Intelligent energy regulation method based on distributed energy microgrid

Abstract

The application discloses a smart energy regulation and control method based on a distributed energy microgrid, relates to the technical field of smart energy regulation and control, and comprises the following steps: collecting multi-source heterogeneous state data, pre-processing the multi-source heterogeneous state data, and forming a standardized microgrid real-time operation data set; based on the standardized microgrid real-time operation data set, combining static electrical parameters, constructing an operation state model fusing dynamic topology and equipment constraints, and synchronously constructing a digital twin operation model corresponding to the physical microgrid according to the operation state model; based on the digital twin operation model, combining ultra-short-term prediction data, rolling simulating future operation trends of the microgrid, and according to the rolling simulation result. The application realizes active defense through risk grading quantification, optimizes strategies in combination with recovery capability evaluation, eliminates current risks while reserving sufficient regulation margins, significantly improves the resilience of the microgrid to continuous disturbances, and finally realizes high-reliability and sustainable smart energy regulation and control.

Description

Technical Field

[0001] This invention relates to the field of smart energy regulation technology, and in particular to a smart energy regulation method based on distributed energy microgrids. Background Technology

[0002] With the advancement of the "dual carbon" goals, the penetration rate of distributed energy sources, represented by photovoltaics and wind power, in microgrids is increasing. Microgrids have evolved from simple distribution network endpoints into independent operating units with source-grid-load-storage coordination capabilities. Current microgrid smart control technologies mainly rely on real-time monitoring data and short-term forecasting algorithms, using model predictive control (MPC) or rule-based expert systems to maintain power balance and voltage stability. Existing technologies typically employ deterministic optimization models, assuming that future load demand and renewable energy output are known or fixed values ​​with only minor errors, and then calculating the optimal equipment scheduling instructions. This approach performs well in scenarios with stable operating conditions and high forecast accuracy, effectively reducing operating costs and improving energy utilization, constituting the mainstream technical route for current microgrid energy management systems (EMS).

[0003] However, the limitations of existing technologies are gradually becoming apparent when faced with the strong randomness and volatility brought about by the high proportion of renewable energy integration. First, traditional methods are mostly based on "point prediction" for decision-making, lacking quantitative assessment of the distribution of prediction errors and extreme scenarios. Once the actual operating trajectory deviates significantly from the predicted value (such as a sudden drop in photovoltaic power due to sudden weather changes), the system often falls into a passive position due to the lack of dynamic reserve margin, and may even lead to frequency collapse or voltage over-limit. Second, existing risk assessments are mostly post-event alarms or single threshold judgments, lacking multi-dimensional quantification of the severity of risks (such as not comprehensively considering the duration, economic losses, and equipment health status), resulting in a "one-size-fits-all" control strategy that cannot distinguish between "rigid risks that must be eliminated immediately" and "defensive risks that suggest reserving resources." Therefore, existing microgrid control technologies mainly suffer from a single dimension of risk perception and a lack of quantitative classification of uncertainty, as well as a short-sighted strategy generation mechanism that ignores the evaluation of the system's subsequent recovery capabilities and equipment health constraints, leading to insufficient system resilience. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a smart energy regulation method based on distributed energy microgrids to solve the problems of lack of uncertainty quantification and classification and short-sighted strategy generation mechanisms.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a smart energy regulation method based on distributed energy microgrids, comprising: Collect multi-source heterogeneous state data and preprocess the multi-source heterogeneous state data to form a standardized microgrid real-time operation dataset; Based on the standardized microgrid real-time operation dataset, an operation status model integrating dynamic topology and equipment constraints is constructed by combining static electrical parameters. Based on the operation status model, a digital twin operation model corresponding to the physical microgrid is constructed synchronously. Based on the digital twin operation model, the future operation of the microgrid is simulated in rolling manner using ultra-short-term forecast data. Based on the rolling simulation results, the dominant disturbance source is identified by multi-dimensional risk quantification and confidence level classification, and physical sensitivity and attention models are integrated to generate hierarchical control demand results. Based on the results of the hierarchical regulation demand, a set of candidate regulation strategies is constructed and input into the digital twin operation model for parallel simulation. Based on the simulation results of the candidate strategies and the recovery margin evaluation, multi-objective decision-making is carried out to determine the target regulation strategy set and execute it. Collect the control results after execution, compare them with the expected effects of the strategy, and revise the operational status model.

[0007] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the acquisition of multi-source heterogeneous state data refers to deploying sensing terminals to collect multi-source raw heterogeneous state data and configuring wireless acquisition and auxiliary power supply units for restricted edge nodes.

[0008] Multi-source raw heterogeneous state data includes electrical quantities such as node voltage, node current, node power, bus frequency, energy storage state of charge, switch status, and grid-connected switching power, as well as equipment state quantities such as temperature and vibration.

[0009] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the step of preprocessing multi-source heterogeneous state data to form a standardized real-time operation dataset refers to adding timestamps and quality labels to the original multi-source heterogeneous state data and performing edge preprocessing to generate a standardized microgrid real-time operation dataset.

[0010] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, wherein: the construction of an operation state model that integrates dynamic topology and equipment constraints based on a standardized microgrid real-time operation dataset and static electrical parameters refers to the construction of a basic microgrid network model based on a standardized microgrid real-time operation dataset and static electrical parameters; A dynamic topology structure with node-branch associations is generated based on the microgrid basic network model, and the dynamic topology structure is loaded with the operating status of each node and the device constraint attributes to form a unified operating status model of the microgrid.

[0011] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the step of synchronously constructing a digital twin operation model corresponding to the physical microgrid based on the operation status model refers to using the unified operation status model of the microgrid, taking the equipment status variables and load attributes as supplementary constraints of the model, and synchronously constructing a digital twin operation model corresponding to the physical microgrid.

[0012] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the step of "based on digital twin operation model, combined with ultra-short-term forecast data to roll simulation of the future operation status of microgrid" refers to constructing a short-term rolling simulation scenario based on digital twin operation model through real-time mapping of physical state and embedding of health constraints. Based on short-term rolling simulation scenarios, the operation process of microgrids in the future short time domain is rolled out using ultra-short-term prediction data to obtain the future operating status of the microgrid in rolling simulation.

[0013] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the method involves: identifying the dominant disturbance source by integrating physical sensitivity and attention models based on rolling simulation results through multi-dimensional risk quantification and confidence level, and generating hierarchical regulation demand results; extracting power voltage branches and topology risk index sets based on the future operating status results of the rolling simulation microgrid, and extracting hierarchical frequency risk quantification results. A comprehensive severity is formed based on the risk indicator set and the stratified frequency risk quantification results, the operational risk results are quantified, and the confidence level of various risks is classified. Construct a time-series vector of real-time operational change characteristics that integrates risk boundary information based on the results of quantitative operational risk analysis; Dominant disturbance sources are identified through physical sensitivity pre-screening and lightweight classification model fine classification, and restricted nodes are marked; Based on the risk confidence level and the identification results of the dominant disturbance source, hierarchical microgrid control requirements are generated.

[0014] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the step of constructing a candidate regulation strategy set based on the hierarchical regulation demand results and inputting it into a digital twin operation model for parallel simulation refers to constructing a candidate regulation strategy set based on the hierarchical microgrid regulation demand results and performing digital twin parallel simulation based on the candidate regulation strategy set to form the strategy effect response results.

[0015] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the method involves: performing multi-objective decision-making based on candidate strategy simulation results and recoverability margin evaluation, determining the target regulation strategy set and executing the strategy effect response results, extracting recovery capability evaluation parameters, and forming a parameter set. The recovery capacity evaluation value is calculated based on the recovery capacity evaluation parameter set, and a comprehensive evaluation result is formed; Based on the comprehensive evaluation results, a tiered screening and ranking process is conducted, and target control strategies are determined. Based on the target control strategy, the device parameters are mapped to generate a set of executable control instructions, which are then reliably transmitted and sent to the end, driving the execution unit to form the actual control response result.

[0016] As a preferred embodiment of the smart energy regulation method based on distributed energy microgrids described in this invention, the following steps are taken: collecting the regulation results after execution and comparing them with the expected effect of the strategy, and correcting the operating state model to form a closed-loop regulation means forming the regulation execution result based on the actual control response result, extracting the actual operating state after execution to form the execution deviation comparison basis, calculating the execution deviation result, identifying the source of the deviation to form the model correction parameter result, and updating the model based on the model correction parameter result to form a closed-loop correction result.

[0017] The beneficial effects of this invention are as follows: By constructing a multi-dimensional risk quantification system that integrates probability distribution boundaries and equipment health, this invention subdivides risks into high-confidence rigid risks and low-confidence defensive risks, realizing a shift from "passive response" to "active defense." Secondly, combined with the "recovery capability evaluation" mechanism, the strategy selection stage not only assesses the elimination effect of current risks but also prioritizes strategies that can retain sufficient energy storage margin, voltage and frequency margin, and power supply capacity for critical loads, ensuring that the microgrid still has the resilience to cope with continuous disturbances after a single regulation, thereby achieving truly sustainable smart energy regulation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a smart energy regulation method based on distributed energy microgrids.

[0020] Figure 2 A flowchart for generating microgrid risk quantification and control requirements based on multi-source data fusion.

[0021] Figure 3 Flowchart for calculating and comprehensively evaluating recovery capacity assessment values. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-3 This is one embodiment of the present invention, which provides a smart energy regulation method based on a distributed energy microgrid, comprising the following steps: S1. Collect multi-source heterogeneous state data and preprocess the multi-source heterogeneous state data to form a standardized microgrid real-time operation dataset. S1.1: Deploy sensing terminals to collect multi-source raw heterogeneous state data.

[0026] Specifically, sensing terminals are deployed at distributed power generation nodes, energy storage nodes, load nodes, and grid connection interface nodes of the distributed energy microgrid to collect raw heterogeneous state data during the operation of the microgrid in real time.

[0027] The distributed power nodes include photovoltaic inverter nodes, wind power converter nodes, and gas micro-source nodes, etc. The energy storage node includes a battery energy storage converter node; The load nodes include critical load nodes and adjustable load nodes; The grid connection interface node includes a common connection point node; The sensing terminal may employ voltage sensors, current sensors, frequency detection units, switch quantity acquisition units, and temperature and vibration composite sensing units, etc., to continuously sample the operating status of the corresponding nodes.

[0028] It should be noted that the multi-source raw heterogeneous state data includes at least electrical quantities such as node voltage, node current, node power, bus frequency, energy storage state of charge, switching status, and grid-connected switching power, as well as equipment state quantities such as temperature and vibration.

[0029] S1.2: Configure wireless acquisition and auxiliary power supply units for restricted edge nodes.

[0030] Specifically, for edge nodes that are difficult to wire, geographically dispersed, or have limited power supply, low-power wireless sensing terminals are configured to wirelessly collect and upload the status data of the corresponding edge nodes. For edge nodes with limited power supply, an environmental energy collection module is further configured on the wireless sensing terminal as an auxiliary power supply unit. This configuration enables the limited edge nodes to maintain continuous collection and uploading of status data even when they cannot be directly connected to wired power or conventional power sources, thereby improving the continuity and integrity of the original heterogeneous status data.

[0031] The environmental energy harvesting module can harvest energy from equipment vibration, mechanical disturbance, or temperature difference, and after rectification, voltage stabilization, or energy storage, provide auxiliary power to the sensors, microprocessors, or wireless communication modules in the wireless sensing terminal.

[0032] S1.3: Add timestamps and quality labels to the original heterogeneous state data from multiple sources.

[0033] It should be noted that the timestamp is used to represent the time when the corresponding data was collected, so as to perform time alignment of data from different nodes in the future, and the quality label is used to represent the validity and reliability of the corresponding data.

[0034] Specifically, quality labels are generated based on factors such as communication link signal strength, data packet loss rate, sampling continuity, sensor self-test results, or the degree of data missing. For example, when the link signal strength corresponding to a certain data is lower than a preset threshold, the packet loss rate is higher than a preset range, or the sensor self-test is abnormal, the data is marked as low-confidence data. When the acquisition process is continuous and stable and the data is complete, the data is marked as high-confidence data. Through appending, timestamps, and quality labels, the original heterogeneous state data has a unified time reference and quality characterization basis, providing a basis for filtering, interpolation, and alignment in subsequent edge preprocessing.

[0035] It should be noted that the preset threshold and preset range are dynamically optimized based on the microgrid operating environment, equipment aging level and communication network load, with typical value ranges of [-105 dBm ~ -115 dBm] and [3% ~ 5%].

[0036] S1.4: Perform edge preprocessing on the multi-source raw heterogeneous state data with added timestamps and quality labels, and generate a standardized microgrid real-time running dataset.

[0037] Specifically, edge preprocessing is performed on the original state data of the attached label. The edge preprocessing includes data cleaning, outlier removal, missing value interpolation, dimension normalization and time sequence alignment. Furthermore, edge feature quantities that characterize the dynamic operation characteristics of the microgrid are extracted, including frequency change rate, power fluctuation rate, voltage fluctuation characteristics or flicker intensity, etc. Through edge preprocessing and feature extraction, a standardized real-time operation dataset of the microgrid is formed, which characterizes the current operation status and dynamic change trend of the microgrid.

[0038] Data cleaning is used to remove duplicate records, records with abnormal formats, or sampled records that are obviously distorted. Outlier removal is used to remove data that exceeds a preset physical range or exhibits abnormal changes. Missing data interpolation is used to fill in data that is missing in a short period of time; Dimensional normalization is used to convert data with different dimensions into a numerical form that can be processed uniformly. Order alignment is used to map data collected from different nodes to a unified timeline based on timestamps; It should be noted that the frequency change rate can be calculated based on the difference between the frequency measurements at adjacent sampling times, the power fluctuation rate can be calculated based on the power change and sampling time interval within adjacent sampling periods, and the voltage fluctuation characteristics can be obtained statistically based on the voltage offset or fluctuation amplitude within the preset sampling window.

[0039] S2. Based on the standardized microgrid real-time operation dataset, a dynamic topology and equipment constraint-integrated operation status model is constructed by combining static electrical parameters. A digital twin operation model corresponding to the physical microgrid is constructed synchronously based on the operation status model. S2.1: Based on the standardized microgrid real-time operation dataset, a basic microgrid network model is constructed by combining static electrical parameters.

[0040] It should be noted that the static electrical parameters are derived from microgrid design drawings, equipment nameplate parameters, protection setting parameter tables, or pre-entered equipment file information, specifically including line impedance, line rated current carrying capacity, transformer turns ratio, transformer rated capacity, distributed power source rated capacity, inverter capacity limit, energy storage rated energy and power limit, etc.

[0041] Specifically, data such as node voltage, current, power, frequency, switching status, and grid-connected switching power in the standardized microgrid real-time operation dataset are mapped to the microgrid electrical connection structure according to the corresponding nodes or branches, and combined with the static electrical parameters to form a microgrid basic network model for describing the microgrid infrastructure and basic operation boundary.

[0042] S2.2: Generate a dynamic topology structure of node-branch associations based on the microgrid basic network model.

[0043] It should be noted that the node-branch association topology is used to characterize the electrical connection relationships between distributed power nodes, energy storage nodes, load nodes, and grid-connected interface nodes, and to reflect the grid-connected status, islanded status, or partial reconfiguration status of the microgrid at the current moment.

[0044] Specifically, the connectivity between nodes is represented by node association tables, branch connection matrices, or adjacency matrices. Based on the switch status, branch on / off status, and grid connection interface status, a node-branch association topology is generated to characterize the current connectivity of the microgrid. When the status of a tie switch, feeder switch, or grid connection switch changes, the corresponding connection relationship in the node-branch association topology is updated, so that the topology can be dynamically refreshed as the microgrid's operating status changes, resulting in a real-time updated dynamic topology.

[0045] S2.3: Load the dynamic topology structure with the operating status of each node and the device constraint attributes to form a unified operating status model of the microgrid.

[0046] Specifically, based on the dynamic topology, each data point in the standardized microgrid real-time operation dataset is mapped and loaded onto the corresponding physical nodes and electrical branches of the dynamic topology. After completing the data mapping, the adjustable capability constraint attributes of each device are dynamically calculated and generated by combining the pre-stored static boundary parameters of the devices. The real-time operation status data is then integrated with the newly generated adjustable capability constraint attributes to construct a unified operation status model for the microgrid.

[0047] The specific generation logic for adjustable capability constraint attributes is as follows: Distributed power source constraints: Read the current output value of the distributed power source in the real-time operation data of the standardized microgrid, and calculate the difference with the pre-stored rated capacity to generate the dynamic adjustable capacity boundary of the node. Energy storage node constraints: Read the real-time state of charge (SOC) of the energy storage nodes in the real-time operation data of the standardized microgrid, and combine it with the pre-stored rated charging and discharging power and safe operation boundary to generate the charging and discharging power range constraints of the node through the preset SOC-power mapping rules. Branch safety constraints: Read the pre-stored branch rated current carrying capacity, voltage allowable deviation range and protection setting threshold, and generate the corresponding branch power flow over-limit boundary.

[0048] It should be noted that the unified operational status model of microgrids includes: Real-time status layer: includes the voltage / current / power / frequency status of each node, energy storage charge status, load demand power, grid-connected switching power, and current switch connection topology; Regulation and constraint layer: including the dynamic adjustable boundary of distributed power sources, the SOC-power mapping constraint of energy storage, the adjustable capacity level of load, and the power flow over-limit boundary of branches.

[0049] S2.4: Based on the unified operation status model of the microgrid, the equipment status variables and load attributes are used as supplementary constraints of the model, and a digital twin operation model corresponding to the physical microgrid is constructed simultaneously.

[0050] Specifically, when the standardized microgrid real-time operation dataset contains equipment status quantities such as temperature and vibration, a health index for key equipment is further generated. This health index is then used as a constraint attribute in the unified operation status model of the microgrid, together with the original node status, branch status, and equipment adjustability boundary attributes, to form a more complete microgrid operation status representation.

[0051] The health index is used to characterize the degree of restricted use of the corresponding equipment in subsequent regulation. The lower the health index, the worse the health status of the equipment. In subsequent regulation, the intensity of use should be reduced or the use should be avoided. A weighted scoring method is used to process the indicators. The temperature deviation score, vibration deviation score, and abnormal duration score are calculated by using an exponential decay function or a probability distribution function based on historical data statistics to reflect the nonlinear degradation characteristics of equipment performance. Preset weights are assigned to each score item. The preset weights can be dynamically adjusted according to the equipment type or operating conditions. Finally, the weighted sums are normalized to form a health index between 0 and 1 or between 0 and 100.

[0052] It should be noted that the indicators include the deviation of the actual temperature of the equipment from the rated operating temperature, the deviation of the vibration amplitude of the equipment from the reference vibration amplitude, the degree of abnormality of the vibration frequency band distribution of the equipment, and the duration of the equipment being in an abnormal state (such as overheating or severe vibration).

[0053] Based on the constructed unified operation state model of the microgrid, energy quality demand level attributes are added to each load node to characterize the tolerance range of the corresponding load for voltage deviation, frequency deviation, transient fluctuations and power outage duration. This allows the subsequent control process to consider not only whether the load is supplied with power, but also "with what power quality". By adding energy quality demand level attributes to each load node in the unified operation state model, the subsequent control strategy generation process can guarantee power quality differently according to load level, prioritize the operational stability of loads with high energy quality demand, and finally obtain the supplemented unified operation state model of the microgrid. A digital twin operation model corresponding to the physical microgrid is then constructed simultaneously on the edge computing platform or cloud.

[0054] It should be noted that the construction process includes mapping the node status, branch status, equipment constraint attributes, health index, and load energy quality demand level attributes in the supplemented microgrid unified operation status model according to the data structure of the digital twin model, and establishing a synchronous refresh mechanism with the real-time data of the physical microgrid, so that the digital twin operation model has: Status synchronization: Dynamically refreshes according to the real-time changes of the physical microgrid to maintain consistency between the virtual and physical networks; Attribute inheritance: Fully inherit all constraint attributes (including health index and energy quality requirement level) in the unified operating state model; Simulation and deduction: Supports "hypothesis-analysis" type future situation deduction based on the current state, providing a simulation environment for subsequent rolling simulation and strategy optimization.

[0055] S3. Based on the digital twin operation model, the future operation status of the microgrid is simulated in rolling manner by combining ultra-short-term forecast data. According to the rolling simulation results, the dominant disturbance source is identified by multi-dimensional risk quantification and confidence level classification, and physical sensitivity and attention model are integrated to generate hierarchical control demand results. S3.1: Based on the digital twin operation model, a short-term rolling simulation scenario is constructed by real-time mapping of physical state and embedding of health constraints.

[0056] Specifically, the node voltage, current, power, frequency, energy storage state of charge, grid-connected switching power, and switch connection relationships of the unified microgrid operation state model are written into the corresponding nodes and branches of the digital twin operation model. At the same time, the dynamic topology is synchronously mapped into the digital twin operation model, so that the digital twin operation model maintains the same network topology and operation boundary as the physical microgrid at the current moment. For devices with a health index lower than the preset lower limit threshold, when constructing a short-term rolling simulation scenario, the corresponding devices are marked as restricted adjustment nodes, and their call restriction conditions are synchronously written, so that the impact of restricted devices on the adjustment boundary can be reflected in the subsequent simulation process.

[0057] It should be noted that the preset lower limit threshold is set according to the equipment type (such as energy storage battery, diesel generator, photovoltaic inverter), the technical specifications provided by the equipment manufacturer, and the safety level requirements for microgrid operation. The typical value range is [0.70, 0.85], and the preferred value is 0.80. The basis for this is that for the core energy storage equipment in the microgrid, its health and cycle life have a non-linear relationship. Industry data shows that lithium-ion batteries are usually considered to have reached the "first end of life" when the SOH drops to 80%. After this inflection point, the battery internal resistance increases significantly, the heat generation rate rises sharply, and continuing to undertake high-frequency, high-power charging and discharging regulation tasks not only significantly reduces the regulation efficiency, but also increases the probability of thermal runaway exponentially. Therefore, using 0.80 as the preferred value can accurately capture the critical moment when the equipment changes from "normal aging" to "high-risk operation".

[0058] S3.2: Based on short-term rolling simulation scenarios, the operation process of microgrids in the future short time domain is rolled out using ultra-short-term prediction data to obtain the future operating status of the microgrid in rolling simulation.

[0059] Specifically, based on short-term rolling simulation scenarios, ultra-short-term prediction data is input into the digital twin operation model, and the microgrid operation process in the future short time domain is rolled out to obtain the future operation trajectory results of bus frequency, bus voltage, branch load rate, energy storage state of charge, grid-connected switching power, and key load power supply status over time.

[0060] The ultra-short-term forecast data includes distributed power generation output forecast data, load forecast data, and grid-connected power exchange change forecast data, etc. Among them, distributed power generation output forecast data can be generated based on historical output curves, current meteorological information, or short-term irradiance and wind speed change trends; load forecast data can be generated based on historical load curves, current load change rate, or time-period energy consumption characteristics; grid-connected power exchange change forecast data can be generated based on current grid connection point power change trends or scheduling plans.

[0061] The rolling simulation uses a preset time window and a preset time step to calculate the microgrid's operating status in the future short time domain step by step.

[0062] S3.3: Extract power voltage branch and topology risk index set based on the results of rolling simulation of the future operation status of microgrid.

[0063] Specifically, based on the power balance relationship between the total output of distributed power sources, energy storage power, grid-connected switching power, and total load demand in the future period, power imbalance indicators are calculated to identify power shortage risks and power surplus risks; based on the voltage prediction results of key nodes in the future operating trajectory, voltage limit exceedance indicators are calculated to identify voltage limit exceedance risks; based on the branch current prediction results in the future operating trajectory, branch overload risk indicators are calculated to identify branch overload risks; simultaneously, based on the topology connection relationships and interconnection status in the future operating trajectory, it is determined whether local power supply paths are restricted; when there are local branches that are unreachable, interconnection capabilities are reduced, or key loads have only a single power supply path, it is identified as a local topology restricted risk; when there are devices in the unified operating status model whose health index is lower than the preset lower limit threshold, and the devices still need to undertake adjustment tasks exceeding their allowable call boundaries in the future period, it is identified as a restricted device call risk, ultimately obtaining a set of basic risk indicators for power, voltage, branches, topology, and devices.

[0064] The formula for calculating the power imbalance index is as follows:

[0065] In the formula, It is an indicator of power imbalance. It is the total output of distributed power sources in the future period. This represents the charging and discharging power of the energy storage system at a given moment; positive values ​​are used for discharging and negative values ​​for charging. It refers to grid-connected power exchange capacity; positive values ​​are taken when purchasing electricity from the main grid, and negative values ​​are taken when supplying electricity to the main grid. It represents the total load demand power within the future time period.

[0066] when When, it is identified as a power deficit risk, when At that time, it was identified as a power surplus risk.

[0067] It should be noted that, The preset power deviation threshold is set using a quantile threshold setting method based on the rolling net power prediction error distribution. A typical value range is [0.90, 0.95], meaning the threshold is set using the 90%–95% quantile of the historical relative net power error rate. The preferred value is 0.95, based on the following... Used for risk identification at the front end, this system can cover most normal forecast error fluctuations for short-term microgrid regulation. Risk identification is only triggered when the future supply-demand deviation exceeds most of the historical normal fluctuation range. It is both conservative and not overly sensitive, making it most suitable for the rolling risk early warning scenario of "smart energy regulation". The formula for calculating the voltage over-limit index is as follows:

[0068] In the formula, It is a voltage over-limit indicator. It is a set of key nodes. It is the predicted voltage of the i-th critical node in the future time period. It is the upper limit of the allowed voltage for the i-th critical node. It is the lower limit of the allowable voltage for the i-th critical node.

[0069] It should be noted that when At that time, it was identified as a risk of voltage exceeding the limit.

[0070] , The normal operating voltage boundary setting method based on the per-unit value of the node rated voltage is adopted, with typical values ​​ranging from [1.03, 1.05] pu to [0.95, 0.97] pu, and preferred values ​​of 1.05 pu and 0.95 pu, respectively, based on the following... , Its purpose is to establish risk identification boundaries, not to set internal optimization targets for secondary controllers. Therefore, the optimal setting here is not to make it too tight, but to adopt the allowable boundary for normal operation. If it is set to a tighter internal target such as 0.96 / 1.04 pu, the risk identification will be too sensitive, and nodes that are still within the normal operating range will be identified as risks prematurely. If it is set to an emergency boundary such as 0.90 / 1.10 pu, the actual voltage risk will be identified too late.

[0071] The formula for calculating the branch overload risk index is as follows:

[0072] In the formula, It is a branch line overload risk indicator. It is a set of key branches. It is the predicted current of the k-th critical branch in the future time period. It is the allowable current-carrying boundary of the k-th critical branch.

[0073] It should be noted that when At that time, it was identified as a risk of branch circuit overload.

[0074] During the system modeling or control parameter initialization phase, the safe upper limit value of the operating current is pre-written into the parameter table of the key branch. The preset value is preferably set within the range of 90% to 100% of the rated operating current of the branch, and preferably 95% of the rated operating current of the corresponding key branch. This is based on the following... It is used for forward identification of overload risks under future operating trajectories, rather than the action boundary of relay protection; setting it to 95% of the rated operating current can both trigger risk identification in advance before the branch approaches overload and avoid normal fluctuations being mistakenly identified as overload risks due to the boundary setting being too low.

[0075] S3.4: Extract the stratified frequency risk quantification results based on the basic risk indicator set.

[0076] Specifically, based on the basic risk index set of power, voltage, branch, topology and equipment constraints, the frequency change rate risk index, frequency minimum point risk index and steady-state frequency risk index are further calculated. The frequency change rate exceeding the limit risk, frequency minimum point risk and steady-state frequency risk are extracted, and a hierarchical frequency risk quantification result is formed.

[0077] The formula for calculating the frequency change rate risk index is as follows:

[0078] In the formula, It is a risk indicator of the rate of change of frequency. It is the absolute value of the predicted rate of change of frequency over a future period. It is the maximum allowed rate of frequency change of the system.

[0079] when At that time, it was identified as a risk of exceeding the frequency change rate limit.

[0080] The formula for calculating the risk index at the lowest frequency point is:

[0081] In the formula, It is the risk indicator with the lowest frequency. It is the lowest frequency predicted in the future period. It is the system's rated frequency. This is the lower limit of the system's allowed frequency.

[0082] when At that time, it is identified as a risk of exceeding the limit at the lowest frequency point.

[0083] The setting method adopts the smaller value between the current-carrying capacity of the branch conductor and the rated current of the branch equipment, multiplied by the warning margin coefficient. The typical value range is [49.0, 49.5] Hz, and the preferred value is 49.0 Hz, based on the following... Used for identifying the lowest frequency point risk, that is, to determine whether the future lowest frequency point will exceed the safety lower limit. 49.0Hz corresponds to the "standard service" lower limit of NREL, which is easy to explain in engineering. Compared with 49.5 Hz, it is not too sensitive to short-term dynamic frequency fluctuations, avoiding too many false alarms. Compared with 48.0 Hz, it can detect the risk of frequency decline that will actually affect power supply quality and control stability earlier.

[0084] The formula for calculating the steady-state frequency risk index is as follows:

[0085] In the formula, It is a steady-state frequency risk indicator. It is the predicted steady-state frequency over a future period. It allows for steady-state frequency deviation.

[0086] It should be noted that when At that time, it was identified as a risk of steady-state frequency shift.

[0087] The allowable frequency deviation setting method based on steady-state service level is adopted, with a typical value range of [0.5, 1.0] Hz and a preferred value of 0.5 Hz, based on the following... It is used to determine whether the steady-state frequency has recovered to an acceptable operating range. 0.5Hz is more in line with the requirement that "good power quality should be achieved after steady-state recovery". It can prevent the system from operating in a state that is significantly deviated from the rated frequency for a long time, even though the system has not collapsed. It is very much in line with the steady-state control objectives in the current "smart energy control" scenario.

[0088] S3.4: Based on the risk indicator set and the stratified frequency risk quantification results, a comprehensive severity is formed, the operational risk results are quantified, and confidence levels are classified for various risks.

[0089] Specifically, the expected occurrence time of a risk is defined as the simulation moment when the corresponding risk indicator first exceeds a preset threshold, and the scope of impact is defined as the set of affected nodes, branches, or load areas. To ensure uniform weighting of all quantities in the comprehensive severity calculation, the magnitude, duration, increase in operating costs, and potential load shedding of the risk are first standardized and normalized, and the severity of the risk is calculated using a comprehensive severity evaluation formula. Based on the sufficiency of historical data, the stability of risk indicators, and the consistency of current prediction results, confidence levels are assigned to various types of risks. When a certain type of risk has sufficient historical samples, stable prediction results, and multiple risk indicators consistently point to the same risk conclusion, it is marked as a high-confidence rigid risk. When there is insufficient historical sample size for a certain type of risk, the prediction fluctuates greatly, or it only occurs in boundary scenarios, it is marked as a low-confidence defensive risk.

[0090] The formula for calculating the severity of the risk is as follows:

[0091] In the formula, It is the overall severity of the type l risk. It is the over-limit value after standardization of type l risk. It is the duration of risk after standardization for type l risk. This refers to the expected increase in operating costs caused by the standardized risk of type l. It is the potential load shedding amount caused by the standardized risk of type l. , , , It is the weighting coefficient.

[0092] It should be noted that, The risk severity weighting method based on the analytic hierarchy process is adopted, with a typical value range of [0,1], and the preferred values ​​are as follows: , , , Based on the priority of objectives that align with the current implementation scenario, The threshold value represents the current risk intensity and is the most direct measure of safety, therefore it has the highest weight. The potential load shedding capacity directly corresponds to the power supply resilience and critical load guarantee capability, which is extremely important in microgrid regulation and control, hence it is the second highest. Duration reflects whether the risk is a transient disturbance or a continuous deterioration; it is important but usually slightly less than the consequences of exceeding limits and load shedding. While the increase in cost cannot be ignored, in the current microgrid risk identification scenario, economic efficiency should be subordinate to security and power supply continuity, and therefore has the lowest weight; security priority > power supply continuity priority > dynamic sustainability > economic efficiency.

[0093] S3.5: Construct a time-series vector of real-time operational change characteristics that integrates risk boundary information based on the quantitative operational risk results.

[0094] Specifically, based on the quantitative operational risk results with risk level labels, the real-time change features are concatenated in chronological order to form a time-series vector of real-time operational change features. At the same time, the risk level labels, expected occurrence time sensitivity, and impact range information in the quantitative operational risk results are integrated so that the one-dimensional time-series feature vector not only reflects the real-time operational phenomena but also reflects the degree of uncertainty and impact boundary corresponding to the current risk.

[0095] The real-time change features include at least the power change rate, frequency change rate, voltage change rate, branch power flow abrupt change, grid-connected switching power abrupt change, and topology change marker.

[0096] S3.6: Based on physical sensitivity pre-screening and lightweight classification model, identify the dominant disturbance source and mark the restricted node.

[0097] Specifically, based on the degree of influence of different disturbance factors on the current risk indicators, the normalized sensitivity coefficient of each disturbance factor is calculated. If the sensitivity coefficient of a certain disturbance factor is significantly higher than that of other disturbance factors, it is selected as a candidate for the dominant disturbance source. The candidate disturbance factors, after being pre-screened by physical sensitivity, are then precisely classified and identified using a lightweight one-dimensional convolutional attention classification model. When there is a device in the unified operating state model with a health index lower than the preset range, and the device is within the influence range of the current quantified operating risk and participates in the current adjustment link, the device is marked as a restricted adjustment node. Furthermore, the lower limit of its maximum available adjustment power in the current scenario is calculated.

[0098] The formula for calculating the normalized sensitivity coefficient is as follows:

[0099] In the formula, It is the normalized sensitivity coefficient of the disturbance factor x to the type l risk indicator. It is the change in the type l risk indicator. It is the benchmark value for the type l risk indicator. It is the change in the disturbance factor. It is the baseline value corresponding to the disturbance factor.

[0100] The lightweight one-dimensional convolutional attention classification model extracts convolutional features and enhances attention on a one-dimensional temporal feature vector, and then obtains the probability distribution of each dominant perturbation source type through the classification output layer.

[0101] The classification output layer uses the Softmax classification function, and the calculation formula is as follows:

[0102] In the formula, It represents the probability that the current input feature vector belongs to the c-th type of perturbation source. is the score of the c-th type of disturbance source output by the classification model, and m is the total number of disturbance source types.

[0103] The final disturbance source type label is determined using the following formula:

[0104] In the formula, It is the label of the dominant disturbance source type that has been identified.

[0105] The formula for calculating the lower limit of the maximum available adjustable power is as follows:

[0106] In the formula, This is the lower limit of the maximum adjustable power available for the device in the current scenario. This is the rated adjustable power of the equipment. This is the health index of the device.

[0107] S3.7: Generate hierarchical microgrid control requirements based on risk confidence level and dominant disturbance source identification results.

[0108] Specifically, based on the quantitative operational risk results and the identification results of the dominant disturbance sources, the risk information and disturbance source information are merged, and hierarchical microgrid control demand results are generated according to the risk confidence level.

[0109] For high-confidence rigid risks, rigid control requirements are generated. These requirements include the minimum regulating power, minimum reserve capacity, minimum response speed, and necessary constraints required to eliminate the current risk. For low-confidence defensive risks, defensive control requirements are generated. These requirements include additional reserve power, additional reserve capacity, additional voltage or frequency safety margins recommended for potential extreme scenarios, and recommended flexible resource capacity. Simultaneously, the lower limit of the maximum available regulating power is set. The control demand results are written into the system so that the subsequent control strategy generation process can automatically avoid the over-limit calls of devices with insufficient health. In the end, the microgrid control demand results will include at least: risk type, expected occurrence time, scope of impact, severity, risk level label, type of dominant disturbance source, influence weight of each disturbance factor, rigid control demand, defensive control demand, and lower limit of maximum available regulation power.

[0110] The minimum regulating power, minimum reserve capacity, and minimum response speed are expressed as follows:

[0111] In the formula, It is the minimum regulating power. This is the minimum standby capacity. It is the minimum response speed. It refers to the power deficit, power surplus correction, or equivalent regulation demand power corresponding to the current risk. It is the minimum duration required to maintain this backup support. This is the maximum allowed response completion time.

[0112] The proposed additional reserve power is expressed as follows:

[0113] In the formula, It is recommended to reserve additional backup power in the context of defensive control requirements. It is a defensive reserve coefficient related to the risk level and the severity of the risk.

[0114] It should be noted that, A direct tiered preset setting method based on risk level labels is adopted, with a typical value range of [0.10, 0.30]. This means the defensive additional reserve power is taken as 10% to 30% of the rigid minimum adjustable power, with a preferred value of 0.20, based on the following... It is neither a market reserve coefficient nor a planned reserve rate, but an additional reserve coefficient for defensive regulatory needs. It needs to address the issue of reserving more resources than rigid needs in the context of "low confidence defensive risk" scenarios, without reserving too much and making subsequent strategy construction too conservative. 0.20 can better cover the uncertainty of prediction errors and boundary scenarios than 0.10, and is less likely to cause excessive conservatism and idle resources than 0.30.

[0115] The method of setting the duration directly based on a single defensive backup support window is adopted, with a typical value range of [10, 60] minutes, and a preferred value of 10 minutes, based on the following... Used in the generation of regulatory demand, the purpose is to provide a demand boundary for "at least how long to hold out" for subsequent strategy construction; the current scenario is the identification and rolling regulation of future short-term operational risks. The 10-minute window is more consistent with short-term rolling regulation and island support, which can ensure that subsequent regulation strategies cover at least one stable transition phase after a key disturbance. Without raising the reserve capacity requirement too high, it still has sufficient resilience significance.

[0116] The response time limit setting method is adopted by directly presetting the maximum allowable completion time of the microgrid fast control closed loop. The typical value range is [1,60]s, and the preferred value is 10s. The basis is that for fast resources such as energy storage and inverters, 10s is conservative enough and realistically achievable; for the microgrid controller, 10s is still within the "sec to min" time scale, which can be coordinated with the secondary control, constraint management and instruction execution closed loop.

[0117] S4. Construct a set of candidate control strategies based on the results of hierarchical control demand, and input them into the digital twin operation model for parallel simulation. Make multi-objective decisions based on the simulation results of the candidate strategies and the recovery margin evaluation, determine the target control strategy set and execute it. S4.1: Construct a candidate control strategy set based on the results of hierarchical microgrid control demand.

[0118] Specifically, based on the microgrid regulation requirements, multiple candidate regulation strategy sets are constructed according to risk type, dominant disturbance source type, impact range, severity, and restricted call node information, combined with constraint principles, to address the current microgrid operation risks.

[0119] When the dominant disturbance source is distributed generation fluctuation type, priority should be given to constructing candidate control strategies based on rapid support from energy storage and correction of grid-connected switching power; when the dominant disturbance source is load impact type, priority should be given to constructing candidate control strategies based on transient support from energy storage and adjustable load shaping; when the dominant disturbance source is topology change type, priority should be given to constructing candidate control strategies based on reactive power support, power flow redistribution, and local power supply path adjustment; when the dominant disturbance source is grid-connected switching fluctuation type, priority should be given to constructing candidate control strategies based on source-side output redistribution and grid-connected power correction; when the dominant disturbance source is equipment constraint type, alternative control strategies should be constructed by bypassing constrained nodes or reducing the degree of constraint node access.

[0120] The constraints are as follows: prioritize ensuring continuous power supply to critical loads; prioritize ensuring the power supply quality boundary of loads with high energy quality demand; prioritize calling energy storage nodes with fast response speed and normal health status; and avoid assigning adjustment tasks beyond their calling limits to devices with health index below the preset range.

[0121] It should be noted that each candidate control strategy in the candidate control strategy set includes at least the following control actions: adjusting the active power output of distributed power sources, adjusting the reactive power support of distributed power sources, adjusting the charging and discharging power of energy storage nodes, reducing or shifting peak loads for adjustable loads, adjusting the switching power of grid connection interfaces, and performing tie switch switching or partial power supply path reconfiguration.

[0122] S4.2: Perform parallel digital twin simulation based on the candidate control strategy set to generate strategy effect response results.

[0123] Specifically, based on the candidate control strategy set, the active power regulation, reactive power regulation, energy storage charging and discharging power, load switching, grid-connected switching power correction, and switch operation status corresponding to each candidate control strategy are written into the digital twin operation model. The microgrid operation process after the execution of each candidate control strategy is gradually simulated according to the same preset time window and time step. The strategy effect response results after the execution of each candidate control strategy are obtained through parallel simulation. When a candidate control strategy causes the bus voltage to exceed the allowable range, the bus frequency to exceed the allowable deviation range, the critical branch load rate to exceed the preset upper limit, or the critical load power supply to be interrupted during the simulation, the candidate control strategy is marked as an infeasible strategy and is eliminated or its priority is reduced in the subsequent comparison process.

[0124] The strategy's response results include at least the lowest frequency point, voltage recovery time, maximum branch load rate, critical load protection results, energy storage remaining charge status, grid-connected switching power variation, and local power supply path variation results.

[0125] S4.3: Extract recovery capability evaluation parameters based on the strategy effect response results and form a parameter set.

[0126] Specifically, the remaining regulation capacity parameters of energy storage are extracted from the post-execution power and current adjustable power boundary of energy storage in the strategy effect response results; the remaining capacity parameters of flexible loads are extracted from the remaining capacity after the execution of adjustable loads in the strategy effect response results; the voltage safety margin parameters of critical nodes are extracted from the predicted voltage values ​​and allowable voltage boundaries of critical nodes in the strategy effect response results; the frequency safety margin parameters are extracted from the predicted minimum frequency values, steady-state frequency values, and allowable frequency boundaries of the strategy effect response results; and the continuous power supply capacity parameters of critical loads are extracted from the power supply status of critical loads and the amount of potential critical load shedding in the strategy effect response results, thus forming a set of recovery capacity evaluation parameters.

[0127] It should be noted that the energy storage remaining regulation capacity parameter is used to characterize the rapid regulation capability that the energy storage node can still provide after the candidate regulation strategy is implemented; the flexible load remaining capacity parameter is used to characterize the adjustable load margin that can still participate in reduction, transfer or peak shifting control after the candidate regulation strategy is implemented; the critical node voltage safety margin parameter is used to characterize the remaining safety space of the critical node from the voltage allowable boundary after the current strategy is implemented; the frequency safety margin parameter is used to characterize the frequency withstand capability when facing the next disturbance after the candidate regulation strategy is implemented; and the critical load continuous power supply capability parameter is used to characterize the continuous power supply capability of the critical load in subsequent disturbance scenarios after the candidate regulation strategy is implemented.

[0128] S4.4: Calculate the recovery capability evaluation value based on the recovery capability evaluation parameter set, and form a comprehensive evaluation result.

[0129] Specifically, based on the recovery capability evaluation parameter set, the remaining regulation capability index of energy storage, the remaining capability index of flexible load, the voltage safety margin index of key nodes, the frequency safety margin index, and the continuous power supply capability index of key loads are calculated respectively. The existing weighted comprehensive evaluation method is used to calculate the recovery capability evaluation value. The recovery capability evaluation value, together with the frequency recovery evaluation result, voltage recovery evaluation result, branch load rate control evaluation result, and operating cost evaluation result corresponding to the candidate control strategy, constitutes the comprehensive evaluation result.

[0130] The formula for calculating the remaining energy storage regulation capacity index is as follows:

[0131] In the formula, It is an indicator of the remaining energy storage regulation capacity. It is the maximum regulation power that the energy storage node can provide under the current operating conditions after the candidate regulation strategy is implemented. This is the actual regulation power that the energy storage node has undertaken after the candidate regulation strategy has been implemented.

[0132] The formula for calculating the remaining capacity index of the flexible load is as follows:

[0133] In the formula, It is an indicator of remaining flexible load capacity. It refers to the remaining adjustable load capacity that can still participate in reduction, transfer, or peak-shifting control after the candidate control strategies are implemented. It is the maximum adjustable capacity of the adjustable load under the current operating conditions.

[0134] The formula for calculating the voltage safety margin index of the key node is as follows:

[0135] In the formula, It is a key node voltage safety margin indicator. It is a set of key nodes. It is the voltage value of the i-th critical node after the candidate control strategy is executed. , These are the upper and lower limits of the allowed voltage for the i-th critical node, respectively.

[0136] The formula for calculating the frequency safety margin index is as follows:

[0137] In the formula, It is a frequency safety margin indicator. It is the lowest frequency in the prediction time domain after the candidate control strategy is executed. It is the predicted steady-state frequency in the time domain after the candidate control strategy is executed. , These are the upper and lower limits of the system's allowed frequency, respectively. The formula for calculating the critical load continuous power supply capability index is as follows:

[0138] In the formula, It is an indicator of the continuous power supply capacity of critical loads. This refers to the power that critical loads may be cut off in subsequent predicted disturbance scenarios after the candidate control strategy is implemented. It is the total power of the critical load.

[0139] The formula for calculating the recovery ability evaluation value is as follows:

[0140] In the formula, It is a recovery ability evaluation value. , , , , These are the weighting coefficients for each evaluation indicator.

[0141] It should be noted that, The setting method is to directly preset the maximum adjustable power of the energy storage system. The typical value range is 0 to the rated power of the energy storage system, and the preferred value is 0.90 times the rated power. This is based on the fact that a 10% operating margin is retained to avoid pushing the equipment to its limit, which is more in line with the evaluation objective of "still having the ability to adjust for subsequent disturbances after the candidate strategy is executed".

[0142] The direct aggregation setting method based on the difference between the adjustable load baseline and the minimum service level is adopted. The typical value range is 0 to the sum of the theoretical maximum interruptible capacity of all adjustable loads under the current baseline. The aggregated value of "baseline load at the current moment - minimum service level load" is used as the preferred value.

[0143] The weight setting method is based on the Analytic Hierarchy Process (AHP), with a typical value range of [0,1], and preferred values ​​are as follows: , , , , Based on Its purpose is to compare "which candidate control strategy has better subsequent recovery capability after execution." In this scenario, the remaining regulation capability of energy storage is the most critical because it directly determines the subsequent rapid compensation capability. The most critical factor is the ability to continuously supply power to critical loads; the next most critical factor is the ability to continuously supply power to critical loads, because smart energy regulation cannot sacrifice critical loads. The second most important factor is the flexible load margin, voltage safety margin, and frequency safety margin. However, these are more of a supportive recovery condition, so they are given equal weight.

[0144] S4.5: Based on the comprehensive evaluation results, conduct stratified screening and ranking, and determine the target control strategy.

[0145] Specifically, based on the comprehensive evaluation results of candidate control strategies, each candidate control strategy is screened and ranked. When multiple candidate control strategies have similar effects on eliminating the current risk, the candidate control strategy with the higher recovery capability evaluation value is selected first. If a candidate control strategy has a good immediate recovery effect, but its recovery capability evaluation value is low, manifested as insufficient remaining energy storage regulation capacity, insufficient remaining capacity of flexible loads, small voltage safety margin of key nodes, low frequency safety margin, or decreased continuous power supply capacity of key loads, its ranking priority is reduced. Finally, the candidate control strategy that meets the current operational risk elimination requirements and has the highest comprehensive ranking score is selected as the target control strategy.

[0146] Filtering based on current risk elimination constraints: First, eliminate all candidate control strategies that cannot meet the requirements for eliminating current operational risks. The requirements for eliminating current operational risks include at least restoring the bus frequency to the preset safe operating frequency range, restoring the critical node voltage to the preset safe operating voltage range, ensuring that the critical branch load rate does not exceed the preset safe current carrying capacity limit, maintaining power supply to critical loads, and returning the grid-connected switching power to the preset allowable interactive power range. If the execution of a candidate control strategy still leads to bus frequency exceeding limits, critical node voltage exceeding limits, critical branch overload, or critical load power failure, then the candidate control strategy is determined to be an infeasible strategy.

[0147] The sorting is based on recovery capability. For feasible candidate control strategies that pass the first-level screening, the existing weighted scoring method is used to calculate the comprehensive ranking score. The calculation formula is as follows:

[0148] In the formula, S is the comprehensive ranking score of candidate control strategies. It is the frequency recovery evaluation value. It is the voltage recovery evaluation value. It is the branch load rate control evaluation value. It is the operating cost evaluation value. , , , , It is the weight coefficient of the corresponding evaluation item.

[0149] It should be noted that, The weight setting method is based on the Analytic Hierarchy Process (AHP), with a typical value range of [0,1], and preferred values ​​are as follows: , , , , Based on this step, the priority is to eliminate the current risk first, and then preserve the subsequent recovery capability. Therefore, the evaluation values ​​for frequency recovery and recovery capability are... It should be ranked the highest because frequency issues are most sensitive in microgrids, and Representing subsequent resilience, voltage recovery, and branch load rate control are important constraints for network security, and therefore are given medium to high weights. Although operating costs should be considered, they should not outweigh safety and resilience, so they are given the lowest weight.

[0150] The preset frequency safe operating range, preset voltage safe operating range, preset current safe upper limit, and preset interactive power allowable range are dynamically set or initialized based on the current operating mode of the microgrid (such as grid-connected mode or islanded mode), the national or industry standards followed, and user-defined safety policies.

[0151] S4.6: Map device parameters based on target control strategies and generate a set of executable control instructions for the device.

[0152] Specifically, the adjustment objects, adjustment directions, adjustment amplitudes, and execution timings in the target control strategy set are mapped to parameter fields that each controller can recognize. Among them, inverter-type equipment is mapped to active / reactive power setpoints, energy storage converters are mapped to charging and discharging power setpoints, load control units are mapped to switching or load reduction commands, and switch execution units are mapped to opening or closing control commands, forming a set of control commands that the equipment can execute.

[0153] The set of control commands that the device can execute includes at least the following: distributed power active power setpoint command, distributed power reactive power setpoint command, energy storage charging power command, energy storage discharging power command, adjustable load switching command, adjustable load unloading command, tie switch opening and closing command, and grid connection interface switching power adjustment command, etc.

[0154] S4.7: Based on the device's executable control instruction set, reliable transmission and end-to-end distribution are completed, and the execution unit is driven to form the actual control response result.

[0155] Specifically, based on the set of control commands that the equipment can execute, various control commands are sent to the corresponding distributed power controllers, energy storage converter controllers, load controllers, and switch execution units through a wired / wireless hybrid communication network to form end-to-end control commands. The corresponding execution units then parse and execute the end-to-end control commands, causing corresponding changes in the actual power balance state, bus voltage state, bus frequency state, branch power flow state, grid-connected switching power state, and local power supply path state of the microgrid, and forming actual control response results that reflect the control effect.

[0156] S5. Collect the control results after execution and compare them with the expected effect of the strategy to correct the operating state model and form a closed-loop control.

[0157] S5.1: Based on the actual control response results, the control execution results are generated, and the actual operating status after execution is extracted to form the basis for comparison of execution deviations.

[0158] Specifically, based on the actual control response results, a new round of node voltage, current, power, frequency, energy storage state of charge, branch load rate, grid-connected switching power, key load power supply status, and switch status are collected after control execution. The post-execution operating status is summarized to form the control execution results. The actual operating quantities after execution, such as node voltage, current, power, frequency, energy storage state of charge, branch load rate, grid-connected switching power, key load power supply status, and switch status, are extracted from the control execution results. At the same time, the expected operating quantities for the same time period are extracted from the digital twin simulation results corresponding to the target control strategy set. A corresponding mapping relationship is established according to nodes, branches, time windows, and control objects, so that the control execution results correspond to the simulation expected effects under the same comparison benchmark, providing a basis for subsequent execution deviation calculation.

[0159] S5.2: Calculate the execution deviation results based on the execution deviation comparison, and identify the sources of deviation to form the model correction parameter results.

[0160] Specifically, based on the comparison of execution deviations, the difference between the execution results of regulation and the expected effect of digital twin simulation is calculated to form the execution deviation results. The execution deviation results are then quantitatively characterized according to the amplitude, duration, or cumulative deviation of various deviations to determine whether the current regulation execution effect is within the allowable range. The main sources of execution deviations are identified, and model correction parameters are formed according to the deviation type.

[0161] The execution deviation results include at least the actual power response deviation, actual voltage recovery deviation, actual frequency recovery deviation, actual branch load rate deviation, actual grid-connected switching power deviation, and actual response time deviation.

[0162] The actual power response deviation can be obtained from the difference between the actual power value and the expected power value after execution; the actual voltage recovery deviation can be obtained from the difference between the actual voltage recovery time and the expected voltage recovery time after execution; the actual frequency recovery deviation can be obtained from the difference between the actual frequency deviation and the expected frequency deviation after execution; and the actual response time deviation can be obtained from the difference between the actual duration from the moment the control command is issued to the moment the target state is reached and the simulation expected duration.

[0163] It should be noted that the model correction parameter results include at least the energy storage response time correction value, load response parameter correction value, equipment call constraint correction value, disturbance source determination parameter correction value, line boundary parameter correction value, and prediction model parameter correction value.

[0164] S5.3: Update the model based on the model correction parameter results to form a closed-loop correction result.

[0165] Specifically, based on the model correction parameter results, the energy storage response time correction value is written into the dynamic response parameter of the corresponding energy storage node in the unified operation state model; the load response parameter correction value is written into the adjustable capacity parameter of the corresponding load node; the equipment call constraint correction value is written into the restricted call boundary of the corresponding equipment; the line boundary parameter correction value is written into the current carrying capacity boundary of the corresponding branch; the disturbance source determination parameter correction value and the prediction model parameter correction value are written into the disturbance attribution and ultra-short-term prediction, respectively; and the risk information, dominant disturbance source type, target control strategy set, control execution result, execution deviation result and model correction parameter result of this round of regulation are stored in the historical database, and the prediction model parameters or comprehensive evaluation weight parameters are periodically updated based on the accumulated historical data to improve the adaptability of subsequent regulation processes to continuous disturbance scenarios.

[0166] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart energy regulation method based on distributed energy microgrids, characterized in that, include: Collect multi-source heterogeneous state data and preprocess the multi-source heterogeneous state data to form a standardized microgrid real-time operation dataset; Based on the standardized microgrid real-time operation dataset, an operation status model integrating dynamic topology and equipment constraints is constructed by combining static electrical parameters. Based on the operation status model, a digital twin operation model corresponding to the physical microgrid is constructed synchronously. Based on the digital twin operation model, the future operation of the microgrid is simulated in rolling manner using ultra-short-term forecast data. Based on the rolling simulation results, the dominant disturbance source is identified by multi-dimensional risk quantification and confidence level classification, and physical sensitivity and attention models are integrated to generate hierarchical control demand results. Based on the results of the hierarchical regulation demand, a set of candidate regulation strategies is constructed and input into the digital twin operation model for parallel simulation. Based on the simulation results of the candidate strategies and the recovery margin evaluation, multi-objective decision-making is carried out to determine the target regulation strategy set and execute it. Collect the control results after execution, compare them with the expected effects of the strategy, and revise the operational status model.

2. The smart energy regulation method based on distributed energy microgrid as described in claim 1, characterized in that... The collection of multi-source heterogeneous state data refers to deploying sensing terminals to collect multi-source raw heterogeneous state data and configuring wireless acquisition and auxiliary power supply units for restricted edge nodes. Multi-source raw heterogeneous state data includes electrical quantities such as node voltage, node current, node power, bus frequency, energy storage state of charge, switch status, and grid-connected switching power, as well as equipment state quantities such as temperature and vibration.

3. The smart energy regulation method based on distributed energy microgrid as described in claim 2, characterized in that... The process of preprocessing multi-source heterogeneous state data to form a standardized real-time running dataset refers to adding timestamps and quality labels to the original multi-source heterogeneous state data and performing edge preprocessing to generate a standardized microgrid real-time running dataset.

4. The smart energy regulation method based on distributed energy microgrid as described in claim 3, characterized in that... The aforementioned construction of an operational state model that integrates dynamic topology and equipment constraints based on a standardized microgrid real-time operation dataset and static electrical parameters refers to the construction of a basic microgrid network model based on a standardized microgrid real-time operation dataset and static electrical parameters. A dynamic topology structure with node-branch associations is generated based on the microgrid basic network model, and the dynamic topology structure is loaded with the operating status of each node and the device constraint attributes to form a unified operating status model of the microgrid.

5. The smart energy regulation method based on distributed energy microgrids as described in claim 4, characterized in that... The aforementioned construction of a digital twin operation model corresponding to the physical microgrid based on the operation status model refers to constructing a digital twin operation model corresponding to the physical microgrid based on the unified operation status model of the microgrid, using equipment status variables and load attributes as supplementary constraints for the model.

6. The smart energy regulation method based on distributed energy microgrid as described in claim 5, characterized in that... The aforementioned simulation of the future operation of a microgrid based on a digital twin operation model, combined with ultra-short-term forecast data, refers to the construction of a short-term rolling simulation scenario based on a digital twin operation model through real-time mapping of physical states and embedding of health constraints. Based on short-term rolling simulation scenarios, the operation process of microgrids in the future short time domain is rolled out using ultra-short-term prediction data to obtain the future operating status of the microgrid in rolling simulation.

7. The smart energy regulation method based on distributed energy microgrid as described in claim 6, characterized in that... Based on the rolling simulation results, the dominant disturbance source is identified by multidimensional risk quantification and confidence level, and physical sensitivity and attention model are integrated. The hierarchical control demand results are generated. Based on the rolling simulation microgrid future operation status results, the power voltage branch and topology risk index set are extracted, and the hierarchical frequency risk quantification results are extracted. A comprehensive severity is formed based on the risk indicator set and the stratified frequency risk quantification results, the operational risk results are quantified, and the confidence level of various risks is classified. Construct a time-series vector of real-time operational change characteristics that integrates risk boundary information based on the results of quantitative operational risk analysis; Dominant disturbance sources are identified through physical sensitivity pre-screening and lightweight classification model fine classification, and restricted nodes are marked; Based on the risk confidence level and the identification results of the dominant disturbance source, hierarchical microgrid control requirements are generated.

8. The smart energy regulation method based on distributed energy microgrid as described in claim 7, characterized in that... The step of constructing a candidate control strategy set based on the hierarchical control demand results and inputting it into the digital twin operation model for parallel simulation refers to constructing a candidate control strategy set based on the hierarchical microgrid control demand results and performing digital twin parallel simulation based on the candidate control strategy set to form the strategy effect response results.

9. The smart energy regulation method based on distributed energy microgrid as described in claim 8, characterized in that... The process involves multi-objective decision-making based on candidate strategy simulation results and recoverability margin evaluation, determining the target control strategy set and executing the strategy effect response results, extracting recovery capability evaluation parameters, and forming a parameter set. The recovery capacity evaluation value is calculated based on the recovery capacity evaluation parameter set, and a comprehensive evaluation result is formed; Based on the comprehensive evaluation results, a tiered screening and ranking process is conducted, and target control strategies are determined. Based on the target control strategy, the device parameters are mapped to generate a set of executable control instructions, which are then reliably transmitted and sent to the end, driving the execution unit to form the actual control response result.

10. The smart energy regulation method based on distributed energy microgrid as described in claim 9, characterized in that... The process involves collecting and executing the control results, comparing them with the expected effects of the strategy, and correcting the operating state model. This process involves forming control execution results based on the actual control response results, extracting the actual operating state after execution to form the basis for execution deviation comparison, calculating the execution deviation results, identifying the sources of deviation to form model correction parameter results, and updating the model based on the model correction parameter results to form a closed-loop correction result.

Images