Energy storage system dynamic capacity configuration method suitable for distributed new energy

Abstract

The invention discloses an energy storage system dynamic capacity configuration method suitable for distributed new energy, relates to the technical field of energy storage configuration, and solves the technical problems of poor scene adaptability and lack of accurate classification and targeted strategies. The power supply gap is subdivided into a new energy output insufficient type and a power grid peak load type, a differentiation strategy is designed for each type of scene, one type of scene and one strategy are achieved, the new energy consumption rate is improved, the energy storage vacancy rate is reduced, energy storage safety constraint, power grid stability constraint and equipment operation constraint are brought into a unified optimization framework, overcharge and overdischarge are avoided, and the energy storage efficiency is improved. The voltage frequency fluctuation amplitude of the power grid is reduced, the isolated grid operation stability of the micro-grid is improved, the load is divided into a first level, a second level and a third level, different load level strategies are formulated, and the problem of disordered power limitation is solved.

Description

Technical Field

[0001] This invention relates to the field of energy storage configuration technology, specifically to a dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources. Background Technology

[0002] As the global energy structure shifts towards cleaner and distributed energy, the installed capacity of new energy sources such as wind and solar power continues to grow. As a core piece of equipment for mitigating the volatility of new energy sources and solving the spatial and temporal mismatch between power generation and load, the rationality of the configuration of distributed new energy storage systems directly affects the new energy absorption rate, grid stability, and operational economy.

[0003] According to patent application number CN202411901190.4, an adjustable resource allocation and energy storage deployment optimization method considering the carrying capacity of new energy sources is disclosed. The method includes: establishing models for wind power generation, photovoltaic power generation, and battery storage; calculating the optimal energy storage system capacity, i.e., battery capacity; using the distribution ratio of wind power and photovoltaic power under the optimal energy storage system capacity as a constraint, and taking the maximum annual total power generation of wind power and photovoltaic energy in each region of the power grid as the objective function, establishing an optimal wind-solar-storage capacity allocation model, and solving for the optimal allocation ratio of each distributed resource capacity; determining the optimal layout location of the energy storage system in the power grid, and taking maximizing energy storage utilization and minimizing energy storage construction and maintenance costs as objective functions, establishing an energy storage deployment optimization model, and solving for the optimal energy storage location.

[0004] Traditional energy storage systems mostly adopt static capacity configuration schemes, which have the following limitations: First, they cannot adapt to different types of surplus and shortage scenarios; second, the prediction methods are singular, relying on a single algorithm or historical data without combining real-time operating conditions for dynamic correction, resulting in a lag in the judgment of surplus and shortage; third, the constraints are considered in a one-sided manner, which can easily cause grid fluctuations or damage to energy storage equipment; fourth, the load guarantee lacks priority classification, and it is impossible to accurately guarantee the power supply of critical loads such as residential life, hospitals / data centers during peak shortages. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources, solving the problems of poor scenario adaptability and lack of accurate classification and targeted strategies.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources, which specifically includes the following steps: By combining AI algorithms to make ultra-short-term predictions for distributed new energy storage systems, the predicted output of new energy is obtained, and the benchmark value of total electricity consumption in the same period is determined. The two are compared. If the former is larger, a signal is generated that the power generation meets the analysis; otherwise, a signal is generated that the power generation does not meet the analysis. For power generation to meet the analysis signal, the difference between the predicted output of new energy and the benchmark value of total electricity consumption is recorded as the surplus electricity. The surplus electricity is classified into new energy consumption type, grid off-peak type, and microgrid type surplus, and dynamic configuration is performed for each of the three. For renewable energy consumption type, calculate the real-time excess output and the predicted excess output and compare their magnitudes. Generate configuration information based on matching excess and satisfying constraints. For grid off-peak type, determine the off-peak period and adjust the charging power according to grid dispatch instructions to generate configuration information. For microgrid type excess, determine the total load of the microgrid and adjust the power according to the relationship between the predicted renewable energy output and the total load to generate configuration information. If the power generation does not meet the analysis signal, the difference between the total power consumption benchmark value and the predicted output of new energy is recorded as the power gap, and the power gap is classified into two types: insufficient output of new energy and peak load of the power grid. Dynamic configuration is carried out for each type. For the type of insufficient output of new energy, the load is acquired and the load level is determined, the predicted shortfall power is calculated, and the power is adjusted according to the standard to generate configuration information. For the type of shortfall due to peak load of the power grid, the duration of the shortfall, the maximum release capacity of energy storage and the allowable capacity are determined, and configuration is carried out according to different load levels to generate configuration information.

[0007] As a further aspect of the present invention, the AI ​​algorithm for ultra-short-term prediction includes: Acquire real-time data, including energy storage parameters, energy and load parameters, and grid parameters, among which energy storage parameters include SOC, SOH, and temperature; Energy and load parameters include real-time output of photovoltaic / wind power and load in the grid connection area; grid parameters include grid connection point voltage and frequency. The LSTM+random forest algorithm is used to predict the power output of new energy sources in the next 10 minutes to 2 hours, and the power output of new energy sources is summarized to obtain the predicted power output of new energy sources.

[0008] As a further aspect of the present invention, the method for dynamic capacity configuration in the renewable energy consumption type is as follows: Real-time excess output = current output of new energy - (local load + available capacity of external transmission channel), predicted excess output = predicted output of new energy - (predicted value of local load + upper limit of external transmission channel). If either output is >0, energy storage charging is triggered; if both are ≤0, standby mode is activated. Based on the principle of matching excess and satisfying constraints, time t is divided into n hourly periods, with the values ​​of t and n set by the operator. A stepped power curve corresponding to the small time period is generated, which is then converted into digital instructions to generate configuration information.

[0009] As a further aspect of the present invention, the grid off-peak surplus configuration includes: The power grid SCADA system collects load data in seconds and uses the ARIMA algorithm to predict the 24-hour load curve. If the load is lower than 40% of the total installed capacity or 60% of the daily average load for 2 consecutive hours, it is determined to be a trough and the fluctuation range is recorded. The system statistically analyzes the basic output range of wind power / solar power during off-peak hours, uses LSTM to predict output fluctuations, collects grid connection frequency and external transmission channel reserve capacity, determines off-peak boundaries, excess power ranges, and grid-allowed charging upper limits. If there are grid dispatch instructions, the instructions are used as the basis; otherwise, the basic charging demand is calculated as excess power × 0.8. Combining energy storage and grid security constraints, the system outputs time-period target charging power curves. Based on the PCS rated power and the battery's maximum charging rate, time t1 is divided into two small cycles with a fixed power. The value of time t1 is set by the operator, with a 5% adjustment margin. When the SOC reaches 78%, the charging power is reduced to the minimum, and configuration information is generated.

[0010] As a further aspect of the present invention, the constraints include energy storage state constraints and grid security constraints; Energy storage state constraints include: remaining rechargeable capacity = total energy storage capacity × (80% of SOC safety limit - current SOC) × charging efficiency, and the corrected power ≤ remaining rechargeable capacity ÷ remaining off-peak duration. Grid safety constraints include: if the grid connection frequency is >50.2Hz, the charging power is increased to 1.2 times the base value and does not exceed the grid upper limit; if the frequency is <49.9Hz, it is reduced to 0.7 times the base value.

[0011] As a further aspect of the present invention, the configuration method for microgrid-type surplus scenarios is as follows: The dynamic output range is obtained by assessing the fluctuation of renewable energy output. The load is classified into three levels: Level 1, Level 2, and Level 3. The load demand range is fitted by a normal distribution and summed to obtain the total load of the microgrid. If the predicted renewable energy output minus the total load is greater than 5% of the microgrid's rated load, or if the frequency is greater than 50.2Hz, charging is initiated. If the voltage is greater than 1.05 times the rated value, the power is increased by 10%-20%, and if it is less than 1.02 times the rated value, the base power is maintained. If the frequency is greater than 50.3Hz, the power is fully utilized, and if it drops to 50.1Hz, the base power is reduced. If only the output is greater than the load and the difference is less than or equal to 5% of the rated load, the system is put into standby mode, and configuration information is generated.

[0012] As a further aspect of the present invention, the dynamic output fluctuation range is obtained as follows: For wind power output, the Weibull distribution is used to fit the wind speed-output relationship; for photovoltaic output, the Beta distribution is used to fit the irradiance-output curve. The distribution parameters are determined by the historical output data of one year, and the output fluctuation range at a 95% confidence level is output. Combined with real-time data, a lightweight LSTM model is used to make ultra-short-term output predictions at the 10-minute level, with the prediction error controlled within 8%. The predicted values ​​are used to correct the historical statistical distribution.

[0013] As a further aspect of the present invention, the configuration method for the type of insufficient new energy output is as follows: Model predictive control is adopted, and the optimal discharge power curve is calculated on a rolling basis with time t3 as the standard. Time t3 is divided into several hour segments, and the value of t3 is set by the operator, and a fixed power is output. The load level is obtained, and the power is adjusted according to the predicted shortfall power = (first-level + second-level load prediction value) - (new energy predicted output + conventional power output) to generate configuration information.

[0014] As a further aspect of the present invention, the configuration method for the peak load type gap in the power grid is as follows: Discharge capacity is allocated according to load level. The LSTM outputs the new energy output and load curves for the next 1-2 hours. The intersection of the two is calculated to obtain the gap duration. The maximum energy storage capacity = total energy storage capacity × (current capacity - capacity lower limit) × discharge efficiency. The allowable capacity based on duration = min(maximum energy storage capacity, capacity lower limit) × gap duration. Then, the required discharge amount is calculated according to load level.

[0015] As a further aspect of the present invention, the method for calculating the required discharge volume according to different load levels is as follows: Primary load gap = Primary load demand - (New energy + conventional power output), Secondary load gap = Secondary load demand - Remaining power output, Demand capacity = (Primary load gap + a × Secondary load gap) × Gap duration, where a is the guarantee coefficient, Discharge capacity = min (Maximum discharge capacity, Allowable capacity, Demand capacity), Generate configuration information.

[0016] This invention provides a dynamic capacity configuration method for energy storage systems applicable to distributed renewable energy sources. Compared with existing technologies, it has the following advantages: This invention categorizes power generation surplus into three types: renewable energy absorption type, grid off-peak type, and microgrid type. It also categorizes power supply shortage into two types: renewable energy insufficient output type and grid peak load type. Differentiated strategies are designed for each type of scenario, achieving one strategy for each scenario. This improves the renewable energy absorption rate, reduces the energy storage idle rate, and incorporates energy storage safety constraints, grid stability constraints, and equipment operation constraints into a unified optimization framework. This avoids overcharging and over-discharging, reduces the amplitude of grid voltage and frequency fluctuations, and improves the stability of microgrid isolated operation.

[0017] This invention divides the load into three levels: Level 1, Level 2, and Level 3, and formulates different load level strategies to solve the problem of disordered power rationing. It adopts a 15-minute rolling optimization mechanism. After each control cycle is completed, the next round of strategy is immediately modified based on real-time output, load, and grid status to ensure that the strategy is always adapted to dynamic operating conditions. Compared with the existing one-time command configuration, it improves the economic efficiency of energy storage operation. Attached Figure Description

[0018] Figure 1 This is a diagram illustrating the steps and methods of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0019] 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.

[0020] Please see Figure 1 and Figure 2 This application provides a dynamic capacity configuration method for energy storage systems applicable to distributed renewable energy sources, which specifically includes the following steps: First Embodiment

[0021] By combining AI algorithms to make ultra-short-term predictions for distributed new energy storage systems, the predicted output of new energy is obtained, and the benchmark value of total electricity consumption in the same period is determined. The two are compared. If the former is larger, a signal is generated that the power generation meets the analysis; otherwise, a signal is generated that the power generation does not meet the analysis. For power generation to meet the analysis signal, the difference between the predicted output of new energy and the benchmark value of total electricity consumption is recorded as the surplus electricity. The surplus electricity is classified into new energy consumption type, grid off-peak type, and microgrid type surplus, and dynamic configuration is performed for each of the three. For renewable energy consumption type, calculate the real-time excess output and the predicted excess output and compare their magnitudes. Generate configuration information based on matching excess and satisfying constraints. For grid off-peak type, determine the off-peak period and adjust the charging power according to grid dispatch instructions to generate configuration information. For microgrid type excess, determine the total load of the microgrid and adjust the power according to the relationship between the predicted renewable energy output and the total load to generate configuration information. If the power generation does not meet the analysis signal, the difference between the total power consumption benchmark value and the predicted output of new energy is recorded as the power gap, and the power gap is classified into two types: insufficient output of new energy and peak load of the power grid. Dynamic configuration is carried out for each type. For the type of insufficient output of new energy, the load is acquired and the load level is determined, the predicted shortfall power is calculated, and the power is adjusted according to the standard to generate configuration information. For the type of shortfall due to peak load of the power grid, the duration of the shortfall, the maximum release capacity of energy storage and the allowable capacity are determined, and configuration is carried out according to different load levels to generate configuration information.

[0022] Second Embodiment

[0023] Step 1: Acquire all distributed new energy storage systems and simultaneously obtain their corresponding real-time data. The real-time data includes the energy storage system's status parameters, energy and load parameters, and grid operation parameters. The energy storage system's status parameters include the battery's state of charge (SOC), state of health (SOH), and operating temperature. The energy and load parameters include the real-time output data of the supporting photovoltaic or wind power and the actual load data of the connected area. The grid operation parameters include the real-time voltage and frequency of the grid connection point. At the same time, use AI algorithms to complete ultra-short-term predictions. The AI ​​algorithms here include LSTM and random forest algorithms. The specific prediction process is existing technology and will not be elaborated on here. Obtain the corresponding ultra-short-term power generation and calculate the predicted output of new energy. Then, simultaneously retrieve the historical electricity consumption data for this period. After outlier removal and trend correction, determine the benchmark value of the total electricity consumption for the same period and compare the two. If the predicted output of new energy is greater than the total electricity consumption, it means that the power generation can meet the corresponding electricity demand, and a power generation satisfaction analysis signal is generated. Otherwise, a power generation dissatisfaction analysis signal is generated. Step 2: Analyze the power generation satisfaction signal obtained from the analysis, calculate the difference between the total overdue power generation and the total power consumption, and record it as the surplus power. At the same time, obtain the scenarios corresponding to the surplus power and classify them into new energy consumption type surplus, grid off-peak type surplus and microgrid type surplus. Then, perform dynamic capacity configuration for the surplus power of different scenario types. Step 3: For scenarios with excess capacity due to renewable energy absorption, calculate real-time excess output and predict excess output. Real-time excess output = current renewable energy output - (local load + available capacity of transmission channels). Available capacity of transmission channels represents the remaining usable transmission capacity at a given moment, after deducting the currently transmitted power from the transmission channel's upper limit. Available capacity of transmission channels = upper limit of transmission channels - current actual transmission power of transmission channels. For example, if the upper limit of the transmission channel is 50MW, and at a certain moment the channel has transmitted 35MW, including current renewable energy output + cross-regional output from other power sources... If the available capacity at that moment is 50-35=15MW, then if the new energy output exceeds 15MW, the excess part needs to be absorbed by energy storage (otherwise it will cause channel overload). The predicted excess output is equal to the predicted output of new energy - (the predicted local load value + the upper limit of the external transmission channel). The upper limit of the external transmission channel represents the maximum allowable transmission power of the physical transmission channel of the new energy power station connected to the grid. The two are compared. If the real-time excess output or the predicted excess output is greater than zero, the energy storage charging absorption process is triggered. If the real-time excess output and the predicted excess output are less than zero, the energy storage remains in standby mode. For the energy storage charging absorption process, corresponding constraints are obtained, including energy storage safety constraints (current SOC + (charging power × time × charging efficiency) ≤ SOC upper limit), grid constraints (actual wind and solar output + energy storage charging power ≤ maximum power of external transmission channel), and equipment constraints (charging power ≤ rated charging power of energy storage). Based on the principle of matching excess fluctuations and meeting constraints, the charging process is divided into three 5-minute sub-periods within a 15-minute control time domain, generating a stepped power curve. Specifically, in the first 5-minute period, the power is set at real-time excess output × 0.9, with a 10% margin to cope with sudden fluctuations. In the second 5-minute period, the power is set at the predicted average excess output × 0.95, with the margin reduced to 5%. In the third 5-minute period, the power is set at the predicted final excess output × 0.85, with a 15% emergency margin. At the same time, the stepped curve is converted into digital instructions that the energy storage system can recognize, generating new energy consumption-type excess configuration information. Step 4: For grid off-peak excess scenarios, collect real-time load data of the entire grid or region through the grid SCADA system, with second-level data acquisition. Combined with the ARIMA time series algorithm, predict the load curve for the next 24 hours. When the load is lower than 40% of the total grid installed capacity or 60% of the daily average load for more than 2 hours, it is determined to be an off-peak period. Simultaneously record the load fluctuation range within the period. ≤±5% is a stable off-peak, and >±10% is a fluctuating off-peak. Statistically calculate the basic output range of wind power or photovoltaic power during the off-peak period. For example, the average output of wind power at night is 200MW and the minimum output is 150MW, while the output of photovoltaic power at night is 0. Use the LSTM algorithm to predict the output fluctuation. At the same time, collect grid connection point frequency data and the reserve capacity of the external transmission channel to obtain the precise boundary of the off-peak period, the excess power fluctuation range within the period, and the upper limit of the grid's allowed charging power. The system determines whether a grid dispatch command exists. If it does, the command value is used as the base reference. If not, the base charging power is calculated by multiplying the excess power by 0.8 to obtain the base charging demand. Simultaneously, the system acquires the energy storage status constraints and grid safety constraints. The energy storage status constraints include: calculating the remaining rechargeable capacity based on real-time BMS data = total energy storage capacity × (SOC safety limit 80% - current SOC) × charging efficiency. For example, if 20MWh of energy storage has a current SOC of 30% and an efficiency of 93%, then the remaining rechargeable capacity = 20 × 50% × 93% = 9.3MWh. The charging power demand is adjusted based on the remaining off-peak hours. The adjusted power is ≤ remaining rechargeable capacity ÷ remaining hours. The grid safety constraints include: if the grid connection point frequency is > 50.2Hz, the charging power can be increased to the base value × 1.2, but must not exceed the grid's allowable limit; if the frequency is < 49.9Hz, it is reduced to the base value × 0.7. The system also outputs the target charging power curves for different time periods during the off-peak hours. Based on the PCS rated power and the battery maximum charging rate, a stepped charging power curve is determined. One hour is divided into two 30-minute periods, with a fixed power set for each period. At the same time, a 5% power adjustment margin is reserved to deal with sudden overcapacity. A SOC protection threshold is set. When the SOC reaches 78% (close to the 80% upper limit), the charging power is automatically reduced to the minimum, generating grid off-peak type overcapacity configuration information. Step 5: For microgrid-type overcapacity scenarios, assess the output volatility of each distributed renewable energy source within the microgrid to obtain the dynamic output volatility range. Specifically, for wind power output, a Weibull distribution is used to fit the wind speed-output relationship; for photovoltaic output, a Beta distribution is used to fit the irradiance-output curve. Distribution parameters are determined using historical one-year output data, and the output volatility range at 95% confidence level is output. Combined with real-time data, a lightweight LSTM model is used to perform 10-minute ultra-short-term output prediction, with the prediction error controlled within 8%. The predicted values ​​are used to correct the historical statistical distribution. Simultaneously, the microgrid load is classified into primary, secondary, and tertiary loads. Primary loads are represented as critical, such as medical equipment and emergency lighting; secondary loads are represented as important, such as industrial production equipment; and tertiary loads are represented as... Generally, for example, office air conditioners and entertainment equipment, a normal distribution is used to fit the demand range of each load level. The mean and standard deviation are determined based on historical electricity consumption records, and the total load of the microgrid is obtained by summing them. Then, the predicted output of new energy sources is compared with the total load of the microgrid. If the predicted output of new energy sources minus the total load of the microgrid is greater than 5% of the rated load of the microgrid, or the microgrid frequency is greater than 50.2Hz, charging is started immediately. When the grid connection point voltage is greater than 1.05 times the rated voltage, the charging power is increased by 10%-20%. When the voltage is less than 1.02 times the rated voltage, the basic charging power is maintained. When the frequency is greater than 50.3Hz, the charging power is increased to the rated upper limit. When the frequency drops to 50.1Hz, the charging power is reduced to the basic charging power. If only the output is greater than the load but the difference is less than 5% of the rated load, standby is maintained, and microgrid-type excess configuration information is generated simultaneously. Step 6: Process the generated power generation failure analysis signal, calculate the difference between the total overdue power generation and the total power consumption, and record it as the shortfall power. At the same time, acquire the scenarios corresponding to the shortfall power and classify them into the new energy insufficient output type and the grid peak load type. Then, perform dynamic capacity configuration for the shortfall power of different scenario types. For scenarios where renewable energy output is insufficient, model predictive control is adopted. The optimal discharge power curve is calculated every 15 minutes. The 15-minute control time domain is divided into three 5-minute sub-periods. A fixed discharge power is output in each period, and the corresponding load level is obtained at the same time. Specifically, there are three levels: level 1, level 2, and level 3. The classification here is the same as that in step 5. Then, the shortfall power is predicted according to the formula = (predicted value of level 1 load + predicted value of level 2 load) - (predicted output of renewable energy + output of conventional power source). The fixed discharge power is adjusted according to the predicted shortfall power as the standard, and configuration information for renewable energy insufficient processing is generated simultaneously. For peak load scenarios, discharge capacity is allocated according to load level, prioritizing residential loads and critical loads such as hospitals / data centers. Interruptible industrial loads can be moderately restricted to avoid wasting energy storage capacity. The intersecting point of the new energy output curve and load forecast curve for the next 1-2 hours, output by the LSTM model, is used to calculate the gap duration. For example, if the intersection point is 20 minutes from the current time, the gap duration is determined to be 20 minutes; if the intersection point is 1.5 hours from the current time, it is determined to be 90 minutes. Then, the total energy storage capacity is obtained. The maximum release capacity is calculated according to the formula: maximum release capacity = total energy storage capacity × (current capacity - capacity lower limit) × discharge efficiency. At the same time, the allowable capacity based on the duration is calculated according to the formula: allowable capacity based on duration = min(maximum release capacity, capacity lower limit) × gap duration. The required discharge amount is also calculated according to different load levels. For primary loads, the primary load gap is calculated using the formula: Primary Load Gap = Primary Load Demand - (New Energy Output + Conventional Power Output). For secondary loads, the secondary load gap is calculated using the formula: Secondary Load Gap = Secondary Load Demand - Remaining Power Output. Combining these two formulas, the required capacity is calculated using the formula: Load-Based Demand Capacity = (Primary Load Gap + a × Secondary Load Gap) × Gap Duration, where a is the secondary load guarantee coefficient. When the capacity is sufficient, a = 1; when it is insufficient, a = 0.5. Then, the discharge capacity is determined using the formula: Discharge Capacity = min(Maximum Discharge Capacity, Allowable Capacity, Demand Capacity), generating peak load configuration information for the power grid.

[0024] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0025] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources, characterized in that, The method specifically includes the following steps: By combining AI algorithms to make ultra-short-term predictions for distributed new energy storage systems, the predicted output of new energy is obtained, and the benchmark value of total electricity consumption in the same period is determined. The two are compared. If the former is larger, a signal is generated that the power generation meets the analysis; otherwise, a signal is generated that the power generation does not meet the analysis. For power generation to meet the analysis signal, the difference between the predicted output of new energy and the benchmark value of total electricity consumption is recorded as the surplus electricity. The surplus electricity is classified into new energy consumption type, grid off-peak type, and microgrid type surplus, and dynamic configuration is performed for each of the three. For renewable energy consumption type, calculate the real-time excess output and the predicted excess output and compare their magnitudes. Generate configuration information based on matching excess and satisfying constraints. For grid off-peak type, determine the off-peak period and adjust the charging power according to grid dispatch instructions to generate configuration information. For microgrid type excess, determine the total load of the microgrid and adjust the power according to the relationship between the predicted renewable energy output and the total load to generate configuration information. If the power generation does not meet the analysis signal, the difference between the total power consumption benchmark value and the predicted output of new energy is recorded as the power gap, and the power gap is classified into two types: insufficient output of new energy and peak load of the power grid. Dynamic configuration is carried out for each type. For the type of insufficient output of new energy, the load is acquired and the load level is determined, the predicted shortfall power is calculated, and the power is adjusted according to the standard to generate configuration information. For the type of shortfall due to peak load of the power grid, the duration of the shortfall, the maximum release capacity of energy storage and the allowable capacity are determined, and configuration is carried out according to different load levels to generate configuration information.

2. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, AI algorithm-based ultra-short-term predictions include: Acquire real-time data, including energy storage parameters, energy and load parameters, and grid parameters, among which energy storage parameters include SOC, SOH, and temperature; Energy and load parameters include real-time output of photovoltaic / wind power and load in the grid connection area; grid parameters include grid connection point voltage and frequency. The LSTM+random forest algorithm is used to predict the power output of new energy sources in the next 10 minutes to 2 hours, and the power output of new energy sources is summarized to obtain the predicted power output of new energy sources.

3. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, The dynamic capacity allocation method for renewable energy consumption is as follows: Real-time excess output = current output of new energy - (local load + available capacity of external transmission channel), predicted excess output = predicted output of new energy - (predicted value of local load + upper limit of external transmission channel). If either output is >0, energy storage charging is triggered; if both are ≤0, standby mode is activated. Based on the principle of matching excess and satisfying constraints, time t is divided into n hourly periods, with the values ​​of t and n set by the operator. A stepped power curve corresponding to the small time period is generated, which is then converted into digital instructions to generate configuration information.

4. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, Off-peak excess capacity in the power grid includes: The power grid SCADA system collects load data in seconds and uses the ARIMA algorithm to predict the 24-hour load curve. If the load is lower than 40% of the total installed capacity or 60% of the daily average load for 2 consecutive hours, it is determined to be a trough and the fluctuation range is recorded. The system statistically analyzes the basic output range of wind power / solar power during off-peak hours, uses LSTM to predict output fluctuations, collects grid connection frequency and external transmission channel reserve capacity, determines off-peak boundaries, excess power ranges, and grid-allowed charging upper limits. If there are grid dispatch instructions, the instructions are used as the basis; otherwise, the basic charging demand is calculated as excess power × 0.

8. Combining energy storage and grid security constraints, the system outputs time-period target charging power curves. Based on the PCS rated power and the battery's maximum charging rate, time t1 is divided into two small cycles with a fixed power. The value of time t1 is set by the operator, with a 5% adjustment margin. When the SOC reaches 78%, the charging power is reduced to the minimum, and configuration information is generated.

5. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 4, characterized in that, The constraints include energy storage state constraints and grid security constraints; Energy storage state constraints include: remaining rechargeable capacity = total energy storage capacity × (80% of SOC safety limit - current SOC) × charging efficiency, and the corrected power ≤ remaining rechargeable capacity ÷ remaining off-peak duration. Grid safety constraints include: if the grid connection frequency is >50.2Hz, the charging power is increased to 1.2 times the base value and does not exceed the grid upper limit; if the frequency is <49.9Hz, it is reduced to 0.7 times the base value.

6. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, The configuration method for microgrid-type surplus scenarios is as follows: The dynamic output range is obtained by assessing the fluctuation of renewable energy output. The load is classified into three levels: Level 1, Level 2, and Level 3. The load demand range is fitted by a normal distribution and summed to obtain the total load of the microgrid. If the predicted renewable energy output minus the total load is greater than 5% of the microgrid's rated load, or if the frequency is greater than 50.2Hz, charging is initiated. If the voltage is greater than 1.05 times the rated value, the power is increased by 10%-20%, and if it is less than 1.02 times the rated value, the base power is maintained. If the frequency is greater than 50.3Hz, the power is fully utilized, and if it drops to 50.1Hz, the base power is reduced. If only the output is greater than the load and the difference is less than or equal to 5% of the rated load, the system is put into standby mode, and configuration information is generated.

7. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 6, characterized in that, The method for obtaining the dynamic output fluctuation range is as follows: For wind power output, the Weibull distribution is used to fit the wind speed-output relationship; for photovoltaic power output, the Beta distribution is used to fit the irradiance-output curve. The distribution parameters are determined by the historical output data of one year, and the output fluctuation range at a 95% confidence level is output. Combined with real-time data, a lightweight LSTM model is used to make ultra-short-term output predictions at the 10-minute level, with the prediction error controlled within 8%. The predicted values ​​are used to correct the historical statistical distribution.

8. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, The configuration method for areas with insufficient new energy output is as follows: Model predictive control is adopted, and the optimal discharge power curve is calculated on a rolling basis with time t3 as the standard. Time t3 is divided into several hour segments, and the value of t3 is set by the operator. The load level is obtained, and the power is adjusted according to the predicted shortfall power = (first-level + second-level load prediction value) - (new energy predicted output + conventional power output) to generate configuration information.

9. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 1, characterized in that, The configuration method for peak load-type power grid gaps is as follows: Discharge capacity is allocated according to load level. The LSTM outputs the new energy output and load curves for the next 1-2 hours. The intersection of the two is calculated to obtain the gap duration. The maximum energy storage capacity = total energy storage capacity × (current capacity - capacity lower limit) × discharge efficiency. The allowable capacity based on duration = min(maximum energy storage capacity, capacity lower limit) × gap duration. Then, the required discharge amount is calculated according to load level.

10. The dynamic capacity configuration method for energy storage systems applicable to distributed new energy sources according to claim 9, characterized in that, The method for calculating the required discharge volume based on different load levels is as follows: Primary load gap = Primary load demand - (New energy + conventional power output), Secondary load gap = Secondary load demand - Remaining power output, Demand capacity = (Primary load gap + a × Secondary load gap) × Gap duration, where a is the guarantee coefficient, Discharge capacity = min (Maximum discharge capacity, Allowable capacity, Demand capacity), Generate configuration information.

Images