0.1 to 0.5 to avoid over-adjustment. The calculation ensures that the corrected charging and discharging power does not exceed the rated power of the energy storage. This enables robust scheduling of energy storage based on prediction deviations during actual operation, thereby improving the microgrid's real-time buffering capacity against power deviations and fluctuations.

[0067] Furthermore, step S6 of this embodiment includes: In this embodiment, to ensure that the uncertainty model for new energy power generation, the uncertainty model for load, and the robust optimization model for energy storage capacity can be continuously and adaptively updated according to the operating characteristics of the microgrid, actual operating data is continuously collected within a preset update cycle and merged with the existing basic dataset. (Note: The last part is a repetition of the first part and can be omitted.) The set of newly added actual running data corresponding to each update cycle is: The previous round's basic dataset was In this embodiment, the updated base dataset is performed after the data merging is completed. Represented as:

[0068] in, For the first The actual operation data set collected within each update cycle includes data such as the measured power of new energy power generation, the measured power of load, the power of external power grid interaction, and the status of energy storage during each scheduling period within that cycle; For the first The base dataset before the update cycle corresponds to the historical running data and prediction data accumulated over several update cycles; The updated base dataset will serve as new data input in the next round of prediction error analysis and robust optimization modeling; symbol " "" indicates a set union operation, used to represent the expansion of data samples.

[0069] In this embodiment, to avoid frequent updates to the robust optimization model when there are insufficient new data samples or the statistical characteristics do not change significantly, an update condition judgment mechanism based on data scale and statistical characteristics is introduced. Let the... The number of new samples added within each update cycle is The preset minimum sample size threshold is Meanwhile, the mean and variance of the prediction errors for new energy power generation in the previous and current base datasets are recorded as follows: , and , Therefore, the update condition criterion is defined as follows in this embodiment:

[0070] in, For the first The number of new actual running data samples added within an update cycle, for example, in this embodiment, a sufficient sample size can be considered when no less than 672 15-minute scheduling periods of data are collected within 7 consecutive days. A minimum sample size threshold is preset to ensure the stability of statistical analysis results; , The first The sample mean and sample standard deviation of the prediction error of new energy power generation on the basic dataset; , These are the sample mean and sample standard deviation obtained after incorporating the newly added data; and The threshold for determining changes in the statistical characteristics of prediction errors, for example, can be set to [value missing] in this embodiment. For 1kW, The value is 2kW, which is used to determine whether the prediction error distribution has drifted significantly, thereby triggering updates to the uncertainty model and the robust optimization model.

[0071] In this embodiment, when the above update conditions are met, the executing entity re-based on the updated base dataset. Steps S2 to S4 are executed sequentially to update the uncertainty model of new energy power generation, the uncertainty model of load, and the robust optimization model of energy storage capacity, thereby obtaining a new optimal configuration scheme for energy storage capacity. This update scheme is then adopted as the capacity parameter in subsequent rolling time-domain optimizations. The change in the capacity parameter before and after the update can be expressed by the following formula:

[0072] in, and For the first The rated capacity and rated charge / discharge power of the energy storage before the upgrade; and For the first Capacity and rated charge / discharge power of energy storage; and These represent the changes in capacity and power parameters caused by this round of model updates. Their magnitude reflects the degree of adaptive adjustment of the energy storage planning scheme to predictive uncertainties and operational characteristics as actual operating data accumulates. In this embodiment, the clean energy microgrid includes at least one new energy power generation unit, at least one set of energy storage devices, and at least one type of dispatchable load. The above-mentioned update mechanism supports the formation of a closed-loop robust optimization process of "operation-update-replanning".

[0073] In this exemplary embodiment, the clean energy microgrid comprises at least one new energy power generation unit, at least one set of energy storage devices, and at least one type of dispatchable load. These three components form a closed-loop structure of energy and power flow through the microgrid's internal power distribution lines, power electronic interfaces, and energy management control relationships. The new energy power generation unit includes wind power, photovoltaic power, or other intermittent renewable energy sources. The power generated by these units during each dispatch period is first connected to the microgrid's AC bus. The energy storage device is connected to the AC bus via a bidirectional conversion interface. When the new energy power generation exceeds the load power, it absorbs excess energy to form charging power; when the new energy power generation is insufficient, it releases energy to the microgrid to form discharging power. The dispatchable load is connected to the AC bus via a load control interface. Its power demand is responded to in an adjustable manner through energy management strategies, thereby undertaking some regulation functions during microgrid operation. The three types of units mentioned above achieve real-time coordination of power flow within the microgrid by jointly satisfying power balance constraints. Among them, the new energy generation unit provides renewable energy input, the energy storage device provides power regulation and energy buffering, and the dispatchable load provides demand-side controllability. This enables the prediction uncertainty model, energy storage capacity robust optimization model, and rolling time-domain energy management strategy constructed in this embodiment to play a role in the "source-load-storage" coupled structure of the microgrid, and achieve comprehensive robust control over power fluctuations, prediction deviations, and extreme scenarios.

[0074] Furthermore, to facilitate the demonstration of the adaptive adjustment effect of the robust optimization method for energy storage capacity after multiple rounds of operational data accumulation, this embodiment lists the rated energy storage capacity, rated energy storage power, and their changes in each update stage in tabular form. Table 2 reflects the actual engineering process of a typical park-level microgrid gradually adjusting its energy storage configuration based on fluctuations in renewable energy output, load change trends, and prediction error drift during long-term operation. In the initial stage, approximately two megawatt-hours of energy storage capacity and approximately eight hundred kilowatts of charging and discharging power were configured to meet basic peak shaving and valley filling and some reserve needs. As operational data accumulated, the first round of updates mainly increased the energy storage capacity to enhance the energy buffering capacity under high renewable energy penetration. In the second round of updates, the charging and discharging power was increased while maintaining the capacity, enabling the energy storage to more effectively cope with rapid fluctuations in net load. The third round of updates further increased the energy storage capacity to adapt to a higher proportion of renewable energy and longer-term energy regulation needs. This table visually demonstrates the closed-loop characteristics of the method of this invention in long-term operation, namely "detecting deviations - updating the model - adjusting capacity," enabling energy storage planning to continuously match the dynamic operating conditions of the park's microgrid and improve the overall stability and robustness of operation.

[0075] Table 2 Iterative update record of rated capacity and rated power of energy storage

[0076] Corresponding to the above methods, such as Figure 2 As shown, this embodiment also provides a robust optimization system for clean energy microgrid energy storage capacity that integrates predictive uncertainty, including: The data acquisition and construction module is used to collect historical operating data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset. The uncertainty modeling module is used to analyze the prediction error distribution based on the historical operating data and the prediction data, and to construct a new energy power generation uncertainty model and a load uncertainty model based on the prediction error distribution, with a set of uncertainty scenarios. The robust optimization modeling module is used to establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. It incorporates the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity, and takes the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. The robust optimization solution module is used to solve the robust optimization model of the energy storage capacity using a robust optimization solution method. Based on the objective function, it iteratively searches for the most unfavorable scenario and adjusts the energy storage capacity decision to achieve robust optimization solution for prediction deviation and obtain the optimal configuration scheme of energy storage capacity. The energy management and robust control module is used to configure the energy storage device capacity according to the optimal configuration scheme, and to build a data-driven energy management strategy based on the basic dataset, and to adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. The model update and adaptive optimization module is used to collect actual operating data, update the parameters of the new energy power generation uncertainty model, the load uncertainty model and the energy storage capacity robust optimization model, and cyclically execute the uncertainty modeling module to the energy management and robust control module to improve the robustness of energy storage capacity configuration.

[0077] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0078] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A robust optimization method for energy storage capacity in clean energy microgrids that incorporates predictive uncertainty, characterized in that, include: S1. Collect historical operation data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset; S2. Analyze the prediction error distribution based on the historical operating data and the prediction data, and construct a new energy power generation uncertainty model and a load uncertainty model according to the prediction error distribution to form a set of uncertainty scenarios; S3. Establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. Incorporate the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity. Use the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. S4. The robust optimization solution method is used to solve the robust optimization model of the energy storage capacity. The objective function is used as the solution basis. By iteratively searching for the most unfavorable scenario and adjusting the energy storage capacity decision, the robust optimization solution for the prediction deviation is achieved, and the optimal configuration scheme of the energy storage capacity is obtained. S5. Configure the energy storage device capacity according to the optimal configuration scheme, and build a data-driven energy management strategy based on the basic dataset. Adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. S6. Collect actual operating data, update the parameters of the new energy power generation uncertainty model, the load uncertainty model and the energy storage capacity robust optimization model, and repeat steps S2 to S5 to improve the robustness of energy storage capacity configuration.

2. The robust optimization method for clean energy microgrid energy storage capacity based on fusion prediction uncertainty as described in claim 1, characterized in that, The historical operating data includes the measured power of renewable energy generation, measured load power, external grid interaction power, and energy storage status of charge of the clean energy microgrid during multiple historical scheduling periods; the forecast data includes the forecasted power of renewable energy generation and the forecasted load power during each scheduling period.

3. The robust optimization method for clean energy microgrid energy storage capacity based on fused predictive uncertainty as described in claim 2, characterized in that, In step S2, the analysis of the prediction error distribution includes: The difference between the predicted power of new energy generation and the corresponding measured power of new energy generation, as well as the difference between the predicted power of load and the corresponding measured power of load, are calculated to obtain the time series of error in new energy generation prediction and the time series of error in load prediction. Statistical analysis is performed on the time series of the load forecasting error to obtain the empirical probability distribution of the forecasting error in each scheduling period, and the empirical probability distribution is used as the forecasting error distribution.

4. The robust optimization method for clean energy microgrid energy storage capacity based on fusion prediction uncertainty as described in claim 3, characterized in that, In step S2, the uncertainty model for new energy power generation and the uncertainty model for load are constructed based on the prediction error distribution. Specifically, the upper and lower bounds of the prediction error for new energy power generation and the prediction error for load are determined according to a given confidence level in each scheduling period. The upper and lower bounds are superimposed with the predicted power of new energy power generation and the predicted power of load to obtain the power range of new energy power generation and the power range of load. The uncertainty scenario set is composed of the power trajectory set formed by the combination of the power range of new energy power generation and the power range of load in each scheduling period.

5. The robust optimization method for clean energy microgrid energy storage capacity based on fused predictive uncertainty as described in claim 1, characterized in that, In step S3, the operational risk is obtained by calculating the power fluctuation amplitude of grid interaction, the amount of wind and solar power curtailment, and the amount of unmet load demand under each scenario within the uncertainty scenario set, and then performing a weighted summation. The objective function is the linear weighted sum of the annualized cost of capacity investment corresponding to the rated capacity and rated power of the energy storage device and the operational risk.

6. The robust optimization method for clean energy microgrid energy storage capacity based on fused predictive uncertainty as described in claim 1, characterized in that, In step S3, the microgrid power balance constraint is that, under any scheduling period and any uncertainty scenario, the sum of the output of distributed power sources, the charging and discharging power of energy storage, and the interaction power of the external power grid equals the load power; the energy storage charging and discharging constraint is that the charging power and discharging power of energy storage are mutually exclusive in any scheduling period and the energy storage state of charge changes within a preset upper and lower limit; the equipment operation constraint is that the output of distributed power sources and the tie-line power operate within the rated capacity range.

7. The robust optimization method for clean energy microgrid energy storage capacity based on fused predictive uncertainty as described in claim 1, characterized in that, In step S4, the robust optimization solution method adopts a column and constraint generation algorithm that alternates between the main problem and sub-problems; The main problem optimizes energy storage capacity decision-making and outputs candidate energy storage capacity schemes under the current finite uncertainty scenario; the sub-problem searches for the most unfavorable scenario that imposes the most stringent constraints on the candidate energy storage capacity schemes within the value range corresponding to the uncertainty scenario set, and adds the most unfavorable scenario to the main problem until no new most unfavorable scenario exists.

8. The robust optimization method for clean energy microgrid energy storage capacity based on fused predictive uncertainty as described in claim 1, characterized in that, In step S5, the data-driven energy management strategy adopts a rolling time-domain optimization method. At the beginning of each rolling optimization period, the optimal configuration scheme of the basic dataset and the energy storage capacity is used to construct a short-term operation optimization model, solve for the energy storage charging and discharging power command and external grid interaction power reference value for each scheduling period in the current rolling optimization period, and correct the energy storage charging and discharging power command according to the real-time prediction deviation during actual operation.

9. The robust optimization method for clean energy microgrid energy storage capacity based on fusion prediction uncertainty as described in claim 1, characterized in that, In step S6, the actual operating data collected within the preset update cycle is incorporated into the basic dataset. When the scale or statistical characteristics of the actual operating data meet the preset update conditions, steps S2 to S4 are re-executed to obtain the updated optimal configuration scheme of energy storage capacity, and the updated optimal configuration scheme of energy storage capacity is adopted in the subsequent rolling time-domain optimization. The clean energy microgrid includes at least one new energy power generation unit, at least one set of energy storage devices, and at least one type of dispatchable load.

10. A robust optimization system for energy storage capacity in clean energy microgrids that integrates predictive uncertainty, characterized in that, include: The data acquisition and construction module is used to collect historical operating data of clean energy microgrids, forecast data of new energy power generation, and forecast data of load to build a basic dataset. The uncertainty modeling module is used to analyze the prediction error distribution based on the historical operating data and the prediction data, construct a new energy power generation uncertainty model and a load uncertainty model based on the prediction error distribution, and generate a set of uncertainty scenarios. The robust optimization modeling module is used to establish a robust optimization model for energy storage capacity that includes the set of uncertain scenarios. It incorporates the preset microgrid power balance constraints, energy storage charging and discharging constraints, and equipment operation constraints into the robust optimization model for energy storage capacity, and takes the minimum weighted sum of energy storage capacity cost and operation risk as the objective function. The robust optimization solution module is used to solve the robust optimization model of the energy storage capacity using a robust optimization solution method. Based on the objective function, it iteratively searches for the most unfavorable scenario and adjusts the energy storage capacity decision to achieve robust optimization solution for prediction deviation and obtain the optimal configuration scheme of energy storage capacity. The energy management and robust control module is used to configure the energy storage device capacity according to the optimal configuration scheme, and to build a data-driven energy management strategy based on the basic dataset, and to adjust the energy storage charging and discharging in real time according to the prediction deviation to achieve robust operation control. The model update and adaptive optimization module is used to collect actual operating data, update the parameters of the new energy power generation uncertainty model, the load uncertainty model and the energy storage capacity robust optimization model, and cyclically execute the uncertainty modeling module to the energy management and robust control module to improve the robustness of energy storage capacity configuration.

Images