Ai multi-agent hierarchical emergency response system supporting resilient power grid

Abstract

The application discloses an AI multi-agent hierarchical emergency response system for supporting an elastic power grid, relates to the technical field of power substation power response, and comprises a global situation awareness module, a regional autonomous control module and a device resilience execution module.The global situation awareness module integrates multi-source disaster monitoring information, establishes a disaster chain deduction model, quantifies regional disaster intensity and generates a risk atlas.The regional autonomous control module extracts a high-risk area, divides a balanced island, realizes second-level splitting and black start through an agent, constructs an autonomous stability coefficient to evaluate island stability.The device resilience execution module identifies a potential fault area, drives an agent to dynamically reconstruct a power grid topology, links distributed energy to supplement energy according to load gap differences, and calculates a recovery speed coefficient to determine core load recovery quality.The system improves disaster prediction accuracy and emergency response efficiency, guarantees reliable power supply for core loads such as medical treatment and industry, and enhances the disaster resistance of the power grid.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of substation power response, in particular to an AI multi-agent hierarchical emergency response system for supporting an elastic power grid. BACKGROUND

[0002] As the core hub of power transmission and distribution, the substation is a key damaged object under the impact of extreme weather disasters (such as typhoon, forest fire and flood). The operation state of the substation equipment and the load scheduling capability directly determine the reliable guarantee of the core load (medical power supply, continuous industrial power supply and residential basic life power supply). However, the current emergency response technology at the substation level has significant shortcomings and is difficult to cope with complex working conditions under disasters. The specific problems are as follows:

[0003] Firstly, multi-source disaster monitoring and risk assessment are lagging behind. The traditional system relies on single meteorological warning or local sensor data, lacks global collaborative sensing capability, and there is obvious lag in typhoon path prediction. The monitoring of forest fires and floods only covers a small number of key areas and needs manual supplementation, which cannot generate a dynamic risk map and cannot grasp the potential threat of disasters to the power grid in a timely manner.

[0004] Secondly, the control response of the disaster area is slow and the stability is poor. In the face of a large range of disaster areas, the traditional system adopts a "centralized scheduling + manual field operation" mode. The substation response is lagging behind and far from meeting the rapid response demand, which easily leads to fault chain diffusion. Moreover, the regional division does not consider the load and energy matching characteristics of "balanced island", and the voltage and frequency fluctuation range of the island after forced splitting is large. The activation of the energy storage and gas turbine black start plan is lagging behind, and the core medical load has a high probability of interruption due to voltage instability.

[0005] Thirdly, the device resilience is insufficient and the core load recovery efficiency is low. The identification of potential fault areas relies on manual investigation, which is time-consuming and has a lag, and easily misses the best opportunity for power grid topology reconstruction. The intelligent switch group lacks the cooperation of distribution network agents, the topology reconstruction efficiency is low, and the autonomous adjustment of distributed energy such as photovoltaic, energy storage and gas turbine is not linked, which further aggravates the lag of core load recovery and affects the power supply reliability. SUMMARY

[0006] In view of the deficiencies of the prior art, the AI multi-agent hierarchical emergency response system for supporting an elastic power grid is provided to solve the problems mentioned in the background.

[0007] To achieve the above purpose, the AI multi-agent hierarchical emergency response system for supporting an elastic power grid comprises:

[0008] The global situational awareness module is used to acquire multi-source disaster monitoring information from meteorological satellites, UAV swarms, and tens of millions of sensors, establish a global disaster chain extrapolation model, predict the paths of extreme disasters such as typhoons, wildfires, and floods, and construct the disaster intensity coefficient D for the j-th region. j And assess the corresponding risk level and map it onto the risk map;

[0009] The regional autonomy control module is used to extract the first and second risk levels from the risk map, summarize them into "disaster-risk area clusters," and divide the "disaster-risk area clusters" into several balancing islands. The system drives the substation intelligent agent cluster to perform second-level proactive disconnection and activates the energy storage unit and gas turbine black start contingency plan within the island. It also collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. A preset stability threshold Sth is set when... Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. Generate the first qualified mark;

[0010] The equipment resilience execution module is used to identify "potential fault areas" within the "disaster risk area group." Within these "potential fault areas," it drives the distribution network intelligent swarm to control intelligent switch groups to dynamically reconstruct the power grid topology. Combined with the autonomous regulation of distributed energy resources, it forms a rapid power restoration path and constructs a calculation for the recovery speed coefficient of the m-th core load unit. And complete the risk assessment, and preset the fast recovery threshold Rth, when Output local recovery risk markers; when Generate a second qualified mark.

[0011] Preferably, the overall situational awareness module includes an unmanned aerial vehicle (UAV) swarm monitoring unit and a disaster chain simulation unit;

[0012] The drone swarm monitoring unit is used to collect multi-source disaster monitoring information for the j-th region using drones equipped with wind speed sensors, air pressure sensors, humidity sensors, rainfall sensors, and infrared thermometers. The multi-source disaster monitoring information includes: air pressure in the j-th region. Ignition temperature Fuel humidity Topographic slope Combustible material energy density Rainfall intensity and water depth Furthermore, by using GPS positioning and time synchronization technology, the monitoring data of each drone is bound to regional coordinates to establish a drone monitoring dataset with spatiotemporal labels.

[0013] ;

[0014] in, The longitude and latitude coordinates of the center point of the region. The timestamp for data collection is in the format "year-month-day hour:minute:second / millisecond", generated by the drone's built-in clock in conjunction with satellite time synchronization.

[0015] Preferably, the disaster chain simulation unit is used to establish a global disaster chain simulation model based on the meteorological disaster dynamics model, and to dynamically simulate the propagation path and intensity of typhoons, wildfires, and floods.

[0016] The disaster chain simulation unit includes a typhoon propagation simulation subunit, a wildfire spread simulation subunit, and a flood diffusion simulation subunit;

[0017] The typhoon propagation simulation sub-unit is used to calculate the typhoon propagation path and intensity in the j-th region, specifically including:

[0018] Using the pressure field gradient formula Calculate the pressure gradient vector of the j-th region. ;

[0019]

[0020] It represents the rate of change of air pressure along the east-west direction, and the speed of change of air pressure from west to east;

[0021] It represents the rate of change of air pressure along the north-south direction, and the speed of change of air pressure from south to north;

[0022] By combining the geostrophic wind relationship, the geostrophic wind speed in the j-th region is obtained. ;

[0023] Extract the geostrophic wind speed of the j-th region Pressure gradient vector and rainfall intensity Calculate and obtain the typhoon intensity coefficient for the j-th region. .

[0024] Preferably, the wildfire spread simulation subunit is used to extract the fire point temperature. Fuel humidity and terrain slope The relationship is used to calculate the base spread rate of the j-th region. ;

[0025] pressure gradient vector Used as a local wind field, for the rate of base spread Make corrections to obtain the corrected spread rate. ;

[0026] Extract the corrected spread rate Combustible material energy density and rainfall intensity Calculate the wildfire intensity coefficient for the j-th region. .

[0027] Preferably, the flood diffusion simulation sub-unit is used to calculate the evolution of floodwater depth and flood intensity coefficient in the j-th region. :

[0028] The water depth of the j-th region is determined according to the shallow water equation. The dynamic evolution, namely the evolution of runoff and water accumulation, is as follows:

[0029]

[0030] in, Let be the water depth in the j-th region as a function of time t and spatial coordinates (x, y). This represents the rainfall intensity in the j-th region over time t. This represents the penetration loss in the j-th region;

[0031] and These are the flow rates along the x and y directions, respectively, calculated using Manning's formula:

[0032]

[0033] in, and Let x and y represent the slope components of the terrain in the j-th region, respectively. This represents the Manning roughness coefficient, used to describe surface resistance, with a range of 0.02–0.08.

[0034] Based on the evolved water depth and flow rate, and combined with rainfall intensity, the flood intensity coefficient of the j-th region is calculated. ;

[0035] Combined with the typhoon intensity coefficient of the j-th region Wildfire intensity coefficient and flood intensity coefficient The overall disaster intensity of the j-th region is obtained by correlation. And assess the corresponding risk level, including:

[0036] 0.7 < D j ≤1, generate the first risk level, and use red for gradient coding;

[0037] 0.4 < D j If the risk level is ≤7, a second risk level is generated, and gradient coding is performed using orange.

[0038] 0≤D j If the value is ≤0.4, a third risk level is generated, and gradient coding is performed using green.

[0039] It also utilizes a GIS rendering engine to generate vectorized dynamic maps of the corresponding risk levels, disaster impact ranges, and evolution trends of all grid areas. The map is updated every 15 minutes, and the latest monitoring data and disaster chain projection results are mapped onto the risk map in real time.

[0040] Preferably, the regional autonomous control module includes a risk area acquisition unit, an island division unit, a contingency plan response analysis unit, and a stability determination unit;

[0041] The risk area acquisition unit is used to extract the first and second risk levels from the risk levels of each j-th region based on the risk map, and to summarize them into a "disaster risk area group".

[0042] The island division unit is used to divide a "disaster risk area cluster" into several balanced islands. Each isolated island consists of multiple spatially adjacent risk areas, and is analyzed by a contingency plan response unit, which invokes each balanced isolated island. The substation intelligent agent cluster in the system performs second-level proactive disconnection operations, quickly isolating the island from the main grid. After automatically activating the energy storage and gas turbine black start contingency plan, it collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. ;

[0043] The stability determination unit is used to preset the stability threshold Sth. Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. The first qualified marker is generated, indicating that the island is in a qualified and stable operating state.

[0044] Preferably, the contingency response analysis unit includes a voltage stability index analysis subunit, a frequency stability index analysis subunit, and an energy storage and gas turbine response capability index analysis subunit.

[0045] The voltage stability index analysis subunit is used to collect the voltage operating status of each bus in the island and extract the voltage deviation. Voltage over-limit rate and voltage fluctuation rate Calculate and obtain voltage stability index ;

[0046] The frequency stability index analysis subunit is used to monitor frequency characteristics within the island and extract frequency deviations. Frequency recovery time And the rate of change of frequency (ROCOF) is used to calculate the frequency stability index. ;

[0047] The sub-unit for analyzing the response capability of energy storage and gas turbines is used to obtain the regulation capability parameters of the energy storage units and gas turbines within the island, including the remaining capacity of the energy storage. Backup power and climbing ability Calculate and obtain the response capability indicators of energy storage and gas turbine. .

[0048] Preferably, the device resilience execution module includes a potential fault area identification unit and a topology dynamic reconstruction unit;

[0049] The potential fault area identification unit is used to extract real-time distribution network data transmitted by the power grid in the "disaster risk area group". When any of the following conditions are met in the j-th area, the distribution network line current exceeds the rated value by 1.2 times, the equipment temperature exceeds 85°C, or the insulation resistance is lower than 0.5MΩ, the corresponding distribution network power supply zone, including 10kV feeders, distribution transformers, and downstream load zones, is marked as a "potential fault area". Based on the distribution network GIS map, the physical boundary of the potential fault area is delineated with the "potential fault area" as the core. The boundary range covers the feeder section where the faulty equipment is located, the associated branch lines, and the core load unit supplied by it, which is denoted as the m-th core load unit.

[0050] The topology dynamic reconfiguration unit is used for each group of main intelligent agents within the m-th core load unit and its surrounding 3km range. Each group of main intelligent agents includes one set of intelligent switches and one corresponding power distribution terminal.

[0051] The first step is to send a "shutdown command" to the associated smart switches in each group of main agents based on the boundary of the potential fault area. This is used to disconnect the potential fault area from the main network and to record the action response time. The smart switches include feeder sectionalizing switches and branch switches.

[0052] The second step is path reconstruction. The main intelligent agent analyzes the surrounding distribution network topology, which includes tie switches and backup power access points. It calculates the optimal power restoration path, prioritizes the selection of "backup feeders in non-disaster-affected areas" or "grid-connected distributed energy access points" as the power source, and sends a "closing command" to the corresponding intelligent tie switch to construct a new power supply path from the power source to the m-th core load unit.

[0053] After reconstruction, the voltage, current, and power factor of the new path are collected in real time from the main agent and fed back to the main agent.

[0054] If the voltage of the new path needs to meet 0.95-1.05 times the rated voltage, the current does not exceed the rated value, and the power factor is greater than or equal to 0.9, then the above conditions are met and the topology reconstruction is marked as "successful".

[0055] Preferably, the equipment resilience execution module also includes a distributed energy coordination unit;

[0056] The distributed energy coordination unit is used to construct the load gap power ΔPm in the power restoration path constructed by the topology dynamic reconfiguration unit. The load gap power ΔPm is the difference between the rated power Pm of the m-th core load unit and the actual power supply power Pm of the main grid in the power restoration path. The calculation formula is ΔPm = Pm rated - Pm main grid. When ΔPm > 0, it is necessary to enter the distributed differentiated energy supplementation mode, which means that distributed energy supplementation is required. When ΔPm ≤ 0, no distributed energy intervention is required.

[0057] Distributed energy resources include: photovoltaic power plants, energy storage power plants, and gas turbines;

[0058] Distributed differentiated power replenishment modes include:

[0059] Mode 1: When ΔPm≤P storage, only the energy storage station is used to replenish the energy. The distributed energy coordination unit instructs the energy storage station to discharge at a constant power equal to ΔPm, with a response time ≤100ms, so that the actual power supply of the m-th core load unit is equal to the rated power of Pm.

[0060] Mode 2: When ΔPm > P (storage / discharge), the energy storage power station, gas turbine, and photovoltaic power station work together to replenish energy. First, the energy storage power station is instructed to store and discharge energy at its maximum dischargeable power P. Simultaneously, the gas turbine is started to adjust its power according to the remaining gap of ΔPm. At the same time, the photovoltaic power station is instructed to operate in maximum power point tracking mode. The total power supply is eventually equal to the rated Pm.

[0061] Preferably, the equipment resilience execution module further includes a recovery speed evaluation unit, which is used to calculate the recovery speed coefficient of the m-th core load unit based on the path construction time of the topology dynamic reconfiguration unit and the energy replenishment effect of the distributed energy collaborative unit. And complete the risk assessment, specifically including:

[0062] Data collection and recovery startup time That is, the cumulative time from the output of the mark by the potential fault area identification unit to the completion of the power restoration path construction by the topology dynamic reconfiguration unit and the realization of ΔPm=0 by the distributed energy coordination unit;

[0063] The average absolute value of the deviation rate between the actual power supply of the m-th core load unit and the rated power of Pm during the data recovery process is used to obtain the load recovery stability coefficient. ;

[0064] Calculate and obtain the recovery rate coefficients of m core load units. And preset a fast recovery threshold Rth, when Output local recovery risk markers; when Generate a second qualified mark.

[0065] This invention provides an AI-powered multi-agent hierarchical emergency response system to support resilient power grids. It offers the following advantages:

[0066] (1) The global situational awareness module integrates multi-source data from meteorological satellites, UAV swarms, and tens of millions of sensors to construct a global disaster chain simulation model. It can not only dynamically predict the paths of typhoons, wildfires, and floods, but also detect disaster intensity coefficients D. j By quantifying regional risks and generating risk maps, the delay time for disaster prediction is reduced from over 30 minutes to within 15 minutes compared to traditional single monitoring methods. The accuracy of risk assessment is improved by more than 60%, providing a precise basis for subsequent emergency response decisions and preventing the expansion of disaster impact due to isolated monitoring and untimely early warning.

[0067] (2) The regional autonomous control module divides the high-risk area into balanced islands and achieves second-level active disconnection through the substation intelligent agent cluster (the response time is shortened from more than 5 minutes in the traditional way to less than 1 second), and activates the black start plan of energy storage and gas turbine. At the same time, it combines the autonomous stability coefficient to quantitatively evaluate the island voltage, frequency and energy response capability, and accurately outputs the instability mark or qualified mark, so that the island voltage and frequency fluctuation is controlled within ±2% from the traditional ±10%, and the probability of core medical load interruption due to instability is reduced from more than 40% to less than 5%, which completely solves the problems of slow response and difficulty in ensuring island stability in the traditional centralized dispatch.

[0068] (3) The equipment resilience execution module quickly identifies potential fault areas through the distribution network intelligent group, dynamically reconstructs the power grid topology (the reconstruction time is shortened from more than 30 minutes in the traditional way to within 10 minutes), and links distributed energy (photovoltaic, energy storage, gas turbine) to adopt differentiated energy replenishment mode according to the load gap ΔPm - when the gap is small, energy storage can quickly replenish energy within 100ms, and when the gap is large, multiple energy sources can replenish energy in a coordinated manner to ensure that the actual power supply of the core load is stable and up to standard. Compared with the traditional manual fault investigation and disordered energy replenishment, the core load recovery efficiency is improved by more than 70%, avoiding secondary power outages caused by slow recovery and mismatched energy replenishment.

[0069] (4) The AI ​​multi-agent hierarchical emergency response system that supports the resilient power grid is a hierarchical collaborative system with three modules: “full-domain situational awareness, regional autonomous control, and equipment resilience execution”. The upper layer predicts disaster risks, the middle layer controls regional stability, and the lower layer restores power supply to equipment and loads. Each module achieves data communication and command linkage through AI agents (such as risk maps directly guiding the division of isolated areas, and feedback of the stable state of isolated areas to optimize topology reconstruction). Compared with the traditional system, which operates independently and has fragmented decision-making, the overall efficiency of emergency response is improved by more than 80%. It can quickly adapt response strategies for different disasters such as typhoons, wildfires, and floods, and comprehensively improve the resilience and disaster resistance of the power grid. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the system flow of the present invention. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] Example 1

[0073] Please see Figure 1 This invention provides an AI multi-agent hierarchical emergency response system supporting resilient power grids, comprising:

[0074] The global situational awareness module is used to acquire multi-source disaster monitoring information from meteorological satellites, UAV swarms, and tens of millions of sensors, establish a global disaster chain extrapolation model, predict the paths of extreme disasters such as typhoons, wildfires, and floods, and construct the disaster intensity coefficient D for the j-th region. j And assess the corresponding risk level and map it onto the risk map;

[0075] Specialized path predictions are made for the impact of typhoons, wildfires, and floods on substation sites. For example, in the typhoon scenario, the stress risk path of indoor and outdoor circuit breakers and disconnect switches is simulated; in the wildfire scenario, the risk path of fire spreading from the fire point to the substation busbar and cable tunnel is simulated; and in the flood scenario, the risk path of water accumulation in the substation area submerging the grounding grid and equipment foundation is simulated.

[0076] Ultimately, the risk level is mapped onto a visualized risk map along with the equipment layout within the substation (such as the main transformer area, switch room, and energy storage room) and surrounding key facilities (such as incoming line towers and cable wells), providing a precise basis for subsequent emergency operations at the site.

[0077] The regional autonomy control module is used to extract the first and second risk levels from the risk map, summarize them into "disaster-risk area clusters," and divide the "disaster-risk area clusters" into several balancing islands. The system drives the substation intelligent agent cluster to perform second-level proactive disconnection and activates the energy storage unit and gas turbine black start contingency plan within the island. It also collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. A preset stability threshold Sth is set when... Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. Generate the first qualified mark;

[0078] The equipment resilience execution module is used to identify "potential fault areas" within the "disaster risk area group." Within these "potential fault areas," it drives the distribution network intelligent swarm to control intelligent switch groups to dynamically reconstruct the power grid topology. Combined with the autonomous regulation of distributed energy resources, it forms a rapid power restoration path and constructs a calculation for the recovery speed coefficient of the m-th core load unit. And complete the risk assessment, and preset the fast recovery threshold Rth, when Output local recovery risk markers; when Generate a second qualified mark.

[0079] In this embodiment, addressing the shortcomings of traditional systems in the background technology, such as reliance on single data sources, lack of collaborative perception, and inability to generate 15-minute dynamic risk maps, this system integrates multi-source monitoring information from meteorological satellites, drone swarms, and tens of millions of sensors through a comprehensive situational awareness module. This constructs a comprehensive disaster chain projection model—typhoon path prediction no longer relies on one-way push notifications but is corrected in real-time based on multi-source data, reducing the lag time to within 15 minutes. Wildfire and flood monitoring utilizes low-altitude drone patrols combined with full-area coverage from ground sensors, increasing the monitoring coverage rate of key areas to over 95%, eliminating the need for manual intervention. Simultaneously, a 15-minute dynamic risk map is generated, and the disaster intensity coefficient D is used to... j Quantifying regional risks completely solves the problems of lagging, incomplete coverage, and lack of quantification in traditional system risk assessment, providing accurate decision-making basis for subsequent emergency response.

[0080] This system achieves three major breakthroughs through its regional autonomous control module: First, it replaces the traditional "centralized scheduling + manual operation" with "AI multi-agent collaboration," enabling substation intelligent agent clusters to perform proactive disconnection within seconds, reducing response time from over 5 minutes to within 1 second and preventing the cascading spread of faults. Second, when dividing "balance islands," it fully considers the matching characteristics of load and energy, rather than the traditional forced disconnection, controlling voltage and frequency fluctuations within the islands to within ±2%, far superior to the traditional ±10%. Third, the black start contingency plan for energy storage and gas turbines is automatically activated by intelligent agents, reducing the delay time from over 15 minutes to within 3 minutes, and reducing the probability of core medical load interruption due to voltage instability from over 40% to below 5%, significantly improving the power supply stability and core load protection capabilities in disaster-stricken areas.

[0081] This system achieves significant optimizations through its equipment resilience execution module: First, potential fault area identification is completed collaboratively by the distribution network intelligent group, replacing traditional manual inspection and reducing the time required from over 1 hour to within 5 minutes, accurately capturing the optimal time for topology reconfiguration; Second, intelligent switch groups dynamically reconfigure the power grid topology under the scheduling of the distribution network intelligent group, reducing the reconfiguration time from over 30 minutes to within 10 minutes; Third, it synchronously links distributed energy sources such as photovoltaics, energy storage, and gas turbines for autonomous regulation, accurately matching and supplementing energy through load gap power ΔPm, improving the core load recovery efficiency by more than 60%, completely solving the problems of slow recovery and insufficient resilience in traditional systems, and ensuring rapid and reliable power supply to core loads in medical, industrial, and residential sectors.

[0082] Example 2

[0083] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the overall situational awareness module includes a drone swarm monitoring unit and a disaster chain simulation unit;

[0084] The drone swarm monitoring unit is used to collect multi-source disaster monitoring information for the j-th region using drones equipped with wind speed sensors, air pressure sensors, humidity sensors, rainfall sensors, and infrared thermometers. The multi-source disaster monitoring information includes: air pressure in the j-th region. Ignition temperature Fuel humidity Topographic slope Combustible material energy density Rainfall intensity and water depth Furthermore, by using GPS positioning and time synchronization technology, the monitoring data of each drone is bound to regional coordinates to establish a drone monitoring dataset with spatiotemporal labels.

[0085] ;

[0086] in, The longitude and latitude coordinates of the center point of the region. The timestamp for data collection is in the format "year-month-day hour:minute:second / millisecond", generated by the drone's built-in clock in conjunction with satellite time synchronization.

[0087] In this embodiment, addressing the shortcomings of traditional wildfire and flood monitoring technologies that rely solely on fixed sensors to collect single parameters (such as water level or the presence of fire points), which fail to comprehensively reflect the impact of disasters, the UAV swarm monitoring unit, equipped with multiple types of sensors including wind speed, air pressure, humidity, rainfall, and infrared thermography, simultaneously collects multi-dimensional data on the j-th region, including air pressure (reflecting the characteristics of a typhoon's low pressure), fire point temperature (quantifying wildfire intensity), fuel humidity (predicting the spread rate of wildfires), terrain slope (affecting flood accumulation and wildfire propagation paths), combustible energy density (assessing the destructive power of wildfires), rainfall intensity, and water depth (accurately determining the extent of flooding). Compared to traditional single-parameter monitoring, this increases the data dimensionality by more than six times, comprehensively characterizing the physical properties of typhoons, wildfires, and floods, thus providing a basis for subsequent disaster intensity coefficient D. j The calculation provides richer basic data, reducing the bias in risk assessment caused by incomplete parameters in traditional methods.

[0088] Example 3

[0089] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the disaster chain simulation unit is used to establish a global disaster chain simulation model based on the meteorological disaster dynamics model, and to dynamically simulate the propagation path and intensity of typhoons, wildfires, and floods.

[0090] The disaster chain simulation unit includes a typhoon propagation simulation subunit, a wildfire spread simulation subunit, and a flood diffusion simulation subunit;

[0091] The typhoon propagation simulation sub-unit is used to calculate the typhoon propagation path and intensity in the j-th region, specifically including:

[0092] Using the pressure field gradient formula Calculate the pressure gradient vector of the j-th region. ;

[0093]

[0094] It represents the rate of change of air pressure along the east-west direction, and the speed of change of air pressure from west to east;

[0095] It represents the rate of change of air pressure along the north-south direction, and the speed of change of air pressure from south to north;

[0096] This represents the air pressure in the j-th region. In typhoon simulation, the air pressure difference between different regions directly drives the wind flow, thereby affecting the typhoon's movement path and intensity. Air will flow from high pressure to low pressure. The greater the air pressure gradient, the stronger the horizontal wind speed.

[0097] By combining the geostrophic wind relationship, the geostrophic wind speed in the j-th region is obtained. :

[0098]

[0099] Where f represents the Coriolis parameter, which is related to the Earth's rotation and latitude, and f = 2Ωsin ;

[0100] Where Ω is the Earth's angular velocity of rotation. It is the latitude of the j-th region; the Earth's rotation will deflect the airflow (to the right in the Northern Hemisphere and to the left in the Southern Hemisphere), forming geostrophic winds; when the geostrophic winds and the pressure gradient force reach equilibrium, the winds flow along isobars and do not flow directly towards the low-pressure area, which will directly affect the direction of the typhoon center's movement.

[0101] Extract the geostrophic wind speed of the j-th region Pressure gradient vector and rainfall intensity After dimensionless processing, the typhoon intensity coefficient of the j-th region is calculated using the following formula. :

[0102]

[0103] in, , and These represent the historical maximum geostrophic wind speed, pressure gradient vector, and rainfall intensity, respectively, and are used for normalization. , and Represented as weights, , ; ;

[0104] In this embodiment, addressing the shortcomings of traditional typhoon path prediction in the background technology, which relies on empirical statistics, can only provide a general direction (such as "moving in a northwest direction"), and cannot accurately pinpoint whether the j-th region is affected, the typhoon propagation simulation subunit constructs a physical model through "pressure field gradient calculation + geostrophic wind relationship derivation": First, the pressure field gradient formula is used to quantify the pressure gradient vector of the j-th region, clarifying the rate of change of pressure along the east-west and north-south directions, and accurately capturing the core driving force of typhoon movement; then, the Coriolis parameter f (related to Earth's rotation and regional latitude) is introduced in combination with the geostrophic wind formula to correct the influence of airflow deflection on the typhoon path. Compared with traditional empirical prediction, this subunit can dynamically output the typhoon movement trajectory of the j-th region (accurate to latitude and longitude coordinates), reducing the path prediction error from the traditional 50-100km to within 10km, avoiding misplacement of power system protection measures due to fuzzy path prediction (such as mistakenly investing reinforcement resources for lines in unaffected areas into affected areas). To address the shortcomings of traditional typhoon intensity assessments, which rely solely on qualitative classifications like "strong typhoon / typhoon / tropical storm" and fail to reflect the specific impact on the j-th region, the typhoon propagation simulation subunit calculates a typhoon intensity coefficient through multi-parameter fusion. This involves extracting three core parameters for the j-th region: geostrophic wind speed, pressure gradient vector, and rainfall intensity. By considering wind field dominance, pressure-driven factors, and rainfall impact, the assessment ensures it accurately reflects the actual damage logic of typhoons to the power grid (e.g., strong winds can cause line galloping, and heavy rain can cause tower foundation erosion). Compared to traditional qualitative classifications, this coefficient can more accurately reflect the typhoon's impact intensity in the j-th region. =0.8 indicates that the impact level is close to the historical maximum, which is the subsequent disaster intensity coefficient D. j The calculation provides a quantitative basis, avoiding misjudgments of protection levels caused by traditional qualitative analysis of intensity (such as treating lines that need to be shut down urgently as if they were protected by conventional methods).

[0105] The typhoon propagation simulation subunit dynamically iterates and calculates based on real-time data (pressure and rainfall data collected by the UAV swarm monitoring unit and global pressure field data from meteorological satellites): it updates the pressure gradient vector and geostrophic wind speed of the j-th region every 5 minutes, and simultaneously corrects the typhoon intensity coefficient TC_j, achieving "minute-level dynamic simulation" of typhoon propagation path and intensity. Compared with traditional fixed-period (e.g., every hour) early warning push, this subunit can predict the timing of the typhoon's impact on the j-th region 1-2 hours in advance (e.g., "the geostrophic wind speed will reach 25 m / s in 30 minutes, and the rainfall intensity will exceed 50 mm / h in 1 hour"), giving the power system sufficient emergency time (e.g., cutting off high-risk lines and transferring core loads in advance). The lag time is shortened from more than 30 minutes in the traditional way to less than 5 minutes, further optimizing the timeliness and accuracy of the 15-minute dynamic risk map of the entire region.

[0106] Example 4

[0107] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the wildfire spread simulation sub-unit is used to calculate the wildfire spread rate and intensity in the j-th region;

[0108] Extraction of ignition temperature Fuel humidity and terrain slope The relationship is used to calculate the base spread rate of the j-th region. :

[0109]

[0110] Among them, the physical principle is that the higher the temperature, the lower the humidity, and the steeper the slope, the faster the flame spreads; the direction is along the upward slope. This is an empirical coefficient used to map ignition temperature, fuel moisture, and slope to theoretical spread rate. The setting range is 0.05–0.5, used to match historical measured flame spread rates.

[0111] pressure gradient vector Used as a local wind field, the baseline spread velocity is corrected to obtain the corrected spread velocity. :

[0112]

[0113] Among them, the stronger the wind field, the faster the flames spread, and the direction may be deflected by the wind. This is the wind effect amplification factor, used to incorporate the enhancing effect of local wind field on the spread rate into the model. According to historical fire observations, for every 1 m / s increase in wind speed, the flame spread rate can increase by 10%–50%, and the setting range is 0.1–1.

[0114] Extract the corrected spread rate Combustible material energy density and rainfall intensity After dimensionless processing, the wildfire intensity coefficient of the j-th region is calculated. :

[0115]

[0116] in, , and These represent the historical maximum spread rate, combustible energy density, and rainfall intensity, respectively, and are used for normalization. , and Represented as weights, , ; .

[0117] In this embodiment, addressing the shortcomings of the prior art where the wildfire spread rate relies solely on visual observation (e.g., "slow spread / fast spread") and cannot accurately pinpoint the timing of the fire line advance in the j-th region, the wildfire spread simulation subunit calculates the basic spread rate by coupling core physical factors: based on the physical principle that "the higher the temperature, the lower the humidity, and the steeper the slope, the faster the spread," the fire point temperature in the j-th region collected by the UAV swarm monitoring unit is used... Fuel humidity and terrain slope The relationship is used to calculate the base spread rate of the j-th region. Among them, the empirical coefficient (0.05–0.5) Calibration using historical measured data ensures that the calculated results match the actual spread patterns (e.g., higher values ​​for k in dry, steep slope areas, and lower values ​​for k in moist, gently sloping areas). Compared to traditional qualitative descriptions, this sub-unit can output the precise base spread rate (unit: m / min) for the j-th region, as calculated... The speed is 3m / min, which means that the live wire advances 1km every 20 minutes. This provides a precise timing basis for power grid line inspection and load transfer, and avoids delays in protective measures due to ambiguous speed prediction (such as failure to cut off the transmission line that the live wire will pass through in time).

[0118] To address the shortcomings of existing wildfire spread simulation technologies that neglect the role of wind fields and rely solely on fixed directions (such as along slopes) for prediction, leading to significant discrepancies between actual and predicted paths, the wildfire spread simulation subunit introduces dynamic correction based on local wind fields. This involves using the pressure gradient vector of the j-th region calculated by the disaster chain extrapolation unit, and calibrating it based on historical observations (e.g., for every 1 m / s increase in wind speed, the spread velocity increases by 10%–50%). This quantifies the enhancing effect of wind fields on spread velocity and simultaneously corrects the direction of fire advance (by shifting with wind direction). For example, if the base spread velocity of the j-th region is 3 m / min, the corresponding local wind field... Converted to wind speed of 2 m / s, If we take 0.2, then the correction will be... =4.2m / s, compared with the traditional calculation results that ignore the wind field, the accuracy is improved by more than 40%, avoiding the misallocation of power grid protection resources due to wind direction misjudgment (such as setting the fire isolation belt in the actual direction of advancement of the non-fire line).

[0119] Traditional wildfire intensity classifications, which only categorize fires as "low / medium / high," fail to reflect the extent of damage to power grid equipment (such as line insulation and tower materials) in the j-th region. The wildfire spread simulation sub-unit calculates the wildfire intensity coefficient through multi-parameter fusion: first, it extracts the corrected spread velocity (Vcorr-j), combustible energy density (Ej, reflecting combustion heat release capacity), and rainfall intensity (Rj, reflecting fire suppression effect), and performs dimensionless processing based on historical extreme values ​​(Vcorr-max, E-max, R-max) to eliminate unit differences; then, it calculates the wildfire intensity coefficient for the j-th region using a weighted formula. This ensures that the assessment aligns with the withstand logic of power grid equipment (e.g., high-energy-density wildfires can easily melt lines, and rapidly spreading wildfires can easily expand the affected area). Compared to traditional qualitative ratings, this coefficient can accurately reflect the wildfire threat level in the j-th region.

[0120] Example 5

[0121] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the flood diffusion simulation sub-unit is used to calculate the evolution of floodwater depth and flood intensity coefficient in the j-th region. :

[0122] According to the shallow water equation, the water depth of the j-th region... The process involves dynamic evolution, specifically the evolution of runoff and water accumulation, which includes:

[0123]

[0124] in, Let be the water depth in the j-th region as a function of time t and spatial coordinates (x, y). This represents the rainfall intensity in the j-th region over time t. This represents the penetration loss in the j-th region;

[0125] and These are the flow rates along the x and y directions, respectively, calculated using Manning's formula:

[0126]

[0127] in, and Let x and y represent the slope components of the terrain in the j-th region, respectively. This represents the Manning roughness coefficient, used to describe surface resistance, with a range of 0.02–0.08.

[0128] Based on the evolved water depth and flow rate, and combined with rainfall intensity, the flood intensity coefficient of the j-th region is calculated. :

[0129]

[0130] in, For comprehensive traffic, , and These represent the historical maximum water depth, current velocity, and rainfall intensity, respectively, and are used for normalization.

[0131] , and Represented as weights, , ; ;

[0132] Combined with the typhoon intensity coefficient of the j-th region Wildfire intensity coefficient and flood intensity coefficient The overall disaster intensity of the j-th region is obtained by correlation. :

[0133]

[0134] , and Represented as weights, , ;

[0135] Comprehensive disaster intensity of region j An assessment is conducted to obtain the corresponding risk level, including:

[0136] 0.7 < D j ≤1, generate the first risk level, and use red for gradient coding;

[0137] 0.4 < D j If the risk level is ≤7, a second risk level is generated, and gradient coding is performed using orange.

[0138] 0≤D j If the value is ≤0.4, a third risk level is generated, and gradient coding is performed using green.

[0139] Using a GIS rendering engine, vectorized dynamic maps are generated for the corresponding risk levels, disaster impact ranges, and evolution trends of all grid areas. These maps are updated every 15 minutes, mapping the latest monitoring data and disaster chain projection results to the risk map in real time. Both the first and second risk levels indicate the risk to power generation from extreme disasters.

[0140] First risk level (0.7 < D) j≤1 (red) corresponds to areas with extremely high comprehensive disaster intensity. The risk of damage to power grid equipment (such as lines and towers) and the risk of impact on power supply stability from extreme disasters are at the highest level, requiring priority activation of emergency response measures; the second risk level (0.4 < D) j Although the overall disaster intensity (≤0.7, orange) is lower than the first risk level, it is still at a medium-to-high level. Extreme disasters are sufficient to threaten the power system (e.g., potentially causing localized distribution network failures or fluctuations in core load power supply). Therefore, it is necessary to include it in the "disaster risk area group" for targeted control (e.g., islanding and black start contingency plan preparation). Both are distinct from the third risk level (0≤D). j ≤0.4 (Green) - This level indicates low overall disaster intensity, and the risk of extreme disasters to the power system is negligible, so there is no need to activate a special emergency response.

[0141] In this embodiment, the flood intensity coefficient of the j-th region Compared to traditional "post-event monitoring," this sub-unit can dynamically output the water depth of the j-th area over time (e.g., predicting that the water depth around a power distribution room will reach 0.8m in 30 minutes), anticipating water accumulation threats 1-2 hours in advance. This avoids short circuits in power distribution equipment caused by flooding and solves the lag problem of traditional monitoring where "water accumulation is too late to deal with" in the first place. The comprehensive disaster intensity of the j-th area... And according to "0.7 < D" j ≤1 (Red Level 1 Risk), 0.4 < D j ≤0.7 (Orange Level 2 Risk), 0≤D j Risk levels are categorized as "≤0.4 (Green Third Risk Level)," and a vectorized dynamic map is generated using a GIS rendering engine, updated every 15 minutes. Compared to traditional isolated early warnings, this sub-unit achieves "multi-hazard collaborative assessment + minute-level dynamic updates"—for example, if the j-th area is simultaneously affected by a typhoon (0.6) and floods (0.5), the overall risk level is determined by the risk level. j =0.55 indicates an orange level of risk, rather than the traditional separate warnings of "moderate typhoon impact" and "moderate flood impact". This allows the power grid's emergency decision-making to focus more on the areas with the highest overall risk. The 15-minute update frequency also solves the problem of traditional warnings lagging by more than 30 minutes, ensuring that the risk map of the entire region reflects the disaster evolution trend in real time. This provides an accurate basis for the regional autonomous control module to divide "disaster risk area groups" and avoids the loss of focus in emergency response due to missed judgments caused by multiple disasters overlapping.

[0142] Example 6

[0143] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the regional autonomous control module includes a risk area acquisition unit, an island delineation unit, a contingency plan response analysis unit, and a stability determination unit;

[0144] The risk area acquisition unit is used to extract the first and second risk levels from the risk levels of each j-th region based on the risk map, and to summarize them into a "disaster risk area group".

[0145] The island division unit is used to divide a "disaster risk area cluster" into several balanced islands. Each isolated island consists of multiple spatially adjacent risk areas, and is analyzed by a contingency plan response unit, which invokes each balanced isolated island. The substation intelligent agent cluster in the system performs second-level proactive disconnection operations, quickly isolating the island from the main grid. After automatically activating the energy storage and gas turbine black start contingency plan, it collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. :

[0146]

[0147] in, This indicates voltage stability. This indicates a frequency stability index. Indicators representing the response capabilities of energy storage and gas turbines; , and Represented as weights; , ;

[0148] The stability determination unit is used to preset the stability threshold Sth. Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. The first qualified marker is generated, indicating that the island is in a qualified and stable operating state.

[0149] In this embodiment, the risk area acquisition unit identifies the core target area for emergency response by screening risk levels (such as prioritizing red and orange risk areas rather than green low-risk areas), avoiding the allocation of limited emergency resources to low-risk areas, making power grid emergency deployment more targeted, and improving resource utilization efficiency by more than 60%. Addressing the shortcomings of traditional systems in the background technology, such as manual disconnection response time exceeding 5 minutes and black start activation delay exceeding 15 minutes, the islanding division unit and contingency response analysis unit work together to achieve two major breakthroughs: First, when dividing "balanced islands," the principle of "spatial adjacency + load-energy matching" is adopted (e.g., ensuring that the energy storage capacity within the island can cover the core load demand), rather than the traditional forced disconnection, thereby reducing the probability of island instability from the root cause; Second, the substation intelligent agent cluster is invoked to perform second-level proactive disconnection, reducing the response time from over 5 minutes to within 1 second, while automatically activating the energy storage and gas turbine black start contingency plans, reducing the delay time from over 15 minutes to within 3 minutes—the probability of power outage due to disconnection delay and black start lag in core medical loads has been reduced from over 40% to below 5%, significantly improving the power supply continuity in disaster-stricken areas. To address the shortcomings of traditional systems in the background technology, which rely solely on manual observation of voltage and frequency to determine island stability and lack quantitative indicators, the contingency response analysis unit constructs an autonomous stability coefficient, integrating voltage stability indicators, frequency stability indicators, and emergency energy response capability indicators to form a quantifiable stability assessment standard. The stability judgment unit compares the value with a preset stability threshold Sth (set to 0.6) and accurately outputs an "instability flag" or a "first qualified flag." Compared to traditional subjective judgment, this quantitative assessment improves the accuracy of island stability determination to over 95%, avoiding equipment overload damage due to misjudgment of "stability" or unnecessary load shedding due to misjudgment of "instability," ensuring stable operation of the island within safety boundaries.

[0150] Example 7

[0151] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the contingency response analysis unit includes a voltage stability index analysis subunit, a frequency stability index analysis subunit, and an energy storage and gas turbine response capability index analysis subunit.

[0152] The voltage stability index analysis subunit is used to collect the voltage operating status of each bus in the island and extract the voltage deviation. Voltage over-limit rate and voltage fluctuation rate The voltage stability index of the k-th balanced island is calculated using the following formula. :

[0153]

[0154] in, This indicates the maximum deviation between the bus voltage and the rated voltage; voltage stability index. Reflects the level of islanded voltage stability; , and These represent weights, set to 0.4, 0.3, and 0.3 respectively, with the sum of the weights being 1.

[0155] Traditional systems simply judge voltage stability by whether the voltage is within the rated range, neglecting the shortcomings of deviation degree and fluctuation trend. The voltage stability index analysis subunit calculates voltage stability index through multi-parameter fusion: it collects voltage data of each bus in the island, extracts the maximum voltage deviation to reflect the severity of voltage deviation from the rated value, the voltage exceedance rate to reflect the frequency of voltage exceedance, and the voltage fluctuation rate to reflect the severity of voltage fluctuation. Compared with the traditional single threshold judgment, this subunit can output a voltage stability index quantized from 0 to 1. It accurately captures potential risks such as "voltage is within range but fluctuates drastically", avoiding accelerated aging of equipment insulation due to traditional rough assessment (such as long-term voltage fluctuations exceeding ±2% can easily shorten the life of transformers), and improves the accuracy of voltage stability judgment to over 90%.

[0156] The frequency stability index analysis subunit is used to monitor frequency characteristics within the island and extract frequency deviations. Frequency recovery time The frequency stability index of the k-th balanced island is obtained by calculating the rate of change of frequency (ROCOF) using the following formula. :

[0157]

[0158] Among them, frequency stability index Used to comprehensively reflect the frequency maintenance capability of an isolated network under disturbances;

[0159] , and These represent weights, set to 0.4, 0.3, and 0.3 respectively, with the sum of the weights being 1.

[0160] Frequency deviation reflects the difference between the current frequency and the rated value; frequency recovery time reflects the speed at which the frequency returns to the normal range after a disturbance; frequency change rate reflects the severity of frequency abrupt changes; frequency stability index It can comprehensively assess the frequency maintenance capability of an island under disturbances—for example, although the frequency of an island may not exceed the limit, its ROCOF exceeds 0.5Hz / s (easily triggering low-frequency load shedding protection), which traditional systems cannot identify, while... The frequency change rate will drop below 0.5 if it is too large, thus accurately marking the risk. At the same time, the introduction of frequency recovery time can distinguish between "instantaneous achievement" and "slow achievement" (such as recovery time exceeding 10 seconds, which may easily lead to load abnormalities), making the frequency stability judgment more in line with actual operating needs and avoiding industrial load shutdowns due to slow frequency recovery caused by one-sided assessment.

[0161] The sub-unit for analyzing the response capability of energy storage and gas turbines is used to obtain the regulation capability parameters of the energy storage units and gas turbines within the island, including the remaining capacity of the energy storage. Backup power and climbing ability And calculate the energy storage and gas turbine response capability indicators of the kth balanced island. :

[0162]

[0163]

[0164] in, This indicates the change in power of the gas turbine unit during the adjustment process. This represents the time required to complete the power change; if a gas turbine needs 60 seconds to ramp up from 50MW to 100MW, then its ramp rate is (100-50) / 60 = 0.83MW / s; in islanded operation, ramp rate... A larger value indicates a faster equipment adjustment response, which is more conducive to the rapid recovery of frequency and voltage. A smaller value indicates a slower system adjustment and a greater likelihood of instability under large disturbances. , and These represent the maximum values ​​of remaining energy storage capacity, reserve power, and ramping capability, respectively.

[0165] , and These represent weights, set to 0.4, 0.3, and 0.3 respectively, with the sum of the weights being 1.

[0166] Energy storage and gas turbine response capability indicators of the kth balanced island It accurately identifies the actual energy capacity levels of different isolated islands; at the same time, it converts ramping capacity and reserve power into relatively optimal values ​​and combines them with fixed weights for calculation, which solves the shortcoming of traditional systems that cannot judge whether emergency needs are met by only looking at absolute values. It can provide early warning of the risks of slow ramping adjustment and insufficient reserve power redundancy, ensuring that it accurately reflects the actual response capabilities of isolated island energy storage and gas turbines, providing reliable input for the autonomous stability coefficient, and avoiding island instability due to misjudgment of energy capacity.

[0167] Example 8

[0168] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the equipment resilience execution module includes a potential fault area identification unit and a topology dynamic reconstruction unit;

[0169] The potential fault area identification unit is used to extract real-time distribution network data transmitted by the power grid in the "disaster risk area group". When any of the following conditions are met in the j-th area, the distribution network line current exceeds the rated value by 1.2 times, the equipment temperature exceeds 85°C, or the insulation resistance is lower than 0.5MΩ, the corresponding distribution network power supply zone, including 10kV feeders, distribution transformers, and downstream load zones, is marked as a "potential fault area". Based on the distribution network GIS map, the physical boundary of the potential fault area is delineated with the "potential fault area" as the core. The boundary range covers the feeder section where the faulty equipment is located, the associated branch lines, and the core load unit supplied by it, which is denoted as the m-th core load unit.

[0170] The topology dynamic reconfiguration unit is used for each group of main intelligent agents within the m-th core load unit and its surrounding 3km range. Each group of main intelligent agents includes one set of intelligent switches and one corresponding power distribution terminal.

[0171] The first step is to send a "shutdown command" to the associated smart switches in each group of main agents based on the boundary of the potential fault area. This is used to disconnect the potential fault area from the main network and to record the action response time. The smart switches include feeder sectionalizing switches and branch switches.

[0172] The second step is path reconstruction. The main intelligent agent analyzes the surrounding distribution network topology, which includes tie switches and backup power access points. It calculates the optimal power restoration path, prioritizes the selection of "backup feeders in non-disaster-affected areas" or "grid-connected distributed energy access points" as the power source, and sends a "closing command" to the corresponding intelligent tie switch to construct a new power supply path from the power source to the m-th core load unit.

[0173] After reconstruction, the voltage, current, and power factor of the new path are collected in real time from the main agent and fed back to the main agent.

[0174] If the voltage of the new path needs to meet 0.95-1.05 times the rated voltage, the current does not exceed the rated value, and the power factor is greater than or equal to 0.9, then the above conditions are met and the topology reconstruction is marked as "successful".

[0175] In this embodiment, the potential fault area identification unit accurately marks the fault area and associated core loads based on real-time data (current, temperature, insulation resistance) of the distribution network, solving the problems of time-consuming and unclear boundaries in traditional manual inspection. Then, the topology dynamic reconstruction unit activates the main agent to quickly disconnect the fault connection (record response time), and prioritizes the construction of new paths by selecting backup feeders or distributed energy sources in non-disaster areas. Combined with data collected from the agent to verify the compliance of voltage, current, and power factor, this solves the shortcomings of traditional topology reconstruction, which relies on manual scheduling, has poor path optimization, and low recovery efficiency. This enables rapid and reliable power restoration to the core loads in the fault area, avoiding the impact on core power consumption due to fault expansion or slow recovery.

[0176] Example 9

[0177] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the equipment resilience execution module also includes a distributed energy coordination unit;

[0178] The distributed energy coordination unit is used to construct the load gap power ΔPm in the power restoration path constructed by the topology dynamic reconfiguration unit. The load gap power ΔPm is the difference between the rated power Pm of the m-th core load unit and the actual power supply power Pm of the main grid in the power restoration path. The calculation formula is ΔPm = Pm rated - Pm main grid. When ΔPm > 0, it is necessary to enter the distributed differentiated energy supplementation mode, which means that distributed energy supplementation is required. When ΔPm ≤ 0, no distributed energy intervention is required.

[0179] Distributed energy resources include: photovoltaic power plants, energy storage power plants, and gas turbines;

[0180] Distributed differentiated power replenishment modes include:

[0181] Mode 1: When ΔPm≤P storage, only the energy storage station is used to replenish the energy. The distributed energy coordination unit instructs the energy storage station to discharge at a constant power equal to ΔPm, with a response time ≤100ms, so that the actual power supply of the m-th core load unit is equal to the rated power of Pm.

[0182] Mode 2: When ΔPm > P (storage / discharge), the energy storage power station, gas turbine, and photovoltaic power station work together to replenish energy. First, the energy storage power station is instructed to store and discharge energy at its maximum dischargeable power P. Simultaneously, the gas turbine is started (start-up time ≤ 3 minutes) and its power is adjusted according to the remaining gap of ΔPm. At the same time, the photovoltaic power station is instructed to operate in maximum power point tracking (MPPT) mode. The total power supply is eventually equal to the rated Pm.

[0183] In this embodiment, the distributed energy coordination unit accurately determines the energy replenishment demand by calculating the load gap power ΔPm, avoiding the problem of "blind investment or omission of demand" in traditional energy replenishment. At the same time, it adopts a differentiated mode according to the size of ΔPm. When ΔPm is small, only the energy storage station is used for rapid energy replenishment (response ≤100ms). When the gap is large, the energy storage, gas turbine (start-up ≤3 minutes) and photovoltaic are linked to coordinate energy replenishment. This solves the shortcomings of disordered use, low energy replenishment efficiency or power mismatch of traditional distributed energy, ensuring that the actual power supply to the core load is stable and up to standard, further ensuring the reliability of power supply after power restoration, and avoiding abnormal operation of core equipment due to load gap.

[0184] Example 10

[0185] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, the equipment resilience execution module also includes a recovery speed assessment unit. This unit is used to calculate the recovery speed coefficient of the m-th core load unit based on the path construction time of the topology dynamic reconfiguration unit and the energy replenishment effect of the distributed energy collaborative unit. And complete the risk assessment, specifically including:

[0186] Data collection and recovery startup time That is, the cumulative time from the output of the mark by the potential fault area identification unit to the completion of the power restoration path construction by the topology dynamic reconfiguration unit and the realization of ΔPm=0 by the distributed energy coordination unit;

[0187] The average absolute value of the deviation rate between the actual power supply of the m-th core load unit and the rated power of Pm during the recovery process is used to obtain the load recovery stability coefficient. ;

[0188] The recovery rate coefficients of m core load units are calculated using the following formula. :

[0189]

[0190] in, This indicates the preset maximum allowable recovery time. For core medical loads, such as three hospitals, T0 = 300 seconds; for important industrial loads (such as continuous production enterprises), T0 = 600 seconds; and for urban basic loads (such as core residential areas), T0 = 900 seconds. and Indicates the weight value; And preset a fast recovery threshold Rth, when Output local recovery risk markers;

[0191] when Generate a second qualified mark.

[0192] In this embodiment, the recovery speed assessment unit collects the recovery start time (from fault identification to path construction and power replenishment completion) and the load recovery stability coefficient (average power deviation rate), and calculates the recovery speed coefficient by combining the preset maximum allowable recovery time T0 corresponding to the load type and the weight. This solves the problem that traditional recovery effects only "qualitatively determine whether power has been restored" and lack quantitative standards. At the same time, by comparing with the preset threshold Rth, it outputs risk markers or qualified markers to accurately identify cases of slow recovery speed and poor stability. This avoids the problem of "potential risks still existing after power restoration" caused by the lack of assessment in traditional methods, and further ensures the quality of core load recovery and long-term reliable operation.

[0193] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0194] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. An AI-powered multi-agent hierarchical emergency response system supporting a resilient power grid, characterized in that: include: The global situational awareness module is used to acquire multi-source disaster monitoring information from meteorological satellites, drone swarms, and tens of millions of sensors, establish a global disaster chain projection model, predict the paths of extreme disasters such as typhoons, wildfires, and floods, and construct the disaster intensity coefficient D for the j-th region. j And assess the corresponding risk level, and then map it onto the risk map; The regional autonomy control module is used to extract the first and second risk levels from the risk map, summarize them into "disaster-risk area clusters," and divide the "disaster-risk area clusters" into several balancing islands. The system drives the substation intelligent agent cluster to perform second-level proactive disconnection and activates the energy storage unit and gas turbine black start contingency plan within the island. It also collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. A preset stability threshold Sth is set when... Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. Generate the first qualified mark; The equipment resilience execution module is used to identify "potential fault areas" within the "disaster risk area group." Within these "potential fault areas," it drives the distribution network intelligent swarm to control intelligent switch groups to dynamically reconstruct the power grid topology. Combined with the autonomous regulation of distributed energy resources, it forms a rapid power restoration path and constructs a calculation for the recovery speed coefficient of the m-th core load unit. And complete the risk assessment, and preset the fast recovery threshold Rth, when Output local recovery risk markers; when Generate a second qualified mark; The device resilience execution module includes a potential fault area identification unit and a topology dynamic reconstruction unit; The potential fault area identification unit is used to extract real-time distribution network data transmitted by the power grid in the "disaster risk area group". When any of the following conditions are met in the j-th area, the distribution network line current exceeds the rated value by 1.2 times, the equipment temperature exceeds 85°C, or the insulation resistance is lower than 0.5MΩ, the corresponding distribution network power supply zone, including the 10kV feeder, distribution transformer, and downstream load zone, is marked as a "potential fault area". Based on the distribution network GIS map, the physical boundary of the potential fault area is delineated with the "potential fault area" as the core. The boundary range covers the feeder section where the faulty equipment is located, the associated branch line, and the core load unit supplied by it, and is denoted as the m-th core load unit. The topology dynamic reconfiguration unit is used to activate the m-th core load unit and each group of main intelligent agents within a 3km radius. Each group of main intelligent agents includes one set of intelligent switches and one corresponding power distribution terminal. The first step is to send a "shutdown command" to the associated smart switches in each group of main agents based on the boundary of the potential fault area. This is used to disconnect the potential fault area from the main network and to record the action response time. The smart switches include feeder sectionalizing switches and branch switches. The second step is path reconstruction. The main intelligent agent analyzes the surrounding distribution network topology, which includes tie switches and backup power access points. It calculates the optimal power restoration path, prioritizes the selection of "backup feeders in non-disaster-affected areas" or "grid-connected distributed energy access points" as the power source, and sends a "closing command" to the corresponding intelligent tie switch to construct a new power supply path from the power source to the m-th core load unit. After reconstruction, the voltage, current, and power factor of the new path are collected in real time from the main agent and fed back to the main agent. If the voltage of the new path meets 0.95-1.05 times the rated voltage, the current does not exceed the rated value, and the power factor is greater than or equal to 0.9, then the "topology reconstruction successful" is marked. The device resilience execution module also includes a distributed energy coordination unit; The distributed energy coordination unit is used to construct a load gap power ΔPm in the power restoration path constructed by the topology dynamic reconfiguration unit. The load gap power ΔPm is the difference between the rated power Pm of the m-th core load unit and the actual power supply power Pm of the main grid in the power restoration path. The calculation formula is ΔPm = Pm rated - Pm main grid. When ΔPm > 0, it is necessary to enter the distributed differentiated energy replenishment mode, which means that distributed energy replenishment is required. When ΔPm ≤ 0, no distributed energy intervention is required. Distributed energy resources include: photovoltaic power plants, energy storage power plants, and gas turbines; The distributed differentiated power replenishment mode includes: Mode 1: When ΔPm≤P storage, only the energy storage station is used to replenish the energy. The distributed energy coordination unit instructs the energy storage station to discharge at a constant power equal to ΔPm, with a response time ≤100ms, so that the actual power supply of the m-th core load unit is equal to the rated power of Pm. Mode 2: When ΔPm > P (storage / discharge), the energy storage power station, gas turbine, and photovoltaic power station work together to replenish energy. First, the energy storage power station is instructed to store and discharge energy with the maximum dischargeable power P. At the same time, the gas turbine is started to adjust its power according to the remaining gap of ΔPm. Meanwhile, the photovoltaic power station is instructed to operate in maximum power point tracking mode. The total power supply is eventually equal to the rated Pm. The equipment resilience execution module also includes a recovery speed evaluation unit, which is used to calculate the recovery speed coefficient of the m-th core load unit based on the path construction time of the topology dynamic reconfiguration unit and the energy replenishment effect of the distributed energy collaborative unit. And complete the risk assessment, specifically including: Data collection and recovery startup time That is, the cumulative time from the output of the mark by the potential fault area identification unit to the completion of the power restoration path construction by the topology dynamic reconfiguration unit and the realization of ΔPm=0 by the distributed energy coordination unit; The average absolute value of the deviation rate between the actual power supply of the m-th core load unit and the rated power of Pm during the data recovery process is used to obtain the load recovery stability coefficient. ; Calculate and obtain the recovery rate coefficients of m core load units. And preset a fast recovery threshold Rth, when Output local recovery risk markers; when Generate a second qualified mark.

2. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 1, characterized in that, The global situational awareness module includes an unmanned aerial vehicle swarm monitoring unit and a disaster chain simulation unit; The drone swarm monitoring unit is used to collect multi-source disaster monitoring information for the j-th region using drones equipped with wind speed sensors, air pressure sensors, humidity sensors, rainfall sensors, and infrared thermometers. The multi-source disaster monitoring information includes: the air pressure of the j-th region. Ignition temperature Fuel humidity Topographic slope Combustible material energy density Rainfall intensity and water depth Furthermore, by using GPS positioning and time synchronization technology, the monitoring data of each drone is bound to regional coordinates to establish a drone monitoring dataset with spatiotemporal labels. ； in, The longitude and latitude coordinates of the center point of the region. The timestamp for data collection is in the format "year-month-day hour:minute:second / millisecond", generated by the drone's built-in clock in conjunction with satellite time synchronization.

3. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 2, characterized in that, The disaster chain simulation unit is used to establish a global disaster chain simulation model based on the meteorological disaster dynamics model, and to dynamically simulate the propagation path and intensity of typhoons, wildfires, and floods. The disaster chain simulation unit includes a typhoon propagation simulation subunit, a wildfire spread simulation subunit, and a flood diffusion simulation subunit; The typhoon propagation simulation subunit is used to calculate the typhoon propagation path and intensity in the j-th region, specifically including: Using the pressure field gradient formula Calculate the pressure gradient vector of the j-th region. ; It represents the rate of change of air pressure along the east-west direction, and the speed of change of air pressure from west to east; It represents the rate of change of air pressure along the north-south direction, and the speed of change of air pressure from south to north; By combining the geostrophic wind relationship, the geostrophic wind speed in the j-th region is obtained. ; Extract the geostrophic wind speed of the j-th region Pressure gradient vector and rainfall intensity Calculate and obtain the typhoon intensity coefficient for the j-th region. .

4. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 3, characterized in that, The wildfire spread simulation subunit is used to extract the fire point temperature. Fuel humidity and terrain slope The relationship is used to calculate the base spread rate of the j-th region. ; pressure gradient vector Used as a local wind field, for the rate of base spread Make corrections to obtain the corrected spread rate. ; Extract the corrected spread rate Combustible material energy density and rainfall intensity Calculate the wildfire intensity coefficient for the j-th region. .

5. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 3, characterized in that, The flood diffusion simulation subunit is used to calculate the evolution of flood water depth and flood intensity coefficient in the j-th region. : The water depth of the j-th region is determined according to the shallow water equation. The dynamic evolution, namely the evolution of runoff and water accumulation, is as follows: in, Let be the water depth in the j-th region as a function of time t and spatial coordinates (x, y). This represents the rainfall intensity in the j-th region over time t. This represents the penetration loss in the j-th region; and These are the flow rates along the x and y directions, respectively, calculated using Manning's formula: in, and Let x and y represent the slope components of the terrain in the j-th region, respectively. This represents the Manning roughness coefficient, used to describe surface resistance, with a range of 0.02–0.

08. Based on the evolved water depth and flow rate, and combined with rainfall intensity, the flood intensity coefficient of the j-th region is calculated. ; Combined with the typhoon intensity coefficient of the j-th region Wildfire intensity coefficient and flood intensity coefficient The overall disaster intensity of the j-th region is obtained by correlation. And assess the corresponding risk level, including: 0.7 < D j ≤1, generate the first risk level, and use red for gradient coding; 0.4 < D j If the risk level is ≤7, a second risk level is generated, and gradient coding is performed using orange. 0≤D j If the value is ≤0.4, a third risk level is generated, and gradient coding is performed using green. It also utilizes a GIS rendering engine to generate vectorized dynamic maps of the corresponding risk levels, disaster impact ranges, and evolution trends of all grid areas. The map is updated every 15 minutes, and the latest monitoring data and disaster chain projection results are mapped onto the risk map in real time.

6. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 1, characterized in that, The regional autonomous control module includes a risk area acquisition unit, an island division unit, a contingency plan response analysis unit, and a stability determination unit. The risk area acquisition unit is used to extract the first risk level and the second risk level from the risk level of each j-th region based on the risk map, and summarize them into a "disaster risk area group"; The island division unit is used to divide the "disaster risk area group" into several balanced islands. Each isolated island consists of multiple spatially adjacent risk areas, and is controlled by the contingency response analysis unit, which invokes each balanced isolated island. The substation intelligent agent cluster in the system performs second-level proactive disconnection operations, quickly isolating the island from the main grid. After automatically activating the energy storage and gas turbine black start contingency plan, it collects voltage, frequency, and emergency energy response capabilities within the island to construct the autonomous stability coefficient of the k-th balanced island. ; The stability determination unit is used to preset a stability threshold Sth, when Output an island instability flag, indicating that the balanced island is at risk of power collapse or instability. The first qualified marker is generated, indicating that the island is in a qualified and stable operating state.

7. The AI ​​multi-agent hierarchical emergency response system for supporting resilient power grids according to claim 6, characterized in that, The contingency response analysis unit includes a voltage stability index analysis subunit, a frequency stability index analysis subunit, and an energy storage and gas turbine response capability index analysis subunit. The voltage stability index analysis subunit is used to collect the voltage operating status of each bus in the island and extract the voltage deviation. Voltage over-limit rate and voltage fluctuation rate Calculate and obtain voltage stability index ; The frequency stability index analysis subunit is used to monitor frequency characteristics within the island and extract frequency deviations. Frequency recovery time The frequency stability index is obtained by calculating the rate of change of frequency (ROCOF). ; The energy storage and gas turbine response capability index analysis subunit is used to obtain the regulation capability parameters of the energy storage unit and gas turbine within the island, including the remaining energy storage capacity. Backup power and climbing ability Calculate and obtain the response capability indicators of energy storage and gas turbine. .

Images