Coordinated optimization method and device for shared electric-hydrogen hybrid energy storage based on multiple microgrids

Abstract

This invention discloses a method and apparatus for collaborative optimization of shared hydrogen-electric hybrid energy storage based on multiple microgrids, belonging to the field of energy storage collaborative optimization technology. It includes the following steps: S1, real-time acquisition of energy storage collaborative operation data and data preprocessing; S2, identification of net load fluctuation behavior, assessment of load change direction and fluctuation intensity, and determination of the regulation demand structure on fast and slow time scales; S3, assessment of the available regulation capability of hydrogen energy storage and determination of the feasibility boundary of slow time scale regulation demand; S4, establishment of a cost prediction model, generation of a shared energy storage capacity allocation scheme based on fast time scale regulation demand, and adaptive correction of scheduling instructions to achieve stable collaborative regulation across microgrids. This solves the problems of difficulty in cross-time scale regulation coordination and insufficient scheduling synchronization when shared hybrid energy storage in multiple microgrids exhibits inconsistent dynamic response characteristics.

Description

Technical Field

[0001] This invention relates to the field of energy storage collaborative optimization technology, specifically to a method and apparatus for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids. Background Technology

[0002] Against the backdrop of a continuously increasing proportion of renewable energy and an accelerated transformation of the energy system structure, the power system is gradually evolving from a traditional centralized supply model to an operational mode that combines distributed, multi-energy complementary, and flexible regulation. Energy storage technology, as a crucial infrastructure supporting renewable energy consumption, power balance, and stable system operation, is evolving from single-form, single-point deployment to a development direction of multi-type collaboration and cross-regional sharing. With the widespread application of microgrids, distributed energy resources, and the energy internet, the role of energy storage in the power system is constantly deepening, and its operation management, resource allocation, and regulation capabilities are of increasingly important macro-level significance in ensuring system security, economy, and flexibility.

[0003] For example, invention patent CN116165910A discloses an electronic switch circuit and a smart home system. The electronic switch circuit includes: an energy storage module; a charging module connected to the energy storage module, used to receive charging signals from external devices, process and regulate the charging signals, and output a charging voltage suitable for charging the energy storage module, so that the energy storage module stores electrical energy; a power supply regulation module connected to the energy storage module, used to receive the power supply voltage output by the energy storage module, process the power supply voltage, and output a power supply drive signal; and a control module connected to the power supply regulation module, used to generate a switch control signal under the action of the power supply drive signal, so that the electronic switch controls the on / off state between the external power supply and the load according to the switch control signal. This invention ensures the continuity of power supply to the load and improves the charging efficiency of the load.

[0004] For example, invention patent CN116578005A discloses an intelligent classroom energy-saving system based on the ISMA-BP network model, including an intelligent classroom system, a data acquisition unit, a load forecasting unit, and a photovoltaic energy storage unit. The intelligent classroom system includes a carbon dioxide sensor, a light sensor, a temperature sensor, an exhaust fan, intelligent lighting, and intelligent air conditioning. The data acquisition unit collects historical power load data and current power load data. The photovoltaic energy storage unit consists of multiple photovoltaic panels and lithium iron phosphate batteries, powering the exhaust fan, intelligent lighting, and intelligent air conditioning. The photovoltaic panels absorb solar energy to provide power and store excess energy. The photovoltaic energy storage unit charges and discharges according to the ISMA-BP-based power load forecast from the load forecasting unit, discharging during peak electricity consumption periods and charging during off-peak periods. Compared with existing technologies, this invention can effectively improve energy utilization, reduce the power grid's supply pressure, create an optimal educational environment, and save electricity.

[0005] However, in collaborative operation scenarios involving shared hybrid energy storage across multiple microgrids, batteries typically possess rapid response characteristics on the order of milliseconds to seconds, while electrolyzers and fuel cells, constrained by electrochemical inertia, experience response delays ranging from several seconds to several minutes. This difference in dynamic characteristics leads to inconsistent regulation rhythms in shared energy storage under rapid load fluctuations: batteries are subjected to excessive transient shocks, while hydrogen energy storage lags behind, making it difficult for collaborative optimization strategies to achieve stable solutions within a single timescale. Furthermore, traditional optimization models cannot characterize the dynamic coupling inertia between electric and hydrogen energy storage, resulting in synchronization offsets and control conflicts in cross-grid energy storage scheduling, impacting overall power balance and operational economy.

[0006] Therefore, in order to address the above issues, there is an urgent need for a collaborative optimization method and device for shared electric-hydrogen hybrid energy storage based on multiple microgrids. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a collaborative optimization method and device for shared electric-hydrogen hybrid energy storage based on multiple microgrids. This solves the problems of difficulty in cross-timescale adjustment and coordination and insufficient scheduling synchronization of shared hybrid energy storage in multiple microgrids under conditions of inconsistent dynamic response characteristics.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a collaborative optimization method for shared hydrogen-electric hybrid energy storage based on multiple microgrids, comprising the following steps: S1, real-time acquisition of collaborative energy storage operation data, and time alignment, anomaly suppression, and scale unification processing of the collaborative energy storage operation data to form an operation feature set; S2, identification of net load fluctuation behavior based on the operation feature set, assessment of load change direction and fluctuation intensity, and determination of the regulation demand structure on fast and slow time scales; S3, assessment of the available regulation capacity of hydrogen energy storage based on the operation feature set, and determination of the feasibility boundary of slow time scale regulation demand based on the available regulation capacity; S4, establishment of a cost prediction model using the operation feature set, and generation of a shared energy storage capacity allocation scheme based on the cost prediction model and fast time scale regulation demand, and adaptive correction of scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable collaborative regulation across microgrids.

[0011] Furthermore, the specific process of collecting real-time energy storage collaborative operation data and performing time alignment, anomaly suppression, and scale unification processing on the energy storage collaborative operation data to form an operational feature set is as follows: Real-time collection of energy storage collaborative operation data includes: load power of each microgrid, renewable energy generation power and real-time electricity price, as well as hydrogen storage tank pressure, hydrogen storage tank temperature, hydrogen flow rate, and fuel cell power, and the shared energy storage call records actually issued and executed within each scheduling cycle, and the synchronously acquired fuel cell rated power, hydrogen storage tank rated nominal pressure, and hydrogen lower heating value; where the scheduling cycle refers to the energy storage collaborative optimization and scheduling instructions. The process involves generating the minimum decision-making time unit; aligning the energy storage collaborative operation data using NTP time synchronization calibration; restoring data continuity by using nearest neighbor interpolation for instantaneous missing points in the energy storage collaborative operation data; removing and smoothing outliers using sliding median filtering; resampling the power data using fixed-step linear interpolation; resampling the hydrogen storage tank pressure, temperature, and flow rate data using piecewise averaging; performing min-max normalization on the energy storage collaborative operation data; and establishing an energy storage collaborative optimization database by writing the original and preprocessed energy storage collaborative operation data, along with timestamps, into the database.

[0012] Furthermore, the specific process for identifying net load fluctuation behavior based on the operating feature set, assessing the direction and intensity of load changes, and determining the adjustment demand structure on fast and slow time scales is as follows: Read the pre-processed load power and renewable energy generation power of each microgrid; calculate the difference between the load power and renewable energy generation power to obtain the net load power of each microgrid; calculate the net load change of each microgrid based on the net load power at adjacent times using differential calculation; extract the median of the absolute value of the net load change as the median deviation of the net load based on a fixed sliding time window; calculate the average value of the net load change within the same window to obtain the moving average of the net load; substitute the net load change into the sign function calculation... The net load change direction term is calculated; the net load change is divided by the sum of the net load median deviation and the minimum constant value to obtain the relative change amplitude of the net load, and the negative of the relative change amplitude of the net load is taken as the exponent for natural exponentiation. The fluctuation amplification term is obtained by subtracting the natural exponentiation result from the constant; the net load offset term is obtained by subtracting the net load sliding mean from the current net load change; the fast-regulation net demand value is obtained by multiplying the net load change direction term, the fluctuation amplification term, and the net load offset term; the slow-regulation net demand value is obtained by subtracting the fast-regulation net demand value from the net load power; the fast-regulation net demand value and the slow-regulation net demand value of each microgrid are calculated, and weighted summation is performed to obtain the total fast-regulation demand value and the total slow-regulation demand value.

[0013] Furthermore, the specific process for evaluating the available regulation capacity of hydrogen energy storage based on the operating characteristic set is as follows: Read the current hydrogen storage tank pressure, temperature, hydrogen flow rate, and fuel cell power; divide the difference between the hydrogen storage tank pressure and its lower limit by the difference between the rated nominal pressure and the lower limit to obtain the pressure margin value; divide the fuel cell power by the product of the hydrogen flow rate and the lower heating value of hydrogen to obtain the fuel cell efficiency value; based on the hydrogen flow rate within the historical window, perform least-squares fitting on the hydrogen flow rate within the window using the structure of a first-order inertial model to obtain the time constant; obtain the scheduling cycle length, divide the scheduling cycle length by the time constant, and take the negative number as the exponent for natural exponentiation, then subtract the natural exponentiation result from the constant to obtain the time inertia value; calculate the ratio of the actual fuel cell power to the rated fuel cell power, and select the ratio corresponding to the confidence quantile based on the sequence of ratios within the sliding time window as the reliability amplification factor; multiply the fuel cell efficiency value, pressure margin value, reliability amplification factor, and time inertia value to obtain the predicted value of hydrogen energy storage regulation capacity.

[0014] Furthermore, the specific process for determining the executability boundary of slow-timescale regulation demand based on available regulation capacity is as follows: the predicted value of hydrogen energy storage regulation capacity is used as the upper limit of slow regulation carrying capacity for the next scheduling cycle and written into the energy storage collaborative optimization database; the total slow regulation demand value is compared with the predicted value of hydrogen energy storage regulation capacity, and it is determined whether the absolute value of the total slow regulation demand value is less than or equal to the predicted value of hydrogen energy storage regulation capacity. If so, an execution permission flag is generated; otherwise, a restricted execution flag is generated.

[0015] Furthermore, the specific process of establishing a cost prediction model using the operating feature set is as follows: For each microgrid, the load power, renewable energy generation power, real-time electricity price, net load power, fast-regulation net demand value, slow-regulation net demand value, and shared energy storage call records corresponding to the microgrid in each scheduling cycle are combined into a microgrid scheduling feature set, and the actual settled operating cost in the scheduling cycle is obtained based on the real-time electricity price as a supervision label; the microgrid scheduling feature set and supervision label are used as inputs, and a cost prediction model is established using a supervised regression algorithm, outputting the non-shared operating cost prediction value and the shared operating cost prediction value for each microgrid.

[0016] Furthermore, combining the cost prediction model with fast timescale adjustment requirements, the specific process for generating a shared energy storage capacity allocation scheme is as follows: For each microgrid, the difference between the predicted non-shared operating cost and the predicted shared operating cost is calculated to obtain the operating cost improvement value. The operating cost improvement values ​​of all microgrids are summed to obtain the total cost improvement value. For the i-th microgrid, the ratio of the operating cost improvement value to the total cost improvement value is calculated to obtain the cost improvement ratio. The net load change of each microgrid within the sliding time window is obtained, and the standard deviation is calculated to obtain the net load fluctuation intensity value. The net load fluctuation intensity values ​​of all microgrids are then calculated. The total fluctuation intensity value is obtained by summing the values. The ratio of the net load fluctuation intensity value of the i-th microgrid to the total fluctuation intensity value is calculated to obtain the fluctuation allocation ratio. For each microgrid, the square root of the product of the cost improvement ratio and the fluctuation allocation ratio is obtained to obtain the comprehensive allocation weight value. The sum of the comprehensive allocation weight values ​​of all microgrids is calculated to obtain the total allocation weight value. The total fast adjustment demand value and the predicted value of hydrogen energy storage adjustment capacity are obtained and added together, and then multiplied by the scheduling cycle length to obtain the total shared energy storage capacity. The shared energy storage capacity allocation value is obtained by multiplying the total shared energy storage capacity by the ratio of the comprehensive allocation weight value of the i-th microgrid to the total allocation weight value.

[0017] Furthermore, during the execution phase, the specific process of adaptively correcting scheduling instructions based on equipment constraints and state feedback to achieve stable and coordinated control across microgrids is as follows: Calculate the shared energy storage capacity allocation value for each microgrid, and calculate the corresponding shared energy storage power constraint value based on the scheduling cycle length; compare the fast-regulation net demand value and slow-regulation net demand value corresponding to each microgrid with the shared energy storage power constraint value, and perform truncation correction to obtain the verified fast-regulation demand value and verified slow-regulation demand value for each microgrid; sum the verified fast-regulation demand value and verified slow-regulation demand value of all microgrids to obtain the verified total fast-regulation demand value and verified total slow-regulation demand value; based on the battery's rated maximum charging power and rated maximum discharging power, The verified total fast regulation demand value is truncated, and charging / discharging commands are issued to the battery converter based on the truncated verified total fast regulation demand value: when the verified total fast regulation demand value is positive, the battery is instructed to discharge; when the verified total fast regulation demand value is negative, the battery is instructed to charge. If an execution permission flag is detected, the hydrogen storage operating mode is adjusted according to the verified total slow regulation demand value: when the verified total slow regulation demand value is positive, the fuel cell power is increased; when the verified total slow regulation demand value is negative, the electrolyzer hydrogen production power is increased. If a restricted execution flag is detected, the amplitude of the verified total slow regulation demand value is limited to ensure that the actual execution power of the hydrogen storage does not exceed the predicted value of the hydrogen storage regulation capacity; at the same time, closed-loop correction is performed based on the actual scheduling execution.

[0018] Furthermore, the specific process of closed-loop correction based on actual scheduling execution is as follows: During the scheduling execution process, the actual electrochemical energy storage battery charging and discharging power, fuel cell power, and electrolyzer hydrogen production power are collected synchronously and compared with the corresponding scheduling instructions. The scheduling execution deviation is calculated and written into the energy storage collaborative optimization database, and the time constant and reliability amplification factor are corrected based on the execution deviation sequence.

[0019] The second aspect of this invention provides a shared electric-hydrogen hybrid energy storage collaborative optimization device based on multiple microgrids, comprising: a data acquisition and preprocessing module for real-time acquisition of energy storage collaborative operation data, performing time alignment, anomaly suppression, and scale unification processing on the energy storage collaborative operation data to form an operation feature set; a net load fluctuation demand decomposition module for identifying net load fluctuation behavior based on the operation feature set, assessing the direction and intensity of load changes, and determining the adjustment demand structure of fast and slow time scales; a hydrogen energy storage adjustment capability assessment module for assessing the available adjustment capability of hydrogen energy storage in conjunction with the operation feature set, and determining the executability boundary of slow time scale adjustment demand based on the available adjustment capability; and a shared energy storage capacity allocation module for establishing a cost prediction model using the operation feature set, generating a shared energy storage capacity allocation scheme by combining the cost prediction model with fast time scale adjustment demand, and adaptively correcting the scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable collaborative control across microgrids.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention analyzes the net load fluctuation behavior and distinguishes the adjustment demand structure of fast time scale and slow time scale, so that energy storage units with different dynamic response characteristics can participate in the adjustment within their respective suitable time scales, avoid adjustment conflicts caused by differences in response characteristics, and thus improve the power balance stability in the process of multi-microgrid collaborative operation.

[0023] (2) This invention evaluates the available adjustment capability of hydrogen energy storage by comprehensively considering the operating status and dynamic inertia characteristics of hydrogen energy storage during the scheduling process, and determines the executable boundary of slow time scale adjustment demand accordingly, thereby avoiding adjustment commands that exceed the actual carrying capacity of hydrogen energy storage and improving the reliability and safety of scheduling decisions.

[0024] (3) By establishing a cost prediction model oriented to the operating characteristics of microgrids and combining the load fluctuation characteristics of each microgrid for comprehensive weight allocation, this invention can reasonably allocate shared energy storage capacity while meeting regulation needs, so that energy storage resources are tilted towards microgrids with higher economic improvement potential and regulation needs, thereby reducing overall operating costs.

[0025] (4) This invention introduces an execution deviation feedback mechanism in the scheduling execution phase to correct the key parameters in the energy storage scheduling model, so that the scheduling strategy can continuously and adaptively adjust with changes in the operating state, reduce the impact of prediction error and model deviation on the collaborative scheduling effect, and improve the long-term adaptability and robustness of multi-microgrid shared energy storage regulation.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 The flowchart shows the collaborative optimization method for shared electric-hydrogen hybrid energy storage based on multiple microgrids.

[0028] Figure 2 This is a module diagram of a shared electric-hydrogen hybrid energy storage collaborative optimization device based on multiple microgrids;

[0029] Figure 3 Multi-factor radar plot for predicting the regulation capacity of hydrogen energy storage;

[0030] Figure 4 This is a trend chart of predicted values ​​for hydrogen energy storage regulation capacity. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. As those skilled in the art will understand, 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.

[0032] Please see Figures 1-4 This invention provides a technical solution: a method and apparatus for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids, such as... Figure 1 As shown, the process includes the following steps: S1, real-time acquisition of energy storage collaborative operation data, time alignment, anomaly suppression, and scale unification processing of the energy storage collaborative operation data to form an operation feature set; S2, identification of net load fluctuation behavior based on the operation feature set, assessment of load change direction and fluctuation intensity, and determination of the regulation demand structure on fast and slow time scales; S3, assessment of the available regulation capacity of hydrogen energy storage based on the operation feature set, and determination of the feasibility boundary of slow time scale regulation demand based on the available regulation capacity; S4, establishment of a cost prediction model using the operation feature set, generation of a shared energy storage capacity allocation scheme based on the cost prediction model and fast time scale regulation demand, and adaptive correction of scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable collaborative regulation across microgrids.

[0033] Specifically, the process of real-time acquisition of energy storage collaborative operation data, followed by time alignment, anomaly suppression, and scale unification processing to form an operational feature set, is as follows: Real-time acquisition of energy storage collaborative operation data includes: load power, renewable energy generation power, and real-time electricity price of each microgrid; hydrogen storage tank pressure, hydrogen storage tank temperature, hydrogen flow rate, and fuel cell power; as well as the shared energy storage call records actually issued and executed within each scheduling cycle, and the synchronously acquired fuel cell rated power, hydrogen storage tank rated nominal pressure, and hydrogen lower heating value. The scheduling cycle refers to the minimum decision-making time unit for energy storage collaborative optimization and scheduling command generation, determined based on the microgrid's operational response speed and control strategy update frequency, preferably 5 to 30 seconds, to ensure both real-time performance and stability during the scheduling process. The load power is determined by the microgrid's... Real-time data is collected by grid-side smart meters; renewable energy generation power is output in real-time by grid-connected inverters and monitoring systems; real-time electricity prices are pushed by the electricity price interface service, which refers to the electricity price signal reflecting the marginal transaction cost of electricity within the current dispatch cycle. The preferred method is to use the day-ahead electricity price consistent with the dispatch cycle timescale, and the price signal is mapped to the corresponding dispatch cycle through time alignment; hydrogen storage tank pressure is obtained through a hydrogen storage tank pressure sensor; hydrogen storage tank temperature is sampled in real-time by temperature probes deployed in the hydrogen storage tank; hydrogen flow rate is measured by a hydrogen mass flow meter; fuel cell power is obtained by collecting the power value output by the fuel cell system controller; shared energy storage call records are automatically recorded by the energy management system (EMS) for dispatch instructions and execution feedback; fuel cell rated power, hydrogen storage tank rated nominal pressure, and hydrogen lower heating value are provided by equipment specifications and hydrogen property standards. NTP time synchronization calibration is used to align energy storage collaborative operation data; NTP is used to ensure consistency of data timestamps from different microgrids and different devices, enabling subsequent time series modeling and dynamic evaluation to be performed on a unified time axis. For instantaneous missing points in the energy storage collaborative operation data, the nearest neighbor interpolation method is used to restore data continuity.Suitable for scenarios with short missing intervals and gradual signal changes, local reconstruction using data from nearby time points helps maintain the true trend of physical quantities changing over time. Outliers are removed and smoothed using sliding median filtering, effectively suppressing spikes caused by sensor jitter, communication fluctuations, or transient interference, making the data distribution more consistent with the actual operating state of the equipment. Power data is resampled using fixed-step linear interpolation to maintain the continuity and high resolution of the time series. The step size is preferably consistent with the scheduling cycle length, ensuring that the resampled time series aligns with the scheduling decision time node, thereby guaranteeing... The timing consistency of scheduling calculations and execution is ensured; piecewise averaging is used to resample hydrogen storage tank pressure, temperature, and flow rate data to improve the stability of relatively slowly changing data, enabling different physical quantities to be used simultaneously for feature extraction on a unified time scale; min-max normalization is performed on energy storage collaborative operation data to facilitate subsequent model construction, feature calculation, and adjustment capability assessment of data with different dimensions, while avoiding the dominance of large-dimensional features on small-dimensional features; an energy storage collaborative optimization database is established, and the original and preprocessed energy storage collaborative operation data are written to the energy storage collaborative optimization database with timestamps.

[0034] In this implementation plan, by uniformly collecting, synchronizing, suppressing anomalies, and normalizing the scale of multi-source energy storage collaborative operation data, operational information from different microgrids and various types of equipment can participate in feature extraction and adjustment analysis under a consistent time base and unified dimensions, ensuring stable data quality, clear sources, and reliable physical meaning. Differentiated preprocessing strategies such as nearest neighbor interpolation, moving median filtering, linear interpolation, and piecewise averaging are employed to improve the accuracy of missing data repair and the robustness of anomaly suppression, making the time series smoother and more continuous. Simultaneously, a structured energy storage collaborative optimization database is established to achieve synchronous storage and traceability of raw and preprocessed data, providing a stable and reliable data foundation for subsequent net load demand decomposition, hydrogen energy storage regulation capacity assessment, and shared energy storage optimized scheduling, thereby significantly enhancing the engineering feasibility and real-time performance of the entire collaborative optimization method.

[0035] Specifically, the process of identifying net load fluctuation behavior based on the operating feature set, assessing the direction and intensity of load changes, and determining the regulation demand structure on fast and slow time scales is as follows: read the preprocessed load power and renewable energy generation power of each microgrid, calculate the difference between the load power and renewable energy generation power to obtain the net load power of each microgrid, and obtain the net load change of each microgrid through differential calculation based on the net load power at adjacent times; among them, the net load power reflects the amount of power that the microgrid needs to compensate externally at the current time and is the basic physical quantity for calculating regulation demand. Differential calculation is used to capture its transient characteristics that change over time. Based on a fixed sliding time window, the median of the absolute value of the net load change is extracted as the median deviation of the net load, and the average value of the net load change within the same window is calculated to obtain the sliding mean of the net load. The sliding time window is preferably 6 to 20 times the length of the dispatch cycle, based on the historical operation monitoring data of the microgrid and the dynamic response characteristics of the equipment. For example, when the dispatch cycle is 10 seconds, the window length can be set to 60 to 200 seconds to cover the short-term fluctuation cycle of the load, so as to smooth short-term fluctuations and prevent the judgment result from being affected by transient anomalies. The median can resist extreme value interference, while the average value is used to characterize the trend of change. The net load change is substituted into the sign function to calculate the direction of net load change, which is used to determine whether the load change is increasing or decreasing, providing a basis for the sign of subsequent fast and slow adjustment demand. The net load change is divided by the sum of the net load median deviation and the minimum constant value to obtain the relative change amplitude of net load. The negative of the relative change amplitude of net load is taken as the exponent for natural exponential operation. The fluctuation amplification term is obtained by subtracting the result of the natural exponential operation from the constant value. The minimum constant value is used to avoid the denominator being zero, and the exponential expression is used to nonlinearly enhance abnormally drastic or excessively slow changes, so that the fluctuation intensity is better mathematically mapped to the adjustment demand. The net load offset term is obtained by subtracting the net load sliding mean from the current net load change. This term is used to determine whether the current change deviates from the short-term trend, thus reflecting the difference between the "trend term" and the "sudden term" in load changes. Multiplying the net load change direction term, the fluctuation amplification term, and the net load offset term yields the fast-regulation net demand value, which characterizes the power regulation required on a fast timescale, such as within the battery response time range, and can be used for real-time rapid control. Subtracting the fast-regulation net demand value from the net load power yields the slow-regulation net demand value, which reflects the power demand that needs to be balanced on a longer timescale, such as within the hydrogen energy storage response range, to avoid relying solely on fast-regulation equipment to bear long-term offsets. The fast-regulation net demand value and the slow-regulation net demand value of each microgrid are calculated and then weighted and summed to obtain the total fast-regulation demand value and the total slow-regulation demand value. The weights in the weighted summation can be adaptively set according to the microgrid capacity ratio, load level, and actual operation strategy, ensuring that the impact of microgrids of different sizes on regional regulation demand is consistent with their actual contribution, thus obtaining the comprehensive regulation demand. The weight values ​​range from 0 to 1.

[0036] The specific formulas for the net demand value of fast adjustment and the net demand value of slow adjustment are as follows:

[0037] ;

[0038] In the formula, It represents the fast-adjustment net demand value of the i-th microgrid, which is used to characterize the instantaneous power adjustment demand of the microgrid on a fast time scale. It is a fast-adjustment indicator that reflects the degree and direction of net load mutation. It represents the slow-regulation net demand value of the i-th microgrid, which is used to characterize the power regulation demand of the microgrid on a slow time scale. It is a slow-regulation index that reflects the trend of net load shift. Indicates a microgrid index; This represents the net load power of the i-th microgrid at the current moment, serving as the baseline quantity for the current regulation target; This represents the net load power of the i-th microgrid at the previous moment, used to determine the direction and magnitude of the change; The net load median deviation of the i-th microgrid is used as a robust scaling factor to normalize the current variation. This represents the moving average of the net load of the i-th microgrid, providing the normal baseline for the current period, and is used to calculate the offset term; This represents a very small constant value. By avoiding zero denominator, preventing excessive amplification of adjustment, and limiting the lower limit of sensitivity, the numerical stability, physical rationality, and engineering robustness of the adjustment model are guaranteed. The preferred value is between 0.05 and 0.2.

[0039] In this implementation scheme, the direction, magnitude, and trend deviation of net load changes are extracted through a sliding time window. Combined with the sign function, exponential amplification, and offset decomposition mechanism, the load dynamics of the microgrid are separated into fast and slow regulation demands, allowing for the differentiated expression of transient disturbances and trend deviations. At the same time, the demands of multiple microgrids are aggregated using weights set by capacity and operating characteristics. This ensures that the resulting total regulation demand reflects the real fluctuation contribution of local networks while avoiding the disproportionate impact of a single microgrid on regional regulation. This achieves multi-timescale decoupling characterization of load fluctuations, laying a feasible, accurate, and physically consistent regulation foundation for the matched regulation of batteries and hydrogen storage.

[0040] Specifically, the process of assessing the available adjustability of hydrogen storage based on operational characteristic sets is as follows: Read the current hydrogen storage tank pressure, temperature, hydrogen flow rate, and fuel cell power; divide the difference between the hydrogen storage tank pressure and its lower pressure limit by the difference between the rated nominal pressure and the lower pressure limit to obtain the pressure margin value. The lower pressure limit originates from the operational safety requirements of the hydrogen storage device, such as the minimum hydrogen supply pressure limit for the fuel cell or the minimum allowable pressure specified in national standards for hydrogen storage tanks. It is typically defined in the equipment specifications as the minimum safe pressure value within the allowable operating range, preferably between 0.2 and 5. The pressure margin value reflects the available pressure space that the current hydrogen storage can maintain for hydrogen release or production operations, and is an important physical indicator for measuring the adjustability margin of the hydrogen system. The fuel cell efficiency is obtained by dividing the fuel cell power by the product of the hydrogen flow rate and the lower heating value (LHV) of hydrogen. The LHV can be set to 120 according to hydrogen property standards, used to convert the hydrogen flow rate into equivalent input energy, giving the efficiency calculation a unified physical meaning. The fuel cell efficiency reflects the current energy conversion state and is used to evaluate the fuel cell's ability to maintain or increase output power. Based on the hydrogen flow rate within a historical window, the time constant is obtained by performing least-squares fitting on the hydrogen flow rate within the window using the structure of a first-order inertial model. The length of the historical window is determined based on the dynamic response characteristics of the fuel cell and electrolyzer, preferably 6 to 20 times the scheduling cycle length, to cover the response trajectory of hydrogen flow rate under typical load variation conditions. The first-order inertial model can reasonably describe the transition dynamics of the hydrogen system under changes in gas supply or consumption. Its time constant reflects the characteristic time required for hydrogen storage to complete a step response and is an important parameter for determining the feasibility of slow regulation. The discretized response of the first-order inertial model can be expressed as: In the formula and This indicates the hydrogen flow rate at two adjacent moments within the window. Indicates the attenuation coefficient. The offset term, the attenuation coefficient, and the offset term are obtained simultaneously by least-squares fitting; the time constant is... The calculation yielded, where Represents the time constant. Indicates the length of the scheduling period. The attenuation coefficient is represented by this factor. The scheduling cycle length is obtained, divided by the time constant, and the negative of this division is used as the exponent for natural exponential calculation. The result of the natural exponential calculation is then subtracted from the constant to obtain the time inertia value. Here, the scheduling cycle length is the previously defined minimum decision unit for scheduling, and the time inertia value describes the dynamic adjustment capability that hydrogen storage can release within the current scheduling cycle. The exponential form ensures that the nonlinear trend of "weak adjustment capability in short cycles and strong adjustment capability in long cycles" is physically satisfied. The ratio of the actual fuel cell power to the rated power of the fuel cell is calculated, and the ratio corresponding to the confidence quantile is selected as the reliability amplification factor based on the sequence of ratios within the sliding time window. The ratio record within the sliding window is used to characterize the stability and output reliability of the fuel cell in historical operation. The confidence quantile is preferably 70% to 95% to exclude extreme operating conditions, making the prediction capability conservative and engineering feasible. The reliability amplification factor reflects the changing trend of fuel cell performance with operating conditions, providing robustness for capability prediction. The predicted value of hydrogen storage regulation capability is obtained by multiplying the fuel cell efficiency value, pressure margin value, reliability amplification factor, and time inertia value. The predicted value of hydrogen energy storage regulation capacity is used to characterize the maximum regulation range that hydrogen energy storage can perform in the next scheduling cycle. It is the core parameter for judging whether slow regulation demand can be undertaken by hydrogen energy storage, and provides a clear physical basis for subsequent scheduling boundary determination and power allocation.

[0041] The specific formula for predicting the hydrogen energy storage regulation capacity is as follows:

[0042] ;

[0043] In the formula, It represents the predicted value of hydrogen storage regulation capacity, which indicates the upper limit of the slow-timescale regulation capacity that hydrogen storage can provide within the current scheduling cycle. It is used to quantify the range of executable adjustable power under the combined effects of the remaining hydrogen in the hydrogen storage tank, fuel cell efficiency, system reliability, and dynamic response characteristics. This represents the power output of the fuel cell, which is the current electrical power that the fuel cell can output. It is the basic power source for hydrogen energy storage to participate in slow regulation within this scheduling cycle. This represents the hydrogen flow rate, indicating the actual mass flow rate of hydrogen consumed by the fuel cell per unit time, used to estimate the releasable energy corresponding to hydrogen consumption. This represents the lower heating value of hydrogen, indicating the effective chemical energy that can be released per unit mass of hydrogen. It is used to convert the hydrogen flow rate into the upper limit of the ideal energy that the fuel cell can release, and its value is 120. This indicates the pressure of the hydrogen storage tank, used to characterize the hydrogen storage level, and can be understood as the current amount of hydrogen available for adjustment. This indicates the lower limit of the hydrogen storage tank pressure, which is the minimum safe operating pressure allowed by the hydrogen storage system. It ensures that the pressure of the hydrogen storage tank will not drop to a critical value that would endanger safety or affect the lifespan during the operation of the fuel cell. It is determined according to the hydrogen storage tank grade and equipment specifications, and the preferred value range is between 0.2 and 5. This indicates the rated nominal pressure of the hydrogen storage tank, which is the rated design pressure of the hydrogen storage tank and serves as a reference value under normal operating conditions for hydrogen energy storage. This represents the reliability amplification factor, a fuel cell output power stability factor obtained through historical sequence statistics, used to mitigate the impact of outliers and make predictions more robust; This represents the length of the scheduling cycle and the minimum time unit for hybrid energy storage optimization and scheduling instruction updates. It represents the time constant, used to characterize the response inertia of hydrogen energy storage caused by electrochemical processes.

[0044] In this embodiment, Table 1 is a data table of predicted hydrogen energy storage regulation capacity. The lower limit of the hydrogen storage tank pressure is 2.0, the rated nominal pressure of the hydrogen storage tank is 10.0, and the scheduling cycle length is 10. The table records in detail the fuel cell power, hydrogen flow rate, hydrogen storage tank pressure, reliability amplification factor, time constant, and predicted hydrogen energy storage regulation capacity at five different times. Specifically, at time 1, the fuel cell power is 12, the hydrogen flow rate is 0.08, the hydrogen storage tank pressure is 5.0, the reliability amplification factor is 0.80, the time constant is 40, and the predicted hydrogen energy storage regulation capacity is 0.08295. At time 2, the fuel cell power is 18, the hydrogen flow rate is 0.10, the hydrogen storage tank pressure is 7.0, the reliability amplification factor is 0.90, the time constant is 50, and the predicted hydrogen energy storage regulation capacity is... The value is 0.15295; the fuel cell power at time 3 is 25, the hydrogen flow rate is 0.12, the hydrogen storage tank pressure is 9.0, the reliability amplification factor is 0.95, the time constant is 60, and the predicted value of hydrogen energy storage regulation capability is 0.22155; the fuel cell power at time 4 is 15, the hydrogen flow rate is 0.07, the hydrogen storage tank pressure is 4.0, the reliability amplification factor is 0.85, the time constant is 35, and the predicted value of hydrogen energy storage regulation capability is 0.09431; the fuel cell power at time 5 is 30, the hydrogen flow rate is 0.15, the hydrogen storage tank pressure is 10.0, the reliability amplification factor is 1.00, the time constant is 70, and the predicted value of hydrogen energy storage regulation capability is 0.22187.

[0045] Table 1. Predicted Values ​​of Hydrogen Storage Regulation Capacity

[0046]

[0047] like Figure 3The image shows a multi-factor radar chart of the predicted features of hydrogen energy storage regulation capacity. It displays key characteristic parameters used to calculate the predicted value of hydrogen energy storage regulation capacity at five different time points, including fuel cell power, hydrogen flow rate, hydrogen storage tank pressure, reliability amplification factor, and time constant. Each line corresponds to data at one time point, and the five-dimensional features are presented in a normalized polar coordinate system to intuitively reflect the relative magnitudes between multiple features and the distribution of feature changes at different time points.

[0048] like Figure 4 The figure shows the trend of predicted hydrogen energy storage regulation capacity. It illustrates the time-series evolution of the predicted hydrogen energy storage regulation capacity, with the horizontal axis representing time number and the vertical axis representing the predicted hydrogen energy storage regulation capacity. (See Table 1 for details.) Figure 3 and Figure 4 It can be seen that the predicted value of hydrogen energy storage regulation capability exhibits a "rise, rise, fall, rise" pattern across five time points: it gradually increases from time 1 to time 3, indicating that the combined effects of fuel cell power, hydrogen flow rate, and hydrogen storage tank pressure give hydrogen energy storage a stronger regulation capability; a significant decrease occurs at time 4, reflecting that the input parameters at that time, such as lower fuel cell power, hydrogen flow rate, and pressure, collectively weaken the regulation capability; and a rebound occurs again at time 5, indicating that after the key factors return to a higher level, the available regulation capability of hydrogen energy storage improves accordingly. Figure 3 The multidimensional distribution of five feature parameters at different time points is shown, with time points 3 and 5 generally exhibiting higher parameter levels. Figure 4 The predicted regulation capacity at time 4 is completely consistent with the predicted high regulation capacity at time 5; however, time 4 is significantly lower in multiple dimensions, corresponding to the lowest predicted regulation capacity. The changing trend of hydrogen energy storage regulation capacity is highly consistent with the combined state of the five key operating characteristics, accurately reflecting the impact of these factors on the regulation capacity of hydrogen energy storage.

[0049] In this implementation plan, a quantifiable and predictable hydrogen energy storage regulation capability assessment mechanism is constructed through joint modeling of four key physical quantities: pressure margin, fuel cell efficiency, historical flow dynamic characteristics, and operational reliability. A first-order inertial model is used to obtain time constants and nonlinear time inertia values, enabling the dynamic response characteristics of hydrogen energy storage to be accurately reflected within the scheduling cycle. Combined with the capability prediction values ​​formed by safety boundaries, energy conversion characteristics, and operational stability, the feasible power range of hydrogen energy storage in the next scheduling cycle can be realistically characterized in engineering, thus making slow-timescale regulation verifiable and feasible, and providing a stable and reliable physical constraint basis for coordinated scheduling.

[0050] Specifically, the process of determining the executability boundary of slow-timescale regulation demand based on available regulation capacity is as follows: The predicted value of hydrogen energy storage regulation capacity is used as the upper limit of slow regulation carrying capacity for the next scheduling cycle and written into the energy storage collaborative optimization database. The upper limit of slow regulation carrying capacity is used to limit the maximum power regulation range that hydrogen energy storage can actually undertake in the next scheduling cycle, so that the subsequent scheduling process has clear physical boundary conditions and can avoid efficiency degradation or safety risks caused by overload operation of fuel cells or electrolyzers. The total slow regulation demand value is compared with the predicted value of hydrogen energy storage regulation capacity to determine whether the absolute value of the total slow regulation demand value is less than or equal to the predicted value of hydrogen energy storage regulation capacity. If so, an execution permission flag is generated; otherwise, a restricted execution flag is generated. Among them, the total slow regulation demand value is the result of weighted summation of slow regulation demands from multiple microgrids, reflecting the amount of power regulation that hydrogen energy storage needs to undertake in the current scheduling cycle, while the predicted value of hydrogen energy storage regulation capacity is a quantitative prediction of the maximum regulation range that hydrogen energy storage can perform in the cycle. The comparison process verifies whether the adjustment demand exceeds the safety and performance boundaries of hydrogen storage. If the total slow adjustment demand is within the capacity range, an execution permission flag is generated, indicating that hydrogen storage can be adjusted normally according to demand. If the demand exceeds the capacity range, a restricted execution flag is generated, indicating that subsequent adjustment amplitudes should be limited. The flag generation results are passed to the subsequent scheduling instruction generation module to drive differentiated control of fuel cell power enhancement or electrolyzer hydrogen production power adjustment processes, making the entire scheduling link a verifiable and closed-loop execution logic. The preferred data type for the flags is discrete logic variables, represented using a Boolean data format.

[0051] In this implementation plan, by establishing a clear upper limit for the slow regulation carrying capacity of hydrogen energy storage and verifying the feasibility of slow regulation demand using the predicted regulation capacity value, the scheduling process has verifiable physical boundaries. At the same time, by relying on the judgment mechanism of the allowable execution flag and the restricted execution flag, the regulation behavior of hydrogen energy storage is made safer, more controllable and in line with equipment operation constraints in engineering, thereby ensuring that the execution process of slow time scale regulation is stable, reliable and feasible.

[0052] Specifically, the process of establishing a cost prediction model using the operational feature set is as follows: For each microgrid, the load power, renewable energy generation power, real-time electricity price, net load power, fast-regulation net demand value, slow-regulation net demand value, and shared energy storage call records corresponding to the microgrid in each scheduling cycle are combined into a microgrid scheduling feature set. In each scheduling cycle, a one-to-one correspondence is performed using timestamps to ensure the feature construction has completeness and temporal consistency. The microgrid scheduling feature set is used to characterize the impact of the microgrid's operating status and potential scheduling behavior on operating costs in the current cycle, and serves as the basic input vector for the cost prediction model; and based on the actual... The actual operating cost settled within the dispatch cycle using the time-of-use electricity price is used as a monitoring label. This operating cost can be calculated according to regional electricity market settlement rules, for example, by summing the product of electricity consumption and the time-of-use real-time electricity price. The monitoring label guides the model to learn the impact of different dispatch behaviors on costs, ensuring that the prediction results conform to economic and physical logic. Using the microgrid dispatch feature set and the monitoring label as input, a supervised regression algorithm is used to establish a cost prediction model. Gradient boosting regression is preferred, as it can handle multidimensional nonlinear relationships and has good generalization ability. The model outputs the non-shared operating cost prediction and the shared operating cost prediction for each microgrid. The non-shared operating cost prediction simulates the independent operating cost of the microgrid without shared energy storage; the shared operating cost prediction simulates the economic performance of the microgrid after participating in cross-grid shared energy storage. The difference between the two reflects the potential economic benefits of the sharing strategy and is an important input for the subsequent shared energy storage capacity allocation mechanism, enabling the dispatch strategy to balance economy and fairness.

[0053] In this implementation plan, a microgrid dispatch feature set containing key operational state variables is constructed. Using actual settlement costs as a monitoring label, a high-precision cost prediction model is established using a gradient boosting regression algorithm. This enables a quantifiable assessment of the economics under both shared and non-shared operation modes. The model accurately characterizes the nonlinear impact of dispatch behavior, load characteristics, and electricity price changes on operating costs, providing a reliable basis for comparing the economic differences between shared and independent operation. This supports the subsequent optimized allocation of shared energy storage capacity, enabling coordinated dispatch to significantly improve overall economic efficiency while ensuring fairness.

[0054] Specifically, the process of generating a shared energy storage capacity allocation scheme by combining a cost prediction model with fast-scale adjustment requirements is as follows: For each microgrid, the difference between the predicted non-shared operating cost and the predicted shared operating cost is calculated to obtain the operating cost improvement value. The operating cost improvement value is used to characterize the degree of economic benefit improvement of the microgrid after participating in sharing, and is an important basis for measuring economic contribution. The operating cost improvement values ​​of all microgrids are summed to obtain the total cost improvement value. For the i-th microgrid, the ratio of the operating cost improvement value to the total cost improvement value is calculated to obtain the cost improvement ratio. The cost improvement ratio is used to reflect the contribution of the microgrid to the overall sharing benefits, so that the capacity allocation takes into account both economic fairness and benefit matching. The net load change of each microgrid within the sliding time window is obtained, and the standard deviation is calculated to obtain the net load fluctuation intensity value. The standard deviation is used to measure the severity of the fluctuation of the microgrid load over time, and is a quantitative indicator for judging the sensitivity to the fast-scale adjustment requirements of shared energy storage. The net load fluctuation intensity values ​​of all microgrids are summed to obtain the total fluctuation intensity value. The net load fluctuation intensity value and the total fluctuation intensity value of the i-th microgrid are calculated. The ratio is used to obtain the fluctuation allocation ratio, which reflects the relative proportion of the microgrid in the overall fluctuation, so that the allocation strategy can reflect the principle of "the greater the fluctuation, the higher the demand" in the regulation fairness. For each microgrid, the square root of the product of the cost improvement ratio and the fluctuation allocation ratio is used to obtain the comprehensive allocation weight value. The square root operation is used to mitigate extreme distribution differences and make the weights of economic factors and physical fluctuation factors more coordinated in capacity allocation, avoiding the excessive dominance of a single factor. The sum of the comprehensive allocation weight values ​​of all microgrids is calculated to obtain the total allocation weight value, which is used to achieve normalization and make the subsequent capacity allocation comparable and physically consistent. The total fast adjustment demand value and the predicted value of hydrogen energy storage regulation capacity are obtained, added together, and then multiplied by the scheduling cycle length to obtain the total shared energy storage capacity. Among them, the total fast adjustment demand value reflects the amount of fast power regulation that the battery side needs to respond to in the current cycle, and the predicted value of hydrogen energy storage regulation capacity reflects the executable upper limit of slow regulation. The sum of the two can constitute the total capacity range that can be reasonably scheduled in the current cycle. The scheduling cycle length is used to convert the power quantity into the capacity quantity, ensuring that the meaning is consistent and has engineering feasibility. The shared energy storage capacity allocation value is obtained by multiplying the total shared energy storage capacity by the ratio of the comprehensive allocation weight value of the i-th microgrid to the total allocation weight value. This value is used to indicate the amount of shared energy storage capacity that the microgrid can call upon in the next scheduling cycle, so as to realize the allocation of shared energy storage on demand, according to contribution and according to fluctuation characteristics in cross-microgrid scenarios, and make the scheduling strategy fair, interpretable and implementable.

[0055] The specific formula for allocating shared energy storage capacity is as follows:

[0056] ;

[0057] In the formula, This represents the shared energy storage capacity allocation value of the i-th microgrid, which represents the shared energy storage capacity allocated to the microgrid in the current scheduling cycle and is the final allocation result used to constrain the scheduling power. This represents the microgrid index; j represents the index variable used for traversing the microgrid set. The total shared energy storage capacity represents the total scale of shared energy storage that can be uniformly dispatched across all microgrids. It is the overall available energy storage capacity determined by both fast adjustment demand and hydrogen energy storage adjustment capability. The value of the non-shared operating cost of the i-th microgrid represents the predicted operating cost of the microgrid without calling shared energy storage. It is an important reference for measuring whether a microgrid needs external energy storage support. denoted as the predicted shared operating cost of the i-th microgrid, and denoted as the predicted operating cost of the microgrid after using shared energy storage, which is used to measure the degree of economic improvement brought about by the intervention of energy storage; It represents the total cost improvement value, characterizing the overall economic benefit scale of all microgrids after using shared energy storage, and is used as the denominator for economic improvement normalization, so that the allocation ratio can reflect the relative contribution of each microgrid to the overall economic improvement. The value represents the net load fluctuation intensity of the i-th microgrid, indicating the degree of fluctuation in the net load of the microgrid within the sliding window. It is an important indicator for measuring the difficulty of regulation, the urgency of energy storage demand, and the pressure of rapid regulation. The total fluctuation intensity value represents the sum of the net load fluctuation intensity of all microgrids. It is used to normalize the fluctuation contribution so that the proportion of regulation pressure brought by the fluctuation of each microgrid is fairly reflected in the whole.

[0058] In this implementation plan, a comprehensive allocation weight is constructed by combining the cost improvement ratio and the load fluctuation allocation ratio, so that the allocation of shared energy storage capacity among multiple microgrids can simultaneously take into account the maximization of economic benefits and the matching degree of physical regulation needs. By introducing a normalized weight system and a total shared capacity constructed based on the predicted value of fast regulation demand and hydrogen energy storage regulation capacity, the capacity calculation has clear physical meaning and the scheduling results are engineering-executable. Ultimately, the coordinated and unified allocation of shared energy storage on demand, by contribution, and by fluctuation in cross-microgrid scenarios is realized, which improves the overall operating economy, regulation efficiency, and collaborative stability.

[0059] Specifically, the process of adaptively correcting scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable and coordinated regulation across microgrids is as follows: The shared energy storage capacity allocation value for each microgrid is calculated, and the corresponding shared energy storage power constraint value is calculated in conjunction with the scheduling cycle length. The shared energy storage power constraint value limits the maximum charging and discharging power that a microgrid can utilize within the current scheduling cycle, ensuring that the allocated capacity is implemented in power form and meets the rated power limits and regulation speed requirements of the energy storage devices. The fast-regulation net demand value and slow-regulation net demand value for each microgrid are compared with the shared energy storage power constraint value and truncated to obtain the verified fast-regulation demand value and verified slow-regulation demand value for each microgrid. This truncation correction ensures that the regulation demand proposed by each microgrid does not exceed its allocated energy storage power limit, thereby preventing out-of-bounds instructions or excessive use of shared energy storage at the execution level. The verified fast regulation demand value and the verified slow regulation demand value of all microgrids are summed to obtain the verified total fast regulation demand value and the verified total slow regulation demand value. The summation process is used to form regional regulation commands across microgrids, so that the regulation actions can be uniformly executed at the level of shared energy storage devices, avoiding command conflicts caused by independent execution. Based on the battery's rated maximum charging power and rated maximum discharging power, the verified total fast regulation demand value is truncated. Charge and discharge commands are then issued to the battery converter based on the truncated verified total fast regulation demand value: when the verified total fast regulation demand value is positive, the battery is instructed to discharge; when the verified total fast regulation demand value is negative, the battery is instructed to charge. If an execution permission flag is detected, the hydrogen storage operating mode is adjusted according to the verified total slow regulation demand value: when the verified total slow regulation demand value is positive, the fuel cell power is increased; when the verified total slow regulation demand value is negative, the electrolyzer hydrogen production power is increased. The execution permission flag indicates that the hydrogen storage has sufficient pressure margin, dynamic capability, and efficiency conditions to fully undertake slow regulation tasks, enabling the fuel cell or electrolyzer to adjust its output within the rated ramp rate limit. If a restricted execution flag is detected, an amplitude limit is applied to the verified total slow regulation demand value to ensure that the actual execution power of hydrogen storage does not exceed the predicted value of hydrogen storage regulation capacity. This amplitude limit ensures that hydrogen storage can still perform some slow regulation tasks within the safety boundary even when capacity is insufficient, avoiding safety risks such as excessively rapid pressure drop, reactor overload, or electrolyzer overheating. Simultaneously, closed-loop corrections are performed based on actual scheduling execution.Among them, the battery charging and discharging power corresponding to the fast-adjustment net demand value is a continuously adjustable quantity, and the truncated demand value can be directly issued as a power command; while the slow-adjustment net demand value reflects the directional demand of hydrogen energy storage in energy balance. Since the fuel cell and the electrolyzer are responsible for hydrogen release and power generation and hydrogen production and energy absorption respectively, they are two physically independent execution units. Therefore, the slow-adjustment demand value cannot be directly output as a single power command. Instead, the corresponding execution equipment must be selected according to the symbol, and the current power must be increased or decreased to gradually approach the demand value within the range that the equipment can execute, thereby meeting the requirements of dynamic adjustability and equipment protection.

[0060] In this implementation plan, safe and controllable scheduling of shared electric-hydrogen hybrid energy storage is achieved through shared energy storage capacity constraints, cross-microgrid demand verification, hierarchical execution of fast and slow regulation, and boundary control of hydrogen energy storage capacity. This ensures that the regulation needs of each microgrid are legally implemented within the allocable power range, avoiding out-of-bounds calls; regional-level aggregation and unified issuance ensure consistency of cross-grid commands and prevent regulation conflicts; fast regulation is continuously executed by batteries, improving regulation speed and fluctuation suppression capabilities; slow regulation adaptively selects fuel cells or electrolyzers based on demand direction, and amplitude is limited in conjunction with capacity limits, enabling hydrogen energy storage to stably participate in regulation while meeting dynamic constraints and safety boundaries.

[0061] Specifically, the closed-loop correction process based on actual scheduling execution is as follows: During scheduling execution, the actual electrochemical energy storage battery charging and discharging power, fuel cell power, and electrolyzer hydrogen production power are simultaneously collected and compared with the corresponding scheduling instructions. The scheduling execution deviation is calculated and written into the energy storage collaborative optimization database. The actual power data is collected in real time by the battery converter measurement module, fuel cell controller, and electrolyzer power metering device, and written into the database using a unified timestamp to ensure accurate temporal consistency in deviation calculation. The deviation between the scheduling instructions and the actual executed values ​​characterizes the response lag, undertracking, or overshoot behavior of the energy storage equipment during dynamic operation due to ramp-up limitations, aging characteristics, or external disturbances. It is an important indicator for measuring the steady-state and dynamic performance of the equipment. The time constant and reliability amplification factor are then corrected based on the execution deviation sequence. The time constant is corrected based on the statistical trend of the execution deviation sequence. The mean square value of the deviation is smoothed using an exponential moving average, and the time constant is gradually adjusted according to the direction of the smoothed deviation change. When the smoothed mean square value of the deviation shows an upward trend within a continuous scheduling cycle, the time constant is adjusted upwards with a preset small step size factor. When the smoothed mean square value of the deviation continues to decrease, the time constant is adjusted downwards with the same or smaller step size factor, so that the dynamic response characteristics of the first-order inertial model gradually approach the actual equipment state. To ensure the stability of the correction process, the updated time constant is limited to a physically reasonable range. This range is determined based on the design parameters of the hydrogen energy storage equipment and historical operating statistics. Updates cease when the correction result reaches the boundary of this range, thus avoiding non-physical inertial amplification or excessive shortening. The reliability amplification factor is corrected based on the robust statistical quantiles of the execution deviation sequence. This is achieved by recalculating the confidence quantiles of the deviation distribution within a sliding time window and adjusting the reliability amplification factor using a quantile backoff mechanism. When the deviation quantiles exceed the confidence threshold, the reliability amplification factor is gradually reduced to reflect the increased uncertainty in fuel cell output capability. When the deviation quantiles fall back into the stable range, the reliability amplification factor is gradually restored to ensure consistency between the capability prediction and actual operating conditions. The correction process for the time constant and the reliability amplification factor is performed at most once per scheduling cycle and automatically converges after the deviation statistics enter the stable range, ceasing significant changes. This ensures the scheduling model possesses numerical stability, physical consistency, and engineering feasibility during long-term operation.

[0062] In this implementation scheme, by continuously monitoring and recording the deviation between the actual executed power and the dispatch command into a historical sequence, it is possible to identify lag, ramp-up limitations, and performance fluctuations in batteries, fuel cells, and electrolyzers during actual operation. Based on this, the time constant and reliability amplification factor are adaptively adjusted to ensure that the dynamic characteristics of hydrogen energy storage remain consistent with model predictions. This mechanism significantly improves the accuracy and robustness of regulation capability assessment, enabling subsequent dispatch commands to have higher accessibility and security at the execution end, thereby achieving long-term stable and coordinated control of shared energy storage.

[0063] Reference Figure 2 As shown, the second aspect of the present invention provides a shared electric-hydrogen hybrid energy storage collaborative optimization device based on multiple microgrids, applied to the aforementioned shared electric-hydrogen hybrid energy storage collaborative optimization method based on multiple microgrids. The device includes: a data acquisition and preprocessing module, used to acquire energy storage collaborative operation data in real time, and perform time alignment, anomaly suppression, and scale unification processing on the energy storage collaborative operation data to form an operation feature set; a net load fluctuation demand decomposition module, used to identify net load fluctuation behavior based on the operation feature set, assess the direction and intensity of load changes, and determine the adjustment demand structure for fast and slow time scales; a hydrogen energy storage regulation capability assessment module, used to assess the available regulation capability of hydrogen energy storage in conjunction with the operation feature set, and determine the executability boundary of slow time scale regulation demand based on the available regulation capability; and a shared energy storage capacity allocation module, used to establish a cost prediction model using the operation feature set, combine the cost prediction model with fast time scale regulation demand, generate a shared energy storage capacity allocation scheme, and adaptively correct scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable collaborative regulation across microgrids.

[0064] In this implementation plan, modular partitioning enables end-to-end collaborative processing of data acquisition, load fluctuation decomposition, hydrogen energy storage capacity assessment, and shared energy storage allocation. This allows for the simultaneous identification of regulation needs at different time scales, dynamic assessment of the feasibility of hydrogen energy storage, and generation of economical and feasible energy storage allocation schemes. Simultaneously, equipment feedback during the execution phase enables adaptive correction of scheduling commands, thereby ensuring the stability, safety, and efficiency of energy storage regulation across microgrids and significantly improving the regional energy regulation capacity and operational economy.

[0065] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0066] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. As those skilled in the art will understand, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A collaborative optimization method for shared electric-hydrogen hybrid energy storage based on multiple microgrids, characterized in that, Includes the following steps: S1 collects energy storage collaborative operation data in real time, performs time alignment, anomaly suppression and scale unification processing on the energy storage collaborative operation data, and forms an operation feature set; S2 identifies net load fluctuation behavior based on the operating feature set, assesses the direction and intensity of load changes, and determines the adjustment demand structure on fast and slow time scales. The specific process of identifying net load fluctuation behavior based on operational feature sets, assessing the direction and intensity of load changes, and determining the adjustment demand structure on fast and slow time scales is as follows: Read the preprocessed load power and renewable energy generation power of each microgrid, calculate the difference between the load power and renewable energy generation power to obtain the net load power of each microgrid, and obtain the net load change of each microgrid by differential calculation based on the net load power at adjacent time points. Based on a fixed sliding time window, the median of the absolute value of the net load change is extracted as the median deviation of the net load, and the average value of the net load change within the same window is calculated to obtain the moving average of the net load. The net load change is substituted into the sign function to calculate the net load change direction term; the net load change is divided by the net load median deviation and the minimum constant value to obtain the net load relative change amplitude; the negative of the net load relative change amplitude is taken as the exponent for natural exponent calculation; the fluctuation amplification term is obtained by subtracting the natural exponent calculation result from the constant; the net load offset term is obtained by subtracting the net load moving average value from the current net load change. Multiply the net load change direction term, the fluctuation amplification term, and the net load offset term to obtain the fast-adjustment net demand value, and subtract the fast-adjustment net demand value from the net load power to obtain the slow-adjustment net demand value. Calculate the net fast-regulation demand and net slow-regulation demand of each microgrid, and then perform a weighted summation to obtain the total fast-regulation demand and total slow-regulation demand. S3, combined with the operating feature set, assesses the available adjustability of hydrogen energy storage, and determines the feasibility boundary of slow timescale adjustment needs based on the available adjustability. S4 utilizes the operating feature set to establish a cost prediction model. Combining the cost prediction model with the fast timescale adjustment requirements, a shared energy storage capacity allocation scheme is generated. During the execution phase, scheduling instructions are adaptively corrected based on equipment constraints and state feedback to achieve stable and coordinated control across microgrids.

2. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process of acquiring real-time energy storage collaborative operation data, performing time alignment, anomaly suppression, and scale unification processing on the energy storage collaborative operation data to form an operation feature set is as follows: Real-time data collection of energy storage collaborative operation data includes: load power of each microgrid, renewable energy power generation and real-time electricity price, as well as hydrogen storage tank pressure, hydrogen storage tank temperature, hydrogen flow rate and fuel cell power, and shared energy storage call records actually issued and executed in each scheduling cycle, and synchronously acquired fuel cell rated power, hydrogen storage tank rated nominal pressure and hydrogen lower heating value; wherein, the scheduling cycle refers to the smallest decision time unit for energy storage collaborative optimization and scheduling instruction generation; NTP time synchronization calibration is used to align the energy storage collaborative operation data; for instantaneous missing points in the energy storage collaborative operation data, nearest neighbor interpolation is used to restore data continuity; outliers are removed and smoothed using sliding median filtering; power data is resampled using fixed-step linear interpolation, and hydrogen tank pressure, hydrogen tank temperature, and hydrogen flow rate data are resampled using piecewise averaging; minimum-maximum normalization is performed on the energy storage collaborative operation data; an energy storage collaborative optimization database is established, and the original and preprocessed energy storage collaborative operation data are written to the energy storage collaborative optimization database with timestamps.

3. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process for evaluating the available adjustability of hydrogen energy storage by combining operational feature sets is as follows: Read the current hydrogen storage tank pressure, hydrogen storage tank temperature, hydrogen flow rate, and fuel cell power; divide the difference between the hydrogen storage tank pressure and the lower limit of the hydrogen storage tank pressure by the difference between the rated nominal pressure of the hydrogen storage tank and the lower limit of the hydrogen storage tank pressure to obtain the pressure margin value; divide the fuel cell power by the product of the hydrogen flow rate and the lower heating value of hydrogen to obtain the fuel cell efficiency value. Based on the hydrogen flow rate within the historical window, the time constant is obtained by performing least squares fitting on the hydrogen flow rate within the window using the structure of the first-order inertial model; the scheduling cycle length is obtained, the scheduling cycle length is divided by the time constant, and the negative number is taken as the exponent for natural exponentiation, and then the time inertia value is obtained by subtracting the natural exponentiation result from the constant. Calculate the ratio of the actual fuel cell power to the rated power of the fuel cell, and select the ratio corresponding to the confidence quantile as the reliability amplification factor based on the sequence of ratios within the sliding time window; The predicted value of hydrogen energy storage regulation capability is obtained by multiplying the fuel cell efficiency value, pressure margin value, reliability amplification factor, and time inertia value.

4. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process for determining the executability boundary of slow-timescale adjustment requirements based on available adjustment capabilities is as follows: The predicted value of hydrogen energy storage regulation capacity is used as the upper limit of slow regulation carrying capacity in the next scheduling cycle and written into the energy storage collaborative optimization database. The total slow regulation demand value is compared with the predicted value of hydrogen energy storage regulation capacity. It is determined whether the absolute value of the total slow regulation demand value is less than or equal to the predicted value of hydrogen energy storage regulation capacity. If so, an execution permission flag is generated; otherwise, a restricted execution flag is generated.

5. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process of establishing a cost prediction model using the operational feature set is as follows: For each microgrid, the load power, renewable energy generation power, real-time electricity price, net load power, fast-adjustment net demand value, slow-adjustment net demand value, and shared energy storage call records corresponding to the microgrid in each scheduling cycle are combined into a microgrid scheduling feature set, and the actual settled operating cost in the scheduling cycle is obtained based on the real-time electricity price as a supervision label. Using the microgrid dispatch feature set and supervision labels as input, a cost prediction model is established using a supervised regression algorithm, and the predicted non-shared operating cost and the predicted shared operating cost of each microgrid are output.

6. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process of generating a shared energy storage capacity allocation scheme by combining the cost prediction model with fast time-scale adjustment requirements is as follows: For each microgrid, the difference between the non-shared operating cost forecast and the shared operating cost forecast is calculated to obtain the operating cost improvement value. The operating cost improvement values ​​of all microgrids are summed to obtain the total cost improvement value. For the i-th microgrid, the ratio of the operating cost improvement value to the total cost improvement value is calculated to obtain the cost improvement ratio. The net load change of each microgrid within the sliding time window is obtained, and the standard deviation is calculated to obtain the net load fluctuation intensity value. The net load fluctuation intensity values ​​of all microgrids are summed to obtain the total fluctuation intensity value. The ratio of the net load fluctuation intensity value of the i-th microgrid to the total fluctuation intensity value is calculated to obtain the fluctuation distribution ratio. For each microgrid, the square root of the product of the cost improvement ratio and the fluctuation allocation ratio is used to obtain the comprehensive allocation weight value. The sum of the comprehensive allocation weight values ​​of all microgrids is then calculated to obtain the total allocation weight value. Obtain the total fast regulation demand value and the predicted value of hydrogen energy storage regulation capacity, add them together, and then multiply by the scheduling cycle length to obtain the total shared energy storage capacity; multiply the total shared energy storage capacity by the ratio of the comprehensive allocation weight value of the i-th microgrid to the total allocation weight value to obtain the shared energy storage capacity allocation value.

7. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 1, characterized in that, The specific process of adaptively correcting scheduling instructions based on equipment constraints and state feedback during the execution phase to achieve stable and coordinated control across microgrids is as follows: Calculate the shared energy storage capacity allocation value for each microgrid, and calculate the corresponding shared energy storage power constraint value in combination with the scheduling cycle length; The fast-regulation net demand value and slow-regulation net demand value corresponding to each microgrid are compared with the shared energy storage power constraint value and truncated and corrected to obtain the verified fast-regulation demand value and verified slow-regulation demand value of each microgrid. The verified fast-regulation demand value and verified slow-regulation demand value of all microgrids are summed to obtain the verified total fast-regulation demand value and verified total slow-regulation demand value. Based on the battery's rated maximum charging power and rated maximum discharging power, the total fast regulation demand value after verification is truncated, and charging and discharging commands are issued to the battery converter according to the truncated total fast regulation demand value: when the verified total fast regulation demand value is positive, the battery is instructed to discharge. When the total fast adjustment demand value after verification is negative, instruct the battery to charge. If the execution permission flag is detected, the working mode of the hydrogen energy storage is adjusted according to the verified total slow adjustment demand value: when the verified total slow adjustment demand value is positive, the fuel cell power is increased; When the total slow regulation demand value after verification is negative, the hydrogen production power of the electrolyzer is increased; if a restricted execution flag is detected, the amplitude of the total slow regulation demand value after verification is limited so that the actual execution power of hydrogen energy storage does not exceed the predicted value of hydrogen energy storage regulation capacity; at the same time, closed-loop correction is performed based on the actual scheduling execution.

8. The method for collaborative optimization of shared electric-hydrogen hybrid energy storage based on multiple microgrids according to claim 7, characterized in that, The specific process of performing closed-loop correction based on actual scheduling execution is as follows: During the scheduling process, the actual charging and discharging power of the electrochemical energy storage battery, the power of the fuel cell, and the hydrogen production power of the electrolyzer are collected synchronously and compared with the corresponding scheduling instructions. The scheduling execution deviation is calculated and written into the energy storage collaborative optimization database, and the time constant and reliability amplification factor are corrected based on the execution deviation sequence.

9. A shared electric-hydrogen hybrid energy storage collaborative optimization device based on multiple microgrids, employing the shared electric-hydrogen hybrid energy storage collaborative optimization method based on multiple microgrids as described in any one of claims 1-8, characterized in that, include: The data acquisition and preprocessing module is used to acquire energy storage collaborative operation data in real time, and to perform time alignment, anomaly suppression and scale unification processing on the energy storage collaborative operation data to form an operation feature set. The net load fluctuation demand decomposition module is used to identify net load fluctuation behavior based on the operating feature set, assess the direction and intensity of load changes, and determine the adjustment demand structure on fast and slow time scales. The hydrogen energy storage regulation capability assessment module is used to assess the available regulation capability of hydrogen energy storage by combining the operating characteristic set, and to determine the feasibility boundary of regulation demand on a slow time scale based on the available regulation capability. The shared energy storage capacity allocation module is used to establish a cost prediction model using the operating feature set. Combining the cost prediction model with the fast time scale adjustment requirements, it generates a shared energy storage capacity allocation scheme. During the execution phase, it adaptively corrects the scheduling instructions based on equipment constraints and state feedback to achieve stable and coordinated control across microgrids.

Images